University of Wollongong Research Online

University of Wollongong Thesis Collection University of Wollongong Thesis Collections

2012 Phonetic variation in Gurindji Kriol and northern Australian English: a longitudinal study of fricatives in maternal speech Heather Buchan University of Wollongong

Recommended Citation Buchan, Heather, Phonetic variation in Gurindji Kriol and northern Australian English: a longitudinal study of fricatives in maternal speech, thesis, Faculty of Education, University of Wollongong, 2012. http://ro.uow.edu.au/theses/3789

Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library: [email protected]

Phonetic Variation in Gurindji Kriol and Northern Australian English: A

Longitudinal Study of Fricatives in Maternal Speech

A thesis submitted in fulfilment of the requirements for the award of the degree

Doctor of Philosophy

from

University of Wollongong

by

Heather Buchan, BPsych (Hons)

Faculty of Education

2012

iii

Abstract

Acquiring the native language involves learning to perceive and produce sound structures (phonology) of the speech input. Speech to children often contains phonological modifications, and across languages the speech input to which children are exposed contains phonetic variation associated with linguistic and socioindexical information. Previous studies have investigated phonetic variation in infant-directed speech (e.g. Davis & Lindblom, 2001; Fernald et al., 1989; Kuhl et al., 1997; Snow,

1977), and differences in phonological variation between child-directed and adult- directed speech (e.g. Bernstein-Ratner, 1984a, 1984b; Lee, Davis & MacNeilage,

2008; Lee & Davis, 2010). There are, however, few studies examining how and when phonetic variation in speech to children changes after infancy, as they get older. This thesis is a longitudinal investigation of phonetic variation in mothers’ speech to children (maternal speech) in two language varieties spoken in northern Australia,

Gurindji Kriol and Australian English.

Gurindji Kriol is an Australian Aboriginal mixed language that contains lexical forms and phonology from both the traditional language Gurindji and Kriol, an

English-lexifier creole. There are no previous systematic quantitative studies on fricatives in Gurindji Kriol, or any Australian contact language, although impressionistically fricatives in Kriol-derived words are highly variable. Australian

English also contains phonetic variation related to speech processes, such as consonant reduction in casual speech. There is, however, little prior research on phonetic variation that children are exposed to in Australian English. The purpose of this thesis was to provide quantitative analyses of fricative variation in maternal speech in Gurindji Kriol and northern Australian English as children aged from approximately 1;6 to 2;0 to 2;6. A subsidiary aim was to examine methodological
iv issues in phonetic transcription when the transcribers are non-native speakers of the language.

This thesis contains three empirical research studies. In Study 1 we added phonetic transcription and analysis to a subsample of a corpus of naturalistic family interactions in Gurindji Kriol created by Felicity Meakins between 2003 and 2007 as part of the Aboriginal Child Language project (ACLA-1, Simpson & Wigglesworth,

2008). Speakers were three Gurindji Kriol speaking women recorded at three timepoints, when the focus children were approximately 1;6, 2;0 and 2;6. We analysed stop-fricative variation in tokens of Kriol-derived words that could potentially be pronounced with a fricative, based on words that had fricatives in their

English cognates. Results showed that words containing stop-fricative variation were more likely to be open-class than closed-class, with the variable segment most frequently word-initial and at labio-dental and alveolar places of articulation. Across all tokens with potential fricatives, the likelihood of fricatives in word-initial position significantly increased when children were 2;6. In tokens of words found to contain stop-fricative variation, word-medial fricatives were significantly more likely in mothers’ speech at child age 2;6. Analyses took into account phonological environments and interspeaker differences.

Study 2 investigated fricative variation in the form of phonological reduction in northern Australian English maternal speech. A longitudinal audiovisual corpus of naturalistic family interactions was recorded and phonetically transcribed for this study. Speakers were five mothers who were native speakers of Australian English, and were recorded when their children were approximately 1;6, 2;0 and 2;6. We analysed deletion in word-initial /h/ and word-final /v/, common processes in casual speech, in mothers’ speech at each timepoint. Results showed a non-linear change in
v
overall deletion over time within a stable set of lexical items. Between child ages 1;6 and 2;0 deletion proportionately increased in mothers’ speech, while between 2;0 and

2;6 deletion proportionately decreased.

Study 3 addressed a methodological issue of checking phonetic transcription with native speakers of Gurindji Kriol. Native speakers have implicit knowledge of

Gurindji Kriol phonology that would be beneficial to understanding transcription ambiguities arising from perceptual bias, and for furthering our interpretation of phonetic variation in Gurindji Kriol. We used visual analogue scales to elicit native speaker perceptions on sound segments potentially pronounced as fricatives in

Gurindji Kriol. Native speaker judgements on the scales were then compared to the

IPA judgements made by non-native speaker phonetic transcribers. Results showed both agreements and discrepancies between native and non-native speakers for different types of judgements and segment word positions.

Empirical findings in this thesis are discussed in terms of processes driving change in phonetic variation in mothers’ speech to children, such as fine-tuning to children’s own receptive and productive language development. Results have implications for theoretical models of children’s phonological acquisition, which must take into account variable phonetic detail in the input. Pedagogical implications are also discussed (in Chapter 7) in terms of how teachers can use information about

Gurindji children’s home language to augment their language and literacy teaching.

vi

Acknowledgements

I would like to acknowledge and offer my heartfelt thanks to the many individuals who have made this thesis possible and my PhD experience one that I will cherish.

I wish to thank my primary supervisor, Dr Caroline Jones, for the many opportunities she has afforded me and for her endless support and patience. Caroline encouraged me to pursue my own research ideas while also contributing her extensive knowledge and expertise, and I thank her for her contribution to this thesis and to my growth as a researcher. I am also thankful to my cosupervisor Dr Amanda Baker, whose involvement and insight helped to develop my ideas and my writing. I am very grateful to my associate supervisor Dr Felicity Meakins for contributing her substantial breadth of knowledge of Gurindji Kriol and linguistics in general. I also owe thanks to Felicity for sharing with us her corpus of Gurindji Kriol and allowing us to build on her extensive previous work in this project. I would also like to acknowledge Pete Ball at the University of Tasmania, whose knowledge and enthusiasm for psycholinguistics started my initial interest in the field.

I would like to acknowledge the Aboriginal Child Language project (ACLA-

1), which was the basis of Felicity’s work on the Gurindji Kriol corpus, and I am thankful to the ACLA PIs Jane Simpson and Gillian Wigglesworth for supporting the research in this thesis. I also acknowledge the Australian Research Council for financial assistance with my stipend, field work and conference trips (ARC Discovery

Grant DP0985395 “Phonological development among child speakers of mixed language”, 2009-2012, C.I. Caroline Jones) and I thank Caroline for supporting my research through this grant.
vii

This research could not exist without the many people who assisted in my fieldwork, especially the speakers and their families, community leaders and organisations, and community Research Assistants in Katherine and Kalkaringi. I thank Gurindji Kriol speakers Cecelia Edwards, Anne-Maree Reynolds and Samantha

Smiler for giving us permission to use the recordings of their families that Felicity

Meakins had made. I am also incredibly grateful to the community research assistants in Kalkaringi: Trisha Morris, Kirsty Smiler, Jessica Vincent, Samantha Smiler, Anne-

Maree Reynolds, Lisa Smiler, Leanne Smiler, and Rosemary Johnson. They taught me a lot, both for the research and on a personal level, and I treasure the friendships we have made.

I also thank traditional owners in Kalkaringi for consulting with us on the project; their advice and direction was invaluable. I am very grateful to Leah Leaman who worked as a liaison and participant advocate, and provided guidance in my community visits. I am very appreciative of our evening chats in the garden and

Leah’s extensive knowledge of the community and her readiness to share this with me. I wish to thank all the people in Kalkaringi and Daguragu who made my trips there interesting, informative and fun in many different ways.

I am very grateful to the Katherine mothers Cathy, Eva, Kellie, Kim and Erin

(pseudonyms used at participants’ discretion) and their children and families, who let me into their homes and their lives. They always greeted me with a smile and were patient when I had to fiddle around with equipment. It was truly a pleasure to work with them and I feel privileged to have watched their kids grow. I am also grateful to all the organisations and businesses in Katherine who helped me to set up the project and find participants. I wish to particularly thank the childcare services Good
viii

Beginnings and Little Mangoes. Also I thank the linguists and others in Katherine who befriended me and offered much-appreciated advice, dinner and coffee.

I would like to thank the project steering committee at Diwurruwurru Jaru (the

Katherine Language Centre) for their advice and recommendations and for letting me sit and work in the Language Centre library in my first trip to Katherine. I particularly thank Ruth Joshua for teaching about Kriol and kinship systems. I am also grateful to

Cerise King for her voluntary cultural mentoring. Cerises’s insight and advice was always helpful.

I deeply appreciate the contributions of the Research Assistants who have worked on the project in various capacities: Shujau Muawiyath, Colleen Moerkerken,

Kim Cayzer, Rachel Groves, Sarah Cutfield, and Alison Hannah. I also thank Yvan

Rose and the team at Memorial University for their support with the phonetic transcription and analysis program Phon and for always providing helpful (and patient) answers to my questions.

My Building 23 (and 22) buddies have made this whole process fun and supportive, and I feel so lucky to have formed many lasting friendships. Thank you to

Rosie Welch and a certain energy drink, Jonnell, Narumi, Sophie, Charles, Kyle,

Ashley Sisco for the chocolate activities, prime coffee buddy Alex Miller, Sam

McMahon for feeding me and her special contributions to the office stationary, and

Kay Prcevich who has been an amazing officemate for the entire time.

Finally, I wish to thank all my friends and family who have supported me along the way. I especially thank Flick for being such a great friend, Alissa for the cloudy apple nights, MV for the mojo, and Yan for her optimism and friendship. My parents Rosemary and Grant Buchan, my sisters, grandparents and other family members have always been supportive and I give them my heartfelt thanks.
ix

Table of Contents

Abstract...... iii Acknowledgements...... vi Table of Contents ...... ix List of Tables...... xiii List of Figures...... xv Publications...... xvii 1 Introduction...... 1 1.1 Gurindji Kriol...... 3 1.2 Australian English...... 6 1.3 Approach to Variation in this Thesis...... 8 1.4 Structure of the Thesis...... 10 1.5 Research Questions ...... 11 2 Empirical and Theoretical Context of Children’s Phonological Acquisition from Speech Input...... 13 2.1 Literature Review...... 14 2.1.1 Child-Directed Speech ...... 14 2.1.2 Sources of Speech Input to Children...... 19 2.1.3 Phonological Development After Infancy ...... 21 2.1.4 Summary of Literature ...... 25 2.2 Thesis Aims and Rationale ...... 26 2.3 Research Questions ...... 31 2.3.1 Study 1 (Chapter 4) ...... 31 2.3.2 Study 2 (Chapter 5) ...... 31 2.3.3 Study 3 (Chapter 6) ...... 32 2.4 Significance...... 32 2.5 Theoretical Framework ...... 33 2.6 Summaries of Research Chapters...... 41 2.6.1 Chapter 4 ...... 41 2.6.2 Chapter 5 ...... 43 2.6.3 Chapter 6 ...... 44 2.7 Chapter Summary...... 45 3 Methodology ...... 47
x

3.1 Timeline of Data Collection...... 48 3.2 Gurindji Kriol Corpus...... 49 3.2.1 Design...... 49 3.2.2 Participants...... 52 3.2.3 Procedure...... 54 3.2.4 Phonetic Transcription ...... 54 3.2.5 Ethical Considerations Relating to the Gurindji Kriol Corpus ...... 55 3.3 Katherine English Corpus...... 56 3.3.1 Design...... 56 3.3.2 Participants...... 56 3.3.3 Procedure...... 60 3.3.4 Ethical Considerations Relating to the Katherine English Corpus ...... 62 3.3.5 Phonetic Transcription ...... 63 3.4 Working with Gurindji Community Members ...... 64 3.5 Phonetic Transcription and Analysis in Phon ...... 65 3.5.1 Overview...... 65 3.5.2 Getting Started ...... 68 3.5.3 Media...... 70 3.5.4 Transcription...... 71 3.5.5 Search...... 73 3.5.6 Conclusion...... 75 3.6 Chapter Summary...... 76 4 Fricative Variation in Maternal Gurindji Kriol ...... 78 4.1 Abstract...... 79 4.2 Background...... 81 4.2.1 Gurindji Kriol...... 83 4.2.2 Phonetic Variation in Child-Directed Speech...... 85 4.2.3 Lenition Processes and Fricative Variation...... 87 4.2.4 The Current Study...... 89 4.3 Method...... 90 4.3.1 Design...... 90 4.3.2 Speakers...... 91 4.3.3 Recordings...... 91 4.3.4 Analysis...... 91
xi

4.3.5 Coding...... 92 4.3.6 Transcription Reliability ...... 93 4.4 Results...... 95 4.4.1 Analysis 1: Fricative Variation Within Gurindji Kriol Lexical Forms. 95 4.4.2 Analysis 2: Relationships Between Fricatives in English Cognates and Pronunciation in Gurindji Kriol Words that Contain Stop-Fricative Variation ...... 100 4.4.3 Analysis 3: Factors affecting fricative variation in maternal Gurindji Kriol...... 108 4.4.4 Analysis 4: Effect of Child Age...... 117 4.5 Discussion...... 118 5 Fricative Reduction in Maternal Northern Australian English ...... 125 5.1 Abstract...... 126 5.2 Background...... 127 5.2.1 Phonetic Variation in Australian English...... 133 5.2.2 Katherine English...... 136 5.2.3 The Current Study...... 137 5.3 Method...... 138 5.3.1 Participants...... 138 5.3.2 Procedure...... 140 5.3.3 Analysis...... 144 5.4 Results...... 144 5.4.1 Analysis 1: Incidence of Word-Initial /h/ and Word-Final /v/ Deletion in Maternal Speech Over Time...... 144 5.4.2 Analysis 2: Exploration of Local Effects on Casual Speech Processes...... 154 5.4.3 Analysis 3: Effect of Speech Rate...... 159 5.5 General Discussion ...... 160 6 Investigating Native Speaker Judgements of Phonetic Variation in Gurindji Kriol Using Visual Analogue Scales...... 169 6.1 Abstract...... 170 6.2 Introduction...... 171 6.3 Background...... 172
xii

6.3.1 Measuring Speech with Visual Analogue Scales...... 172 6.3.2 Limitations of Phonetic Transcription ...... 174 6.3.3 Language Background Affects Phonetic Transcription ...... 176 6.3.4 Current Applied Context...... 178 6.4 Method...... 181 6.4.1 Non-Native Speaker IPA Transcriptions ...... 181 6.4.2 Visual Analogue Scale Development...... 182 6.4.3 Phonological Awareness Training ...... 187 6.4.4 Visual Analogue Scale Administration...... 188 6.5 Results...... 189 6.5.1 Word-Initial...... 190 6.5.2 Word-Medial...... 192 6.5.3 Word-Final...... 193 6.6 Discussion...... 198 6.6.1 Effects of Language Background...... 200 6.6.2 Word Position Effects ...... 204 6.6.3 Limitations...... 206 6.6.4 Conclusion...... 207 7 Conclusion ...... 209 7.1 Core Findings...... 211 7.2 Implications...... 214 7.3 Limitations...... 217 7.4 Directions for Future Research...... 219 7.5 Conclusion...... 222 References...... 223 Appendix A: Participant Information and Consent Forms...... 245 Appendix B: Gurindji Kriol Words Containing Potential Fricatives ....261
xiii

List of Tables

Table 3.1 Gurindji Kriol speaking mothers and age of the focus children...... 52

Table 3.2 Total amounts of recording and number of mothers’ utterances in the

Gurindji Kriol subsample at each stage...... 53

Table 3.3 Characteristics of Katherine English speaking mothers and the focus

children ...... 57

Table 3.4 Total amounts of recording and number of mothers’ utterances in the

Katherine English corpus at each stage ...... 59

Table 4.1 Number and percentage of feature agreement on ‘potential fricative’

segments between two transcribers ...... 94

Table 4.2 Frequencies of word-initial fricatives in English cognates and

corresponding phones in Gurindji Kriol variable words ...... 104

Table 4.3 Frequencies of word-medial fricatives in English cognates and

corresponding phones in Gurindji Kriol variable words ...... 105

Table 4.4 Frequencies of word-final fricatives in English cognates and

corresponding phones in Gurindji Kriol variable words ...... 106

Table 4.5 Logistic regression results for word-initial fricative variation in

all words...... 109

Table 4.6 Logistic regression results for word-initial fricative variation in

variable words...... 110

Table 4.7 Logistic regression results for word-medial fricative variation in

all words...... 111

Table 4.8 Logistic regression results for word-medial fricative variation in

variable words...... 112

xiv

Table 4.9 Logistic regression results for word-final fricative variation in

all words...... 113

Table 4.10 Logistic regression results for word-final fricative variation in

variable words...... 114

Table 5.1 Participant characteristics...... 140

Table 5.2 Total and mean lengths of recording sessions, hh:mm:ss...... 142

Table 5.3 Raw frequencies of word-initial /h/ deletion out of possible contexts by

each speaker at each average child age...... 145

Table 5.4 Logistic regression for word-initial /h/ deletion: Child age and

speaker ...... 147

Table 5.5 Raw frequencies of word-final /v/ deletion out of possible contexts by

each speaker at each average child age...... 149

Table 5.6 Logistic regression for word-final /v/ deletion: child age and

speaker ...... 151

Table 5.7 Lexical items containing word-initial /h/ and word-final /v/ deletion at

each stage...... 156

Table 5.8 /h/ deletion in male and female pronouns by mothers of boys (n=2) and

girls (n=2) at each child age...... 157

Table 6.1 Proposed Gurindji Kriol consonant inventory...... 180

Table 6.2 Number of each type of judgements in the Visual Analogue Scales

(N = 83)...... 183

Table 6.3 Word-initial judgement type frequencies ...... 184

Table 6.4 Word-medial judgement type frequencies...... 185

Table 6.5 Word-final judgement type frequencies ...... 185

xv

List of Figures

Figure 1.1 Map of the Victoria River District and its communities

(Meakins, 2011)...... 8

Figure 3.1 Stages of fieldwork and timeframes of data collection ...... 51

Figure 3.2 Photo of a typical recording session ...... 61

Figure 3.3 Example of a customised session editor display in Phon ...... 69

Figure 3.4 Using the segmentation tool to create a record that is linked to the

Media ...... 71

Figure 3.5 Example of Phon’s syllabification and alignment feature...... 73

Figure 4.1 Proportions of closed-class and open-class words within lexical form

categories ...... 96

Figure 4.2 Proportions of word positions of potential fricatives within lexical form

categories ...... 97

Figure 4.3 Proportions of places of articulation of fricatives in English cognates

within Gurindji Kriol lexical form categories ...... 99

Figure 4.4 Relationships between fricatives in English cognates and corresponding

pronunciation in Gurindji Kriol words...... 107

Figure 4.5 Proportions of 'potential fricatives' pronounced as fricatives across child

ages by each speaker in all tokens ...... 115

Figure 4.6 Proportions of word categories at each child age ...... 116

Figure 4.7 Proportions of 'potential fricatives' pronounced as fricatives across child

ages in variable words, other words, and the total sample ...... 117

Figure 5.1 Spectrogram of utterance "There we go, we'll just leave him there,

Shall we?"...... 142
xvi

Figure 5.2 Percentage word-initial /h/ deletion out of word-initial /h/ contexts by

child age and speaker...... 146

Figure 5.3 Percentage word-final /v/ deletion out of word-final /v/ contexts by child

age and speaker...... 150

Figure 5.4 Most frequent lexical items containing word-initial /h/ deletion at each

timepoint...... 155

Figure 6.1 Example Visual Analogue Scale for [b] and [v]...... 187

Figure 6.2 Word-initial native speaker judgements (box-and-whisker plot) and

transcriber judgements (circles)...... 195

Figure 6.3 Word-medial native speaker judgements (box-and-whisker plot) and

transcriber judgements (circles)...... 196

Figure 6.4 Word-final native speaker judgements (box-and-whisker plot) and

transcriber judgements (circles)...... 197 xvii

Publications

This thesis, Phonetic variation in Gurindji Kriol and northern Australian English: A longitudinal study of fricatives in maternal speech, includes three chapters, and one part of one chapter, that have been written as the following four journal articles. Two of these articles have been submitted to peer-reviewed journals. Of these, one of the articles has been published, and one is under review:

Chapter 3, Section 3.5 Buchan, H. (2011). Phon: Free software for phonological transcription and analysis. Language Documentation and Conservation, 5, 81-87.

Chapter 4 Buchan, H., Jones, C., & Meakins, F. (2012). Fricative variation in maternal Gurindji Kriol. Manuscript in preparation. To be submitted to Australian Journal of Linguistics.

Chapter 5 Buchan, H., & Jones, C. (2013). Phonological reduction in maternal speech in northern Australian English: Change over time. Journal of Child Language. doi: 10.1017/S0305000913000123

Chapter 6 Buchan, H., Jones, C., & Meakins, F. (2012). Investigating native speaker judgements of fricative variation in Gurindji Kriol using visual analogue scales. Manuscript in preparation. To be submitted to Language Resources and Evaluation.

As the primary supervisor, I, Dr Caroline Jones, declare that the greater part of the work in each article listed above is that of the candidate, Heather Buchan. In each of the co- authored manuscripts above, Heather contributed to the study design, was primarily responsible for data analysis and data interpretation, and led the writing.

In terms of data collection and transcription, Heather was primarily responsible for data collection for Chapters 5 (English speech corpus) and 6 (Perceptual judgements). The thesis also draws on a subset of a pre-existing corpus of Gurindji Kriol, recorded and orthographically transcribed by Dr Felicity Meakins (2003-07, as part of the Aboriginal Child Language project through University of Melbourne, funded by Australian Research Council DP0343189). Transcripts from the pre-existing Meakins corpus were converted to Phon format by Caroline and Heather. Phonetic transcription was added to specific words in the transcripts by Heather and a casual research assistant, with occasional assistance from Caroline. This formed the data for Chapter 4 and the materials for the perceptual study in Chapter 6. Two casual research assistants phonetically transcribed 10% of the data in Chapters 4 and 5 as a reliability check. Gurindji Research Assistants contributed to data collection for Chapter 6 during Heather’s visits to Kalkaringi, Northern Territory.

Regarding data analysis and writing, Heather performed all the data analyses including statistical analyses. Heather wrote the first draft of each manuscript and was then responsible for responding to the suggestions of her co-authors in subsequent versions of each manuscript. The co-authors, Dr Caroline Jones (Chapters 4-6) and Dr Felicity Meakins (Chapters 4 and 6), assisted in literature review, study design, data analysis,

CHAPTER ONE

Introduction

CHAPTER 1 2

Learning the sound structure of the home language from speech input is a critical process in children’s language acquisition. The input provides children with language-specific information about the systematic organisation of speech sounds, including the frequency distributions of sounds, probabilities of sound sequences within words and across word boundaries, and phonetic variability in sounds and words. The importance of sound patterns in the input to children’s phonological development has long been recognised in the literature, with systematic studies on the phonology of caregivers’ speech to children (infant- or child-directed speech) dating back decades

(see Snow & Ferguson, 1977). Since this early research on speech input, our understanding of the amount and types of information that children encode from the input has developed. Current acquisition theories recognise the critical role of the input and posit mechanisms for how detailed phonetic features of input can shape children’s phonological representations. However, while it has been well established that there are phonological differences in adult speech directed to children and to adults, surprisingly little research has been done on how and when phonological variability in the input changes as children get older – that is, as child-directed speech develops into adult- directed speech. The overall aim of this thesis is to investigate how phonological variation in the input to children changes from later infancy into early childhood. The data used in these studies were primarily speech recordings transcribed phonetically using the IPA (International Phonetic Alphabet), and a subsidiary aim of this thesis was to explore a methodology for checking phonetic transcription with native speakers of

Gurindji Kriol.

This thesis is a longitudinal investigation of phonetic variation in fricatives in the input children are exposed to in two language varieties of northern Australia,

Gurindji Kriol and northern Australian English. This Introduction chapter provides brief CHAPTER 1 3

essential background to the three empirical studies that comprise the thesis. The first and second sections set out the context of previous research into Gurindji Kriol and

Australian English, respectively. The third section, ‘Approach to variation in this thesis’, defines key terms and describes the analytical approach taken here. In the closing section of Chapter 1, I lay out the structure of the thesis and present the overarching research questions.

1.1 Gurindji Kriol

Gurindji Kriol is the main language of the Gurindji people in northern Australia.

It is a contact language variety that systematically combines lexical items and grammatical structures of two languages, Gurindji and Kriol (Meakins, 2011, pp. 2, 12-

13). Gurindji is an endangered traditional Australian Aboriginal language spoken mainly by older generations of Gurindji people at Kalkaringi and Daguragu in the

Victoria River District in the Northern Territory, Australia (see map in Figure 1.1).

Kriol is an English-lexifier creole, that is it is a language variety in which most of the word forms are derived from English although the grammar and semantics reflect regional Aboriginal languages (Hudson, 1985). The grammar of Gurindji Kriol has been studied in detail by Meakins (2009, 2010, 2011). Prior to this project there are no detailed accounts of the phonetics and phonology of Gurindji Kriol, though the phonological inventory has been described briefly (Meakins, in press; Meakins, 2007).

This thesis is part of a larger project on the phonology of Gurindji Kriol1 (Jones, the CI, is the primary supervisor of this thesis). Studies in the larger project have examined

Gurindji Kriol’s vowel system (Jones, Meakins & Buchan, 2011; Jones, Meakins &

1 Australian Research Council Discovery Grant DP0985395, Phonological development in child speakers of mixed language (CI Caroline Jones, 2009-2012) CHAPTER 1 4

Muawiyath, 2012), voicing variation in stop consonants (Jones & Meakins, 2012a), and the phonetic characteristics of baby-talk (Jones & Meakins, 2012b).

The Gurindji Kriol studies in the current thesis use data from a longitudinal audiovisual corpus of naturalistic family interactions mainly in Gurindji Kriol recorded and orthographically transcribed by Felicity Meakins between 2003 and 2007 as part of the Aboriginal Child Language (ACLA-1) project (Simpson & Wigglesworth, 2008;

Meakins, 2011). For the current research, we built on this corpus to add phonetic transcription and analysis to a subsample of recordings and transcripts; in this thesis I added phonetic transcription to Kriol-based words whose English cognates contain fricatives.

Gurindji Kriol is the home language of children in Kalkaringi community

(Meakins, 2011, p. 42). The speech community is multilingual – children’s home language is Gurindji Kriol, but they are also exposed to traditional Gurindji from older speakers and to Kriol, which is spoken with Kriol-speaking visitors. The neighbouring

Aboriginal language Warlpiri is also spoken by some community members and visitors.

Children are taught in English at the school, but they do not generally hear English in the preschool years, except in fun or role-playing (e.g., doctors and nurses). Children are thus exposed to some code-switching into English, as well as code-switching into other language varieties, chiefly Kriol, Gurindji, and Warlpiri, depending on their family and connections. See Meakins (2008; 2011, pp. 58-70) for further detail on the language situation in the community.

The phonology of Gurindji Kriol is of both theoretical and educational interest.

Gurindji Kriol is a mixed language that contains a structural split between the noun phrase system, supplied mostly by Gurindji, and the verb phrase system, supplied mostly by Kriol (Meakins, 2011, pp. 12-13). Importantly, both Gurindji and Kriol make CHAPTER 1 5

substantial lexical – and thus phonological – contributions. Gurindji Kriol therefore contains phonology from both Gurindji and Kriol. Due to the originally English derivation of most Kriol words, Gurindji Kriol also has some phonological similarities with English. In addition, Gurindji Kriol phonology is fundamentally characterised by the high phonological variability in Kriol (Hudson, 1985; Sandefur, 1979; Sandefur,

1991). In Sandefur’s (1979) description of Kriol spoken in the Roper River region (to the east of the current study sites) he suggested that the high variation in Kriol phonology is due to the way in which it changed over time. When speakers of

Aboriginal languages first came into contact with English, English words were pronounced with the phonological structures of Aboriginal languages. According to this view, for example, fricatives and affricates were pronounced as stops and consonant clusters were reduced. (Traditional Aboriginal languages of the region do not have phonemic fricatives, and have restricted consonant cluster types, especially in syllable- onset position.) Once Kriol developed, speakers continued to have contact with English and traditional Aboriginal languages, and words were pronounced with some English phonology in addition to the phonology from traditional languages (Sandefur, 1979).

Sandefur (1979, 1991) suggests that this resulted in the high phonological variability in

Kriol, where many words have several alternate phonetic forms.

From a language acquisition perspective, the variation in Australian Aboriginal contact language varieties opens up research questions about the amount and type of variability that children are exposed to in Gurindji Kriol and the phonological structures that they are learning from their home language in the preschool years. From an educational viewpoint, systematic descriptions of the phonological variation in Gurindji

Kriol and the sound structures children learn in their early years can provide information to assist formal teaching of both Aboriginal languages and English. For CHAPTER 1 6

example, sounds that are rare in Gurindji Kriol, such as some English fricatives, may require explicit teaching, while sounds that are the same or similar in both language varieties are generally easier to learn and can be used as a basis for teaching the harder sounds.

Phonological variation occurs in all languages to some extent. For example pronunciation of sounds and words can vary according to processes such as coarticulation, gestural overlap, and lenition or fortition in particular phonological environments, as well as speaker variables and the situational and stylistic context. It is therefore useful to investigate phonetic variation in a more documented language variety such as Australian English, as a reference point and to further our interpretation of the variation that occurs in Gurindji Kriol. This thesis includes research on a type of phonetic variation in a variety of Australian English from the town Katherine, Northern

Territory, which is in the same geographic region as Kalkaringi and is the major regional service centre (Figure 1.1).

1.2 Australian English

The phonological system of standard Australian English has been previously documented in detail (e.g. Cox, 2006; Cox, 2012; Cox & Palethorpe, 2007; Harrington,

Cox & Evans, 1997; Mitchell & Delbridge, 1965; Tollfree, 2001; Yallop, 2003). As well as differing from other varieties of English in some lexical items, the phonology differs in the pronunciation of vowels and in some prosodic patterns (Harrington et al.,

1997). Australia is a multicultural country and there are dialect differences among the many different cultural backgrounds, and there have also been some suggestions of regional differences (Cox & Palethorpe, 2003; Loakes, Hajek & Fletcher, 2010). There CHAPTER 1 7

is, however, a historical perception of relatively low regional variation (i.e. dialect differences) in Australian English and it has not been studied in depth2.

Sociolinguistic variation has been previously investigated in Australian English spoken in capital cities in the 1980s. The Sydney Urban Dialect Survey (Horvath, 1985) was a sociolinguistic survey on 117 speakers of Australian English in Sydney, New

South Wales. Findings revealed differences in both vowel and consonant variation according to the speaker’s gender, age, socioeconomic status and ethnicity (Horvath,

1985). Ingram (1989) researched variation in connected speech processes in adolescent speakers from Brisbane, Queensland and found differences in some types of phonetic reduction according to socioeconomic status. Research on sociolinguistic variation in

Australian English spoken today is scarce. Further, to our knowledge there is no previously published research on infant- or child-directed speech in Australian English.

Thus I recorded a longitudinal corpus of Australian English maternal speech for this thesis, in order to examine phonetic variation in local English as a reference point for investigating variation in Gurindji Kriol. The Australian English corpus is of a regional variety from northern Australia from the same geographical region as Gurindji people, in part so that we do not overestimate the variability in Gurindji Kriol or recognise features as special to Gurindji Kriol when they are in common with regional Australian

English.

2 Though see AusTalk (https://austalk.edu.au/) - a project currently in the process of collecting a large corpus of Australian English adult speech from all Australian states and territories. Also see the Australian Voices website (Cox & Palethorpe, 2010: http://clas.mq.edu.au/australian-voices/australian-voices) for more information and audio samples of Australian English. CHAPTER 1 8

Figure 1.1. Map of the Victoria River District and its communities (Meakins, 2011, p. xxi).

1.3 Approach to Variation in this Thesis

The current research involves discussion and analysis of phonetic variation at a segmental level. Here the terms segments and phones are used to refer to units of speech sounds that are perceptually discrete, at least to literate adults, and possess distinct articulatory and acoustic properties. This includes both phonemic and allophonic CHAPTER 1 9

segments (Pierrehumbert, 2003). Segmental variation refers to within- and between- speaker variability in the acoustic, articulatory, and/or perceptual characteristics of phones. Phonetic variation in speech input may index linguistic variables as well as social variables. The latter is termed sociophonetic variation and includes variability in phonetic forms that is related to speaker identities and social groups in the speech community (Foulkes & Docherty, 2006).

The studies presented here focus on phonetic variation in fricatives. Fricatives are consonants that are produced by constricting the space between two articulators so that the airstream is partially restricted, resulting in a turbulent airflow (Ladefoged &

Maddieson, 1996). English contains voiceless and voiced fricatives at four places of articulation: labio-dental e.g. fat, vat [f, v], dental e.g. bath, bathe [θ, ð], alveolar e.g. sip, zip [s, z], and palato-alveolar e.g. mission, vision [ʃ, ʒ]. In addition, [h] (who) is traditionally considered a voiceless glottal fricative, although its status is not the same as a typical consonant and has also been described as a voiceless vowel in some environments (Ladefoged & Maddieson, 1996). Fricatives require precise articulatory coordination – if the airflow is fully obstructed then the sound becomes a stop, affricate, or a fricated stop, and if there is too much space between the two articulators then it becomes an approximant (voiced or voiceless). Cross-linguistically fricatives are relatively difficult for children to produce and tend to be acquired later than stops and glides (Dinnsen, 1992; Kent, 1992).

In Gurindji Kriol fricatives are mainly present in Kriol-based words as they do not occur phonemically in traditional Gurindji. (In fast Gurindji, fricatives occur occasionally as reductions of intervocalic stops, e.g., tarukap ‘swim’, can be pronounced with an intervocalic voiced velar fricative.) The fricatives known from previous research to occur in Gurindji Kriol are at three places of articulation: CHAPTER 1 10

labiodental [f], alveolar [s] and palatal [ʃ]. There is voicing variation, and fricatives are also used in variation with stops (Meakins, in press). There is no previous research that provides a systematic quantitative description of fricatives in Gurindji Kriol, or indeed in any Australian contact language variety to our knowledge. An investigation of the distributions and nature of fricatives that children are exposed to in their home language would contribute toward our understanding of Gurindji children’s acquisition of fricatives and the acquisition of highly variable phonology in general. This would provide information for language professionals serving the community such as teachers and speech pathologists.

1.4 Structure of the Thesis

This thesis is an investigation of phonetic variation in the input children are exposed to at the segmental level, and how it changes over time as children age from approximately 1;6 to 2;0 to 2;6. The thesis includes three research studies. The first study is of fricative variation in Gurindji Kriol maternal speech. We investigated variation within and between lexical forms, and the relationships between sounds in

Kriol-derived words and fricatives in corresponding English cognates. We also examined change in fricative variation in maternal speech over time as children aged.

The second study is an investigation of input to children in a monolingual environment in Katherine, a regional town in northern Australia. Some types of phonetic variation in fricatives in this variety of Australian English are related to casual speech and different variants may carry social meanings. We examined how mothers’ use of casual phonetic variants changed as their children got older.

Studies one and two are based on analyses of phonetic transcription data. The transcribers were native speakers of Australian English speakers, but there may be differences in how Gurindji Kriol phones are perceived between native and non-native CHAPTER 1 11

speakers of Gurindji Kriol. This is because phonetic transcription is a perceptual activity involving categorical judgements, and the transcriber’s perceptual judgement about the category a sound belongs to depends to some extent on their own language background

(e.g. Coussé, Gillis, Kloots & Swerts, 2004; Strange & Jenkins, 1978). Depending on the questions being investigated, native speaker phonetic transcriptions are generally considered valid because perceptual judgements are made based on the sound categories of the speech community (Edwards & Beckman, 2008). In the current context, however, it was not practical to train speakers to the technical level required to do phonetic transcription, as is often the case in field linguistics and under-documented languages.

Therefore in the third study of this thesis we used Visual Analogue Scales (VAS) to elicit native speaker judgements of transcribed segments and compared how native speakers and non-native transcribers perceived fricative variation in Gurindji Kriol.

Results were used to further our interpretation of Gurindji Kriol fricative variation.

1.5 Research Questions

The overall aim of this thesis was to investigate variation in fricative production in mothers’ speech in Gurindji Kriol and northern Australian English as children aged from approximately 1;6 to 2;0 to 2;6. The overarching research questions of the empirical studies in this thesis are:

1) What is the nature and extent of variation in fricative production in

Gurindji Kriol maternal speech?

2) How does phonological reduction change over time in northern

Australian English maternal speech?

3) How can we elicit judgements of fricatives in Gurindji Kriol from

native speakers without phonetic training, and how do native speaker

perceptions add to phonetic transcription? CHAPTER 1 12

The three studies each make up a chapter of this thesis (Chapters 4, 5 and 6) written in journal article style3 and are followed by a general conclusions chapter

(Chapter 7). In the next two chapters I discuss the literature and conceptual framework of children’s acquisition of phonology (Chapter 2), and describe in detail the methodology of this research (Chapter 3).

3 Because this thesis is written in the journal article format of the University of Wollongong, whereby some chapters are styled as journal articles, there is necessarily a small amount of repetition in some sections, particularly the methodology.

CHAPTER TWO

Empirical and Theoretical Context of Children’s

Phonological Acquisition from Speech Input CHAPTER 2 14

This thesis contributes to our understanding of how phonological variation in the input changes as children get older, in the age range 1;6 to 2;6, in two relatively underdescribed language varieties. The aim of this chapter is to contextualise the empirical research on phonological variation in maternal speech and to provide the theoretical rationale for investigating speech input to children. As background to the empirical studies (Chapters 4, 5, 6), this chapter provides a review of the literature in child-directed speech and children’s phonological development after infancy, as well as a discussion of exemplar theories of phonological representations as the theoretical rationale underlying this research.

The organisation of this chapter is as follows. First, a review of the literature on child-directed speech and how it is conceptualised in the current studies and the development of children’s phonological knowledge after infancy. Second, the aims, rationale and research questions of each empirical study in this thesis. Third, a discussion of exemplar-based models of phonological acquisition as the theoretical framework of this research. The chapter ends with summaries of the three research studies that comprise Chapters 4, 5 and 6.

2.1 Literature Review

2.1.1 Child-Directed Speech

Consistent use of a different speech register when communicating with children occurs in male and female adult speakers and is widespread across many (if not all) cultures and language varieties (Byrant & Barrett, 2007; Fernald et al., 1989; Grieser &

Kuhl, 1988). Child-directed speech contains simpler lexical and syntactic structures than adult-directed speech (Ferguson, 1977), as well as a slower tempo, longer pauses, more variability in pitch contours, and higher fundamental frequency (Fernald et al.,

1989; Katz, Cohn & Moore, 1996; Snow, 1977). Infants prefer to listen to infant- CHAPTER 2 15

directed over adult-directed speech (Cooper & Aslin, 1990; Fernald, 1985), regardless of the sex or language of the speaker (Fernald & Morikawa, 1993; Werker & McLeod,

1989), likely due to the positive affect conveyed in infant-directed speech (Singh,

Morgan & Best, 2002). Speaking in a register that young children prefer to listen to has the effect of guiding their attention to the speech signal (Dominey & Dodane, 2004) and encouraging communication between children and caregivers (Cruttenden, 1994). The linguistic functions of child-directed speech remain unclear and are likely to vary depending on the type of modification made, the language variety and cultural context, the kind of interaction taking place, and the child’s age and receptive and productive language abilities (Cruttenden, 1994; Kitamura, Thanavishuth, Burnham &

Luksaneeyanawin, 2002; Stern, Spieker, Barnett & MacKain, 1983; Soderstrom, 2007).

Research on phonological features of child-directed speech suggests they tend to be exaggerated, compared to adult-directed speech. Throughout the literature different terminology has been used to characterise the differences. In an early cross-linguistic study of baby-talk, Ferguson (1977) talked about baby-talk as a “simplified register” (p.

210) that contains ‘simplifying’ and ‘clarifying’ processes. These processes include use of limited sounds and syllable forms with a tendency toward reduplicated consonant- vowel (CV) forms (e.g., mama, dada), and substituting articulatory easy sounds for articulatory complex ones (e.g. replacing rhotics with glides). Ferguson (1977) suggested that adults make modifications in baby-talk to adjust to children’s own language abilities to make the input easier to understand. One function of baby-talk as a simplified speech register may be “a response to the need for improved communication when one of the participants has only a limited ability to use language normally”

(Ferguson, 1977, p. 232). CHAPTER 2 16

Later studies on child-directed speech in the 1980s also used the terms

‘clarification’, contrasted with ‘reduction’, and ‘(un)intelligibility’. ‘Clarification’ referred to exaggerated vowel sounds in mothers’ speech to children, as defined by greater dispersion and less overlap between vowel categories as measured by formants

(Bernstein Ratner, 1984a). ‘Reduction’ referred to processes such as consonant deletion, glottalisation, and palatalisation (Bernstein Ratner, 1984b; Shockey & Bond, 1980).

‘Intelligibility’ is a perceptual term used in earlier research on segmental phonology of interadult conversational speech (see Cole & Jakimik, 1980, for a review). In this literature, ‘unintelligibility’ (Bard & Anderson, 1983) refers to the perceived degradation (reduction, deletion) of segments that characterises informal adult speech or distorted or degraded speech signals.

The variability related to the above processes has also been conceptualised as being on a continuum with hyperspeech (over-articulation) at one end and hypospeech

(under-articulation) at the other (Lindblom, 1990). Hypoarticulation is characteristic of casual speech and refers to when articulators do not reach the target, resulting in lenition

(weakening) or deletion in consonants and reduction in vowels, that is, articulation away from the extremes and toward the centre of vowel space (Browman & Goldstein,

1990). At the other end of the continuum, hyperarticulation characterises careful speech

(Picheny, Durlach & Braida, 1986) where segments are distinctly articulated (de Jong,

Beckman & Edwards, 1993) and vowels are more separated in acoustic space (Davis &

Lindblom, 2001). The continuum represents the dynamic between the perceived needs of the listener and ease of articulation for the speaker (Lindblom, 1990). In the sociolinguistics literature variation along the hyper-hypospeech continuum is associated with style-shifting, referring to the process whereby speakers accommodate their speech to the situation and perceived needs of the audience (Labov, 1984). While social CHAPTER 2 17

variation refers to variation due to interspeaker factors such as age, gender and socioeconomic status, style-shifting variation is due to ‘audience design’ factors (Bell,

1984). These stylistic factors include the speaker’s attention to their own speech, as well as the audience and the social meaning the speaker wishes to construct with the audience. Situational factors such as the topic and setting also affect variation as the speaker dynamically responds to the situation (Bell, 1984; Eckert, 2001). In the psychology literature the concept of audience design is reflected in communication accommodation theory, which explains the social and psychological factors that influence how speakers vary their speech to either converge with or diverge from their audience’s speech style (Ball, Giles & Hewstone, 1985; Giles, Coupland & Coupland,

1991). ‘Style’ and ‘register’ are often used interchangeably, though ‘register’ is also used as a broader term encompassing both situational and stylistic variation (Finegan &

Biber, 2001). Child-directed speech may therefore be a speech register, but a better understanding of the functions of child-directed speech is needed in order to place it on a style continuum4.

Against this backdrop of dynamic variability in speech, child-directed speech can be conceptualised as variation in adult speech whereby caregivers make modifications, either consciously or unconsciously, potentially in response to the linguistic needs of the child. Considerable research has examined the phonological modifications caregivers make in their speech during the child’s first year of life (e.g.

Cristia, 2010; Ferguson, 1977; Fernald, 1989; Fernald et al., 1989; Fernald &

Morikawa, 1993; Grieser & Kuhl, 1988; Kirchhoff & Schimmel, 2005; Kitamura et al.,

2002; Kuhl et al., 1997; Lee & Davis, 2010; Lee, Davis & MacNeilage, 2008).

4 Though see Wassink, Wright and Franklin (2007) for evidence that variation in infant- directed speech resembles that of style-shifting and may relate to audience design. CHAPTER 2 18

Relatively few studies, however, have investigated longitudinally the characteristics of adult speech to children after infancy and how and when these change as children get older.

The studies that have examined phonological features of speech to older children

(e.g. Foulkes, Docherty & Watt, 2005; Ko, 2012; Liu, Tsao & Kuhl, 2009; Smith,

Durham & Fortune, 2007) are discussed in more detail in the literature reviews of

Chapters 4 and 5. Overall, these studies suggest two things. First, a combination of social and stylistic factors affects phonological variation in speech to children as they get older. Second, the change from child-directed to adult-directed speech may not necessarily be a gradual linear shift in style. The input that children are exposed to in their early years is dynamic, likely changing over time as the perceived needs of the children as listeners and the functions of child-directed speech develop.

So far I have been referring to ‘infant-directed speech’ and ‘child-directed speech’, as these are common terms in the literature and the ones generally used by the studies discussed above. However throughout this thesis I use the term ‘maternal speech’, particularly when describing the speech data recorded and analysed for this research. This is because ‘maternal speech’ is a more general term which encompasses all of the mother’s speech to which children are exposed, rather than just speech directed specifically to them. ‘Maternal speech’ also covers both infant- and child- directed speech, which refer to the same concept with the use of ‘infant’ or ‘child’ depending on the age of the children in the study. In the current research children were aged between approximately 1;6 and 2;6, which is just after infancy and into early childhood. Much of the literature on the relationship between input and phonological acquisition has focused on infant-directed speech, but research also shows that phonological acquisition is an ongoing process and the input continues to play an CHAPTER 2 19

important role after infancy. My rationale for using the concept of maternal speech and the supporting evidence is discussed next. This is followed by a brief discussion of the literature on phonological development after infancy.

2.1.2 Sources of Speech Input to Children

Child-directed speech is not the only type of input that children are exposed to, as evidenced by van de Weijer (2002) who recorded all the speech that one infant heard during her waking hours from age 6 to 9 months. He transcribed and analysed in depth three weeks of these recordings (the first, middle, and last weeks) and found that only

14 per cent of utterances were spoken by an adult directly to the infant. The infant had a sister 2 years older than her, and 60 per cent of the speech transcribed was between an adult and the older sister. Inter-adult speech made up 19 per cent of the total, which is still a higher proportion than the infant-directed speech. Most of the infant’s activities during the recording period were with one or both parents, at home with the older sister and an adult babysitter, and at a daycare centre with the older sister, other children, and the daycare centre staff. Although this study only involves one infant, it makes sense that the types and amount of input to which children are exposed would be highly dependent on family structure and functions of family roles.

Studies have also shown that other types of input may be highly relevant for children’s language learning, as children can encode and store in long-term memory at least some aspects of speech that are merely overheard. This was demonstrated by Au,

Knightly, Jun and Oh (2002), who found that participants who overheard informal

Spanish spoken by native speakers in childhood were significantly better at producing a native-like Spanish accent when learning Spanish as an additional language in adulthood than participants who had not been exposed to Spanish in childhood. The

‘overhearing’ group comprised participants who had overheard Spanish as children CHAPTER 2 20

(with the most exposure occurring between birth and age 6;0) but were spoken to directly in Spanish only to a minimal extent, i.e. no more than occasional words or short phrases (Au et al., 2002). The finding that overhearing a language in childhood had a positive impact on adult learners’ Spanish accents has two implications. First, phonetic information can be learned implicitly and retrieved later in life for further language development (i.e. learning a second language). Second, children may attend to and encode information in speech they are exposed to above and beyond that directed specifically to them – that is, child-directed speech is not the only type of speech input that children are exposed to and encode into memory.

Children’s ability to learn through overhearing was tested more directly by

Akhtar (2005), who examined the ability of children (mean age 2;1) to learn novel words in third-party conversations between two adults while being distracted with an activity. Akhtar (2005) found that the presence of a distraction did not significantly affect children’s learning of a new word, demonstrating that at age 2;1 children can not only learn words through overhearing adult conversation, but can continue to learn when being distracted with another activity.

Infant- and child-directed speech are important components of children’s home language in their early years, but children are also exposed to other sources of speech input (e.g. from overhearing inter-adult speech) that have been known to contribute to their language learning. Therefore, descriptions of speech input available to children should include speech spoken within earshot even if it is not directed specifically to the child. In a review of the nature of speech input to infants, Soderstrom (2007) noted the problems with defining what exactly constitutes the input, and contended that

‘maternal’ speech from the female primary caregiver plays a key role in children’s early language experience. This is perhaps particularly true of Western families. CHAPTER 2 21

In the current research, maternal speech is the focus of analysis. Maternal speech is defined as all speech spoken by the mother around as well as to the target child. This is not to say that the speech of fathers (and aunts, uncles, grandparents and others) is not also important, but with the relatively small numbers of speakers used in corpus based studies we cannot reliably analyse interspeaker differences, and so must control for characteristics of speakers such as sex that may influence the variables being investigated. In practical terms, there is also a cultural issue in research in northern

Australian Aboriginal societies. The Meakins (2011) corpus that was built on for these studies contains mostly women speakers, in part because it would have been culturally inappropriate for female researchers to work with young men who speak Gurindji Kriol.

Both the Meakins (2011) Gurindji Kriol corpus and the Australian English corpus collected for this research are longitudinal and comprise maternal speech to children as they age between approximately 1;6 and 2;6. There is currently relatively little research on how maternal speech changes over time between these ages, despite research showing that input continues to play a role in phonological and lexical development in children of these ages and beyond. The following section discusses this research and provides the rationale for investigating phonetic characteristics of input to children after infancy.

2.1.3 Phonological Development After Infancy

There are many studies on speech input and language development that focus on infant-directed speech in children’s first year of life. During this first year infants acquire information about sound patterns in the ambient language. Specifically, between ages 0;6 and 0;9 they develop sensitivity to prosodic patterns and permissible segment sequences in the language being acquired (Jusczyk, Cutler & Redanz, 1993; Jusczyk,

Luce & Charles-Luce, 1994). Infants display sensitivity to vowel contrasts specific to CHAPTER 2 22

their native language at age 0;6 (Kuhl, Williams, Lacerda, Stevens & Lindblom, 1992), and by age 0;9 they prefer to listen to words that are consistent with phonetic, phonotactic and prosodic patterns in their native language (Jusczyk, Friederici, Wessels,

Svenkerud & Jusczyk, 1993). The speech input during infancy plays a critical role in developing sensitivities to phonetic contrasts and a knowledge base of how sounds are organised in the ambient language. Phonological acquisition, however, is a protracted process and speech input continues to play a role throughout development.

The notion that the development of phonological knowledge is an ongoing process is supported by research that shows a relationship between vocabulary and phonological development. These studies suggest a feedback loop between the lexicon and phonological processing, meaning that as the lexicon develops over time, children’s phonological knowledge will develop as well. Edwards, Beckman and Munson (2004) used nonword repetition as a measure of phonological processing and tested it in children aged 3;2 to 8;10. Nonwords contained segment sequences that have both high and low phonotactic probabilities in the native language (English). It was found that nonword repetition had a significant positive correlation with expressive vocabulary size, and that vocabulary size mediated the effect of frequency on nonword repetition scores (Edwards et al., 2004). The first result suggests that a larger vocabulary supports children’s phonological knowledge – knowing more words helps children to perceive and reproduce new word forms, possibly because they are able to make more accurate generalisations of phonological patterns. The finding that children scored higher on nonwords with high-frequency segments than low-frequency segments again demonstrates the effect of the lexicon on phonological development. Low-frequency segments are not as well represented in the lexicon, which means children have less top- down information to draw on when trying to produce them. That vocabulary size CHAPTER 2 23

mediated the effect to some extent supports the argument that a larger vocabulary allows children to make more robust phonological generalisations (Edwards et al.,

2004), as there is more data in the lexicon to draw on for the creation of an articulatory representation for the new nonword. Vocabulary growth therefore helps to refine phonological knowledge.

Perceptual aspects of phonological acquisition also develop gradually over time.

Slawinksi and Fitzgerald (1998) investigated development of perceptual boundaries of the /ɹ-w/ contrast using a forced-choice identification task for children aged 3;0, 4;0 and

5;0 and adults. They found that the phonemic boundary between /ɹ/ and /w/ was closer to the /ɹ/ endpoint for children relative to adults, and that it increasingly shifted toward the adult-like boundary with child age. That is, older children judged more stimuli on the [ɹ-w] continuum as belonging to the /ɹ/ category. Older children were also more consistent with judgements, which suggests greater confidence in their perceptual judgements (Slawinksi & Fitzgerald, 1998). Further, the children aged 5;0 made adult- like judgements of the /w/ category, but not of the /ɹ/ category. These results show that perceptual boundaries of phonemic categories are still developing throughout childhood and must continue to be refined after 5 years of age.

Munson, Swenson and Manthei (2005) also showed that lexical and phonological development is an ongoing process. They investigated two groups of children of mean ages 4;3 and 7;2 and tested their ability to repeat real words and nonwords that varied in neighbourhood density (i.e. phonological similarity to other words in the lexicon). Munson et al. (2005) found that neighbourhood density affected real word repetition in the older children but not the younger children, and suggest that because the younger children have smaller lexicons there is less competition from phonetically similar words in lexical processing. In the nonword repetition tasks they CHAPTER 2 24

found evidence of phonological facilitation that increased with age. The older children were significantly faster at repeating nonwords with high-probability than low- probability phonotactic segments (Munson et al., 2005). Together these findings show that both lexical competition, perhaps due to neighbourhood density, and phonological facilitation continue to develop throughout childhood.

Research on phonemic categorisation in older children suggests that the development of adult-like strategies for making perceptual decisions continues throughout childhood and into adolescence (Hazan & Barrett, 2000). Gradient stimuli were presented across four different phonemic contrasts to children aged between 6;0 and 12;6 and to an adult control group (mean age 29;8) in a forced-choice identification task. To test children’s ability to integrate acoustic cues the researchers also manipulated cues to the contrast in the stimuli so that in some conditions only one acoustic cue was varied, while in “combined-cue” conditions multiple cues were systematically varied. Hazan and Barrett (2000) found a significant effect of age even after controlling for possible attention effects (by excluding participants who did not reach 85% consistency in any condition). Pairwise comparisons revealed a significant difference between the youngest (6;0-7;6) and oldest (11;6-12;6) groups of children, and that the largest difference was between all child groups and adults. Thus, although the older children were significantly more consistent than the younger children in categorising phonemic contrasts, they were still significantly below the adult level.

Although infancy appears to be a critical period for some aspects of phonological development that set the basis for future word learning, it is also clear that it is a gradual process continuing well into at least early adolescence. Further research is still needed on the nature of the speech input beyond infancy. Even throughout adulthood new input is presented that may refine phonological representations (for CHAPTER 2 25

example, perception of a new accent may introduce phonetic variation that an adult has not heard before but must still somehow process). Periods of large vocabulary growth may significantly influence phonological representations (Rost & McMurray, 2009), and in particular growth in expressive vocabulary. In order to produce a segment children are likely to require articulatory as well as acoustic representations (Edwards et al., 2004), and children’s representations receive further refinement through the production-perception feedback loop (Pierrehumbert, 2003). Therefore, speech input continues to play a role in language development throughout childhood, and may be especially important in periods of expressive vocabulary growth where input contributes to both lexical development and refinement of the phonological system.

2.1.4 Summary of Literature

The sound structures in speech input to children play a central role in children’s language development. Learning to perceive and produce words in the native language requires children to extract sound segments from the input and organise them in memory (e.g. Jusczyk et al., 1993; Jusczyk et al., 1994; Kuhl et al., 1992; Stoel-

Gammon, 2011; Swingley, 2007, 2009; Werker, Fennell, Corcoran, Stager, 2002).

There is a large literature on the phonology of speech to children during infancy (e.g.

Cristià, 2010, 2011; Davis & Lindblom, 2001; Fernald et al., 1989; Grieser & Kuhl,

1988; Katz et al., 1996; Kirchhoff & Schimmel, 2005; Kuhl et al., 1997; Snow, 1977;

Stern et al., 1983), as this is an important period for acquiring sound contrasts and phonotactics of the native language (Aslin, Saffran & Newport, 1998; Coady & Aslin,

2004; Jusczyk et al., 1993; Jusczyk et al., 1994; Kuhl et al., 1992). Speech input to children consists of more than specific infant-directed and child-directed speech, and continues to contribute to phonological development after the first year of life. CHAPTER 2 26

Early childhood is also an important time in children’s language acquisition, as their knowledge of phonological processes in the native language develops and they start to consistently produce whole words (e.g. Nelson, 1973; Swingley, 2007), and expressive vocabulary is important to the phonological development process. Previous research has found differences in phonological features of child-directed speech to children after infancy compared with adult-directed speech. There are, however, few longitudinal studies on phonetic variation in the input to children after infancy that show how and when variation in child-directed speech develops as children get older.

2.2 Thesis Aims and Rationale

The purpose of this thesis is to investigate fricative variation in mothers’ speech to children in two language varieties in northern Australia, Gurindji Kriol and northern

Australian English. To realise this purpose there are two specific aims:

1) To longitudinally analyse fricative variation in mothers’ speech in

Gurindji Kriol and northern Australian English as their children age from

approximately 1;6 to 2;0 to 2;6.

2) To add further information to our phonetic transcription of fricatives in

Gurindji Kriol (a phonetically underdescribed language) by eliciting

native speakers’ own perceptions of Gurindji Kriol fricative variation.

In this section I discuss the rationale for the three empirical research studies in this thesis, that is, our motivation for investigating fricatives in Gurindji Kriol and northern Australian English maternal speech and for eliciting native speaker judgements of fricative variation in Gurindji Kriol. CHAPTER 2 27

The first study in the thesis is a longitudinal analysis of stop-fricative variation in Gurindji Kriol maternal speech. The aim of this first study is to investigate the variation in fricatives that Gurindji children are exposed to in their home language before they go to school, where English is the main language. Gurindji Kriol is a relatively new mixed language and the Kriol-derived words have inherently high phonetic variation (Sandefur, 1991). Fricatives are of particular interest in Gurindji

Kriol for several reasons. Traditional Gurindji does not contain phonemic fricatives, though there are allophonic fricatives in some phonological contexts. Kriol, on the other hand, is English-lexified, and English contains phonemic fricatives at several places of articulation. Kriol contains some fricatives due to its English influence, and these are highly variable with stops as Kriol words often have alternate phonetic forms (Sandefur,

1991). Impressionistically, Kriol-derived words in Gurindji Kriol also contain high variation between stops and fricatives (e.g. Meakins, 2007, p.358).

From a language acquisition perspective, fricatives are of interest across languages. Children tend to acquire fricatives later than stops in production, as fricatives require more precise motor control to constrict the degree of airflow between articulators and are therefore more difficult to produce (Dinnsen, 1992, pp.196-199;

Kent, 1992, pp.75-76). There is also some cross-linguistic evidence of differences in distributions of fricatives and stops in mothers’ speech to children compared to their speech to adults (Lee, Davis & MacNeilage, 2008; Lee & Davis, 2010), likely due to differences in lexical choice (Daland, 2013). Thus across languages there appears to be variation in children’s production of fricatives as they acquire them over time, as well as in the input they are exposed to through variation in segmental distributions.

Whether due to inherent features of the language or developmental and stylistic factors, fricative variation in speech input to children has implications for our CHAPTER 2 28

theoretical understanding of children’s phonological acquisition. Phonetic variation is a broad phenomenon that includes sociolinguistic and phonological processes in production and perception. Language varieties that do not contain the same extent of phonetic variability as Gurindji Kriol still contain variation as a result of, for example, phonological context effects, lexical factors, speech rate and speech style. Across languages children are exposed to phonetic variation in speech. According to exemplar- based models of children’s phonological acquisition (see §2.4 Theoretical Framework), phonetic variation in the input can be perceived, encoded and incorporated into children’s phonological representations in long-term memory (e.g. Pierrehumbert,

2003). Analysis of different types of phonetic variation across languages is necessary to further our understanding of the patterns of variation to which children may be exposed, how these patterns change as children develop, and the range of factors that potentially contribute to the variation.

The second empirical study in this thesis investigates phonological reduction in mothers’ speech in northern Australian English, and how it changes over time as children age. The English variety we analysed is from the town of Katherine in northern

Australia, which is in the same geographic region as the communities where Gurindji

Kriol is spoken. This variety is henceforth referred to as Katherine English. Katherine

English does not have the same inherent high phonetic variation as Gurindji Kriol.

Instead, we examined possible stylistic variation, i.e. variation due to factors such as the audience and formality of the speech style (for further discussion see §2.1 Literature

Review pp.16-17), taking into account phonological and lexical effects. In this study we analysed the reduction processes word-initial /h/ deletion and word-final /v/ deletion.

Both these processes involve fricatives and in some contexts the reduction indicates non-citation speech, or a less carefully articulated speech style. Therefore in maternal CHAPTER 2 29

speech /h/ and /v/ deletion may be affected by child age, if mothers (consciously or not) modify phonetic realisations of word forms as their children’s receptive and productive language develops.

Study 1 provides the first quantitative analysis of fricatives in Gurindji Kriol.

Phonetic variation is high and it is difficult to untangle the many contributing factors to stop-fricative variation particularly in a naturalistic sample5. Findings did suggest that, in addition to the variability inherent in the language variety, some of the variation may have been due to mothers modifying their speech as children got older. In Study 2 we further explored the possibility that mothers may differentially modify their speech as children aged in another language variety, Katherine English. The reduction in fricatives that we analysed in Katherine English is a different kind of phonetic variability to stop- fricative variation in Gurindji Kriol. Study 1 and 2 together provide us with a bigger picture of patterns of phonetic variation in the input children are exposed to between ages 1;6 and 2;6 and potential influences on variation. These findings deepen our understanding of the phonetic information children may be perceiving and storing as phonological representations and associated sociolinguistic information in memory.

The data for studies 1 and 2 consist of IPA phonetic transcription of naturalistic speech recordings. Phonetic transcription is a useful clinical and analytical tool and is widely-used in phonological development research. It does, however, have some limitations as it is a record of the transcriber’s perception of auditory events and thus subject to some perceptual biases. For the Australian English data we addressed this using the standard method of checking transcription agreement with a second,

5 The study of Gurindji Kriol fricatives is based on phonetic transcription and not acoustic analysis, in part because the recordings are naturalistic and therefore contain background noise and amplitude variation that would make spectral analysis of frication and stop bursts relatively difficult to use. CHAPTER 2 30

independent transcriber who is a native speaker of Australian English (e.g. Shriberg &

Lof, 1991; Edwards & Beckman, 2008). For Gurindji Kriol we also checked segmental agreement with an independent transcriber; however, both transcribers were native

Australian English speakers as there is no straightforward procedure for checking phonetic transcription with native speakers of Gurindji Kriol. Although native speakers have implicit linguistic knowledge that can improve our (as non-native speakers) understanding of their language, it was not practical to train native speakers of Gurindji

Kriol to the high technical level required for IPA phonetic transcription. This is our broad rationale for conducting the third study included the thesis.

The aim of Study 3 was to develop and test a method for eliciting native speaker perceptual judgements of fricative variation in Gurindji Kriol, and to use the results in conjunction with phonetic transcription of non-native speakers to further our knowledge of phonetic and phonological variation in Gurindji Kriol. To do this we used a series of visual analogue scales, which have previously been used to investigate variation in children’s productions (Julien, Munson & Edwards, 2012; Munson, Edwards,

Schellinger, Beckman & Meyer, 2010; Schellinger, 2008; Urberg-Carlson, Munson &

Kaiser, 2009). To our knowledge, Study 3 is the first study to use visual analogue scales to examine variation in adult speech in a phonetically and phonologically under- described language. We found that the method was largely successful and offer recommendations for refining the procedure in future research.

In conjunction with the results from Study 1, findings from the visual analogue scale study provide information about the places of articulation and word positions where fricative variation occurs in Gurindji Kriol, and where it is less likely to occur.

Comparing native speaker judgements on the visual analogue scales with non-native speaker phonetic transcription enabled us to examine possible effects of language CHAPTER 2 31

background on perception of fricative variability and whether these may have affected the findings of Study 1. Further, Study 3 gives insight into transcriber discrepancies and ambiguities, that is, when the two Australian English phonetic transcribers disagreed or were unsure of a transcription. Overall, the third study in this thesis allowed us to include native speaker judgements of fricative variation in Gurindji Kriol to our phonetic analyses and provided further information for our interpretation of data based on phonetic transcription in Study 1.

2.3 Research Questions

This thesis contains three research articles. The overarching research questions of each study are presented here along with the specific sub-questions.

2.3.1 Study 1 (Chapter 4)

What is the nature and extent of variation in fricative production in Gurindji

Kriol maternal speech?

• What is the nature and extent of fricative variation within common Gurindji

Kriol lexical forms?

• What are the relationships between fricatives in English cognates and their

pronunciation in Gurindji Kriol words that contain stop-fricative variation?

• How does fricative variation within and between Gurindji Kriol lexical

forms change in mothers’ speech as children get older, taking linguistic

factors (phonological environment and word position) into account?

2.3.2 Study 2 (Chapter 5)

How does phonological reduction change over time in northern Australian

English maternal speech?

• How do rates of word-initial /h/ deletion and word-final /v/ deletion change

in mothers’ speech as children age from 1;6 to 2;0 to 2;6? CHAPTER 2 32

• Are there lexical effects on word-initial /h/ and word-final /v/ deletion? If so,

what are they and how do they change over time?

• What is the effect of speech rate on word-initial /h/ and word-final /v/

deletion?

2.3.3 Study 3 (Chapter 6)

How can we elicit judgements of fricatives in Gurindji Kriol by native speakers without phonetic training, and how do native speaker perceptions add to phonetic transcription?

• How can Visual Analogue Scales be used in the field to record native

speaker perceptions of fricatives in Gurindji Kriol?

• How do native speakers’ judgements of fricatives on a continuous scale

differ from non-native transcribers’ categorical judgements, and do

differences depend on word position and type of judgement?

2.4 Significance

There has been little previous research on phonetic variation in speech input to children of the age group in this project, 1;6 to 2;6, and to the author’s knowledge this is the first study of maternal speech in regional Australian English and is among the first studies of phonology in Gurindji Kriol (see also Jones, Meakins & Buchan, 2011; Jones,

Meakins & Muawiyath, 2012; Jones & Meakins, 2012a, 2012b). The project contributes to our understanding of phonetic variation that children are exposed to between ages 1;6 and 2;6, the phonetic modifications that mothers make in their speech to children at these ages, how variation in the input changes as children get older, and linguistic and sociolinguistic factors that influence phonetic variation in maternal speech. Findings are discussed in terms of theoretical models that explain how variation in speech input is processed by the listener/speaker and acquired by children. This thesis also provides CHAPTER 2 33

novel findings on fricative variation in an Australian Aboriginal language and suggests avenues for further research on fricatives in Kriol varieties spoken in other parts of

Australia.

The current project contributes to a systematic description of phonetic variation in maternal Gurindji Kriol, which may have future use in the development of measures to assess and compare the phonological skills of child speakers of mixed languages and

Kriol spoken in northern Australia. Findings can be used to inform teachers of the likely phonological strengths and weaknesses of Gurindji Kriol speaking children and in facilitating learning of traditional languages and English, as discussed in Chapter 7.

Practically, Study 1 (Chapter 4) contributed phonetic transcription and analysis to an existing audiovisual database of conversational Gurindji Kriol. Study 2 (Chapter

5) resulted in the first phonetically transcribed corpus of regional Australian English maternal speech. Study 3 (Chapter 6) documents a novel methodology for using Visual

Analogue Scales in the field to elicit native speaker perceptual judgements of phones.

Theoretically, the research in this thesis contributes to our understanding of how mothers modify their speech differentially as children age, and offers some speculation as to why this may occur. This is discussed further in the discussion sections of Studies

1 and 2. Further, by examining a situation where children are acquiring language under conditions of highly variable input, these findings provide evidence to inform theoretical models of children’s acquisition of phonology and variability. In the next section I describe and discuss the theoretical framework underpinning the empirical research in the thesis.

2.5 Theoretical Framework

This thesis is based on episodic theories of speech perception and production, that is the theory that individual perceived auditory events leave episodic traces in long- CHAPTER 2 34

term memory that are used in categorisation tasks. At a phonological level this means that each perceived phonetic episode, or exemplar, is encoded into long-term memory with fine-grained detail intact. A key feature of this theory is that variation in the speech input is retained in memory rather than filtered out as noise, thus variation contributes to the organisation of cognitive representations. Therefore with exemplar-based theories we can gain a greater understanding of phonetic variation that children may be acquiring by analysing and describing the speech to which they are exposed. This section contains a description of general exemplar theory and a brief review of the evidence for exemplar models of speech processing, followed by an examination of how these models are applied to children’s acquisition of phonology and a discussion of the theoretical implications of processes explored in this thesis.

Exemplar models of phonological representations are useful for the purpose of this thesis as they provide an account of how phonetic variation and associated information, both linguistic and socioindexical, may be processed and used in the categorisation tasks required for speech perception and production. The purpose of this thesis is to examine variation in the input and potential contributing factors rather than to specifically test exemplar theories of phonological development. Given this, the discussion here is of the theoretical rationale that underlies the current research and will be limited to the general features of exemplar models relevant to this research without making claims as to the precise mechanics of the various proposed models in the literature.

Central to exemplar theories is the distinction between memory of general information about concepts and memory of specific perceived events, a distinction that is long-standing in the psychology literature. One of the earliest arguments for the differentiation of episodic and semantic information was made by the psychologist CHAPTER 2 35

Tulving (1972), who defined episodic memory as a processing system for temporally dated perceptual events and semantic memory as organised knowledge about general concepts and their rules, meanings, and interrelations. Perceptual episodes can be stored in episodic memory purely on the basis of the perceived features of episodes, while semantic memory involves the cognitive referents and more general abstracted information of these episodes (Tulving, 1972). In some literature the terms ‘episodes’ and ‘exemplars’ are synonymous; however, in other memory literature ‘episodes’ refers more specifically to autobiographical memories that are consciously retrieved (e.g.

Wheeler, Stuss & Tulving, 1997). The more general term ‘exemplar’ will therefore be used from this point forward to refer to a perceptual event.

The critical difference between exemplar theories and abstractionist theories is the explanation of how memories of specific exemplars and memories of general concepts are processed and stored in long-term memory. The more traditional abstractionist theories, such as Tulving’s (1972) model, posit separate memory systems for these different types of information. Abstractionist theories of language processing therefore postulate abstract forms of lexical representations that are separate from the perceptual events (e.g. Becker, 1980; Forster, 1976; Morton, 1979), with phonological processing occurring either through additional pathways (dual-route models, e.g.

Behrmann & Bub, 1992; Coltheart, Curtis, Atkins & Haller, 1993) or in parallel with lexical access (connectionist models, e.g. Seidenberg & McClelland, 1990). On the other hand, exemplar theories propose that representations in long-term memory are of specific individual episodes and abstractions emerge from this store of exemplars (for an in-depth review of abstractionist and exemplar theories of word identification see

Tenpenny, 1995). CHAPTER 2 36

Exemplar theories are supported by the robust finding that listeners retain information about phonetic detail after long delays, which suggest that fine acoustic details of speech are in fact represented in long-term memory (Church & Schacter,

1994; Goldinger 1996, 1998; Goldinger & Azuma, 2004; Nygaard & Pisoni, 1998;

Palmeri, Goldinger & Pisoni, 1993; Pisoni, 1997). Exemplars are organised in clusters according to similarity across attributes, so categories consist of distributions of exemplars associated with a category label. Exemplars can be distributed along any dimension and so the same exemplar or attribute can contribute to many categories, for example the attribute of fundamental frequency may contribute to both linguistic categories such as vowel phone and socioindexical categories such as the speaker’s gender. This is a strength of exemplar models, as it means that information from speech input can be stored in multiple ways and different kinds of information can be extracted and used for different tasks (Hawkins, 2003). Exemplars can therefore be thought of as individual data points in memory, with categories emerging from clusters of these data points. The ‘dataset’ can be drawn on to make generalisations about the language as required.

Different models propose different mechanisms for exactly how exemplars are encoded and stored. Depending on the model, a new perceptual event may be organised according to either its similarity to existing exemplars or to the generalisations that emerge from categories, and each encoded perceptual event may either form a new exemplar or raise the activation of an existing exemplar (Wedel, 2006). If individual events are all stored as exemplars then each new event forms an extra case in the exemplar dataset and will contribute to its associated categories, while in activation models each new event contributes to the saliency of an existing exemplar. The important point here is that either way, every encoded perceptual event slightly modifies CHAPTER 2 37

the exemplar store and categories in long-term memory. Another important feature that these models allow for is a production-perception feedback loop (e.g. Pierrehumbert,

2003). I will return to this in more detail later on in this section when Pierrehumbert’s

(2003) model is described in relation to children’s acquisition of phonology, but in the context of general exemplar theory such a loop allows for produced speech to replicate fine-grained phonetic detail from perceived speech and to feed back into the exemplar system. This provides a mechanism for a general production-perception feedback loop in the speech community in which allophonic variation is retained and acquired by children learning the language.

As mentioned earlier, evidence supporting exemplar theories of speech processing has accumulated since at least the early nineties. One phenomenon that started distinguishing exemplar from abstractionist theories is long-term repetition priming. In lexical decision tasks repetition priming refers to the facilitation effect of a brief earlier presentation of a word on the speed and accuracy of identification of words in a later task. Thus, the first presentation of the stimulus ‘primes’ the second presentation for identification. Priming effects have been found in a variety of tasks that involve lexical retrieval (for a review see Tenpenny, 1995). Abstractionist models explain the prime as temporarily improving the accessibility of the cognitive representation, which then goes back to base level after a short period of time. On the other hand, Tenpenny’s (1995) review of repetition priming in word identification tasks showed consistent findings of a priming effect after relatively long delays, and furthermore that the effect is enhanced when phonetic detail is matched in both presentations. These effects suggest that not only are individual perceptual episodes (in this case the prime) encoded into long-term memory, but also that phonetic detail remains intact. Tenpenny’s (1995) findings are supported by studies of familiarity CHAPTER 2 38

effects in recognition tasks, that is that recognition memory of speech is improved when the prime and the stimulus to be identified (the target) are spoken by the same voice

(Church & Schacter, 1994; Nygaard & Pisoni, 1998; Palmeri, Goldinger & Pisoni,

1993). Again these effects have been found with long delays between the prime and the target (Goldinger, 1996, 1998), supporting exemplar models of speech processing whereby speaker-specific acoustic information is retained in long-term memory and drawn on in recognition tasks.

The studies mentioned above involve typical adult participants, who have had sufficient time to develop an exemplar store and form generalisations for use in categorisation tasks. Subsequent research has suggested a hierarchical organisation of generalisations of at least two levels of abstraction: one in parametric phonetic space, where perceptual memory traces are distributed along various dimensions, as discussed earlier in this section, and another level involving broader, less fine-grained generalisations that can be processed ‘on demand’ as required (Beckman et al., 2007).

Evidence for these two encoding systems comes from research on children’s phonological processing (Munson, Edwards & Beckman, 2005; Munson, Kurtz &

Windsor, 2005) and supports hierarchical models of phonological development such as

Pierrehumbert’s (2003) model of how children acquire the phonology of their native language.

Pierrehumbert (2003) posits that the aim of phonological development is to achieve minimally five levels of representation: parametric phonetics, phonetic encoding, the lexicon, the phonological grammar, and morphophonological correspondences, and as these levels are hierarchical they develop in that order

(Pierrehumbert, 2003). In this model categories are initiated through statistical learning of speech segments, i.e. conditional probabilities are computed from encoded perceptual CHAPTER 2 39

events (for more detailed accounts of infants’ statistical learning of speech segments see

Anderson, Morgan & White, 2003; Maye Weiss & Aslin, 2008; Werker & Tees, 1999).

Once categories have been initially formed, information from generalisations can be used to further refine them (Pierrehumbert, 2003). Probability-based information in the input therefore helps to shape phonological development, which is an ongoing process as categories are continually refined from both bottom-up processes (i.e. perceptual experience with new exemplars) and top-down processes – forming generalisations at different levels of abstraction, at least some of which can feed back down into lower levels (Pierrehumbert, 2003).

The purpose of this thesis is not to specifically test exemplar models of phonological development or make a claim as to the exact mechanisms and levels of representation; rather these theories have framed the current studies as they offer concise explanations for processes explored in the thesis. The main features of exemplar models relevant here are:

• Phonetic variation is retained in memory and used in speech processing tasks.

Rather than being filtered out at the encoding stage, fine-grained phonetic detail is

stored in long-term memory and is available for use in identification and

categorisation of linguistic information in the input. Further, this enables a

pathway for perceived variation to be imitated in production, suggesting a

production-perception feedback loop from which a child can acquire the phonetic

variation of their speech community (Pierrehumbert, 2003).

• Exemplars can contribute distributional information to multiple categories

simultaneously. Phonetic detail can, for example, provide information to both

segmental and prosodic generalisations (Hawkins, 2003). More broadly this CHAPTER 2 40

means that socioindexical information may be encoded simultaneously with

linguistic information (Foulkes & Docherty, 2006).

• Phonological development is an ongoing process. Each perceptual event at least

slightly alters the categories to which it contributes information. Phonetic variants

in the input contribute to the range and distribution of exemplars stored in

memory, which means changes in the input over time influence the distributions

from which categories form.

• Gradient differences in features are accounted for. Exemplars are distributed in

categories based on similarity across that category dimension rather than assigned

‘all or nothing’ category membership. This concept is particularly important to

Chapter 6, which raises the possibility of continual variation in some segments.

In summary, exemplar theories of speech processing and phonological development provide a useful, evidence-based account of how phonetic variation in input may be processed and acquired by children. According to this theoretical framework, in order to understand phonological representations that children develop and the associated linguistic and socioindexical generalisations that may form from information contained in the input, it is necessary to understand the patterns of phonetic variation in the speech input to which children are exposed.

The theories of exemplar-based phonological development discussed here provide the theoretical background and rationale for investigating phonetic variation in the input children are exposed to in their home language. The overall purpose of this thesis is to examine phonetic variation in fricatives in speech input to children in two language varieties that contain differing phonological variability. The next section contains summaries of the three research studies contained in the thesis that investigate phonetic CHAPTER 2 41

variation in maternal Gurindji Kriol and northern Australian English. This is followed by the final section, a summary of this chapter.

2.6 Summaries of Research Chapters

This thesis has been written in the journal article style format approved by the

University of Wollongong. It contains three research articles (Chapters 4, 5 and 6) that have been prepared for peer-reviewed academic journals, as well as introduction, context, methodology and conclusion chapters (Chapters 1, 2, 3 and 7) that have been written in traditional thesis chapter format. Each research article has its own abstract, background, methodology, results and discussion sections. The research chapters of the thesis are summarised in the following paragraphs.

2.6.1 Chapter 4. Fricative Variation in Maternal Gurindji Kriol

The purpose of this study was to describe systematically fricative variation in

Gurindji Kriol maternal speech and analyse the factors that potentially contribute to variation in fricative production. The literature review discusses influences of linguistic factors and child age on phonetic variation in maternal speech and how they may explain fricative variation in Gurindji Kriol. In particular the linguistic process of lenition and effects of phonological environment on phone realisation are explored.

Sociolinguistic factors discussed include the language situation and social status of languages spoken in the community, the potential effect of child age, and the situational context.

In the methodology are details of the Gurindji Kriol corpus (Meakins, 2011), the procedure of converting existing files to use in Phon (Rose, Hedlund, Byrne, Wareham

& MacWhinney, 2007) to add phonetic transcription and analysis, and the procedure for constructing a dataset of Gurindji Kriol tokens of Kriol-derived words that could potentially be pronounced with fricatives, which we termed ‘potential fricatives’ and CHAPTER 2 42

defined based on fricatives in English cognate words. Words in the dataset containing a

‘potential fricative’ were categorised into three groups: Always pronounced with a fricative, never pronounced with a fricative, and pronounced variably with a stop or fricatives.

Results contain four analysis sections. In Analysis 1 we compared the phonological characteristics of potential fricatives in the three categories of words.

Words that contained stop-fricative variation were more likely to be open-class than closed-class, and the variable segments were most frequently word-initial and at labio- dental and alveolar places of articulation. Analysis 2 is a description of the relationships between potential fricatives in variable Kriol-derived words as pronounced in Gurindji

Kriol maternal speech and fricatives in the English cognate words, in initial, medial and final word positions. Analysis 3 reports the results of logistic regression models that were ran to investigate the effects of child age, phonological environment, and speaker in all tokens and in tokens of variable words, in each word position. In initial position in all tokens and medial position in tokens of variable words, fricatives the likelihood of fricatives increased in mothers’ speech as children aged between approximate ages 1;6,

2;0 and 2;6. Analysis 4 describes in more detail the effects of child age on stop-fricative variation in mothers’ speech.

In the discussion the results are discussed in terms of contributing factors to stop-fricative variation in maternal Gurindji Kriol. An important factor appears to be lexical choice, as mothers used more words that can be pronounced variably when children were older than when they were younger. These results provide support for the concept of fine-tuning in speech to children, that is, that phonetic variation may change over time in mothers’ speech as mothers make modifications to the child’s age and language abilities. CHAPTER 2 43

2.6.2 Chapter 5. Reduction in Fricatives in Maternal Northern Australian English

This research article focuses on casual speech variation in Katherine English maternal speech and how it changes over time as children get older. The literature review for this study includes previous research on the differences in phonological variation between child-directed speech and adult-directed speech, which overall suggests that in English variation in maternal speech must change over time as mothers shift from using child- to adult-directed speech. The literature review also explores the theory that mothers may not only fine-tune their speech to children’s linguistic development, but also to their understanding of social categories indexed in speech.

Thus the literature on sociophonetic variation in mothers’ speech is also examined here.

Finally in the literature review there is a brief discussion of the previous research on phonetic variation in Australian English and casual speech processes that may occur in regional varieties.

The methods section details how the Katherine English maternal speech recordings were made and transcribed, and how searches and analyses were performed using the program Phon (Rose et al., 2007).

The results section is divided into the three analyses conducted for this study, with a brief discussion following each analysis. Analysis 1 examined the incidence of phonological reduction in maternal speech over time, focusing on two specific reduction processes, word-initial /h/ deletion and word-final /v/ deletion. Findings showed an inverted-v effect: deletion in mothers’ speech was more likely when children were 2;0 than 1;6, but was unexpectedly less likely at 2;6 than 2;0. This change over time effect was significant for /h/ but not /v/ deletion. Analysis 2 examined the local effects on these reduction processes and it was found that changes in deletion rates occurred in lexically specific contexts, usually function words. In Analysis 3 we investigated the CHAPTER 2 44

effect of speech rate on deletion and found deletion did not appear to be driven by speech rate, as there was no significant difference in speech rate between child ages 1;6 and 2;0, and between 2;0 and 2;6 speech rate increased while deletion decreased.

The discussion considers possible explanations for the inverted-v effect found for word-initial /h/ and word-final /v/ deletion in Katherine English maternal speech as children aged from 1;6 to 2;0 to 2;6. Analyses 2 and 3 show that this effect is not likely to be an artefact of changing lexical distributions or changes in speech rate over time.

Results are interpreted in terms of fine-tuning: phonetic variation in speech to children may serve different functions at different stages of linguistic and social development, and so mothers may fine-tune their speech differentially as children age.

2.6.3 Chapter 6. Investigating Native Speaker Judgements of Phonetic Variation in

Gurindji Kriol Using Visual Analogue Scales

This study arose primarily from methodological questions about how we can elicit native speaker perceptual judgements of Gurindji Kriol fricatives and use the information to increase our understanding of phonetic transcription data. The literature review discusses the use of visual analogue scales in speech perception research and how they can be used to get continuous judgements of phonetic variation. Also discussed here are potential biases and other limitations of phonetic transcription, in particular the influence of the listener’s language background on auditory judgements, as well as the applied context of the language situation where Gurindji Kriol is spoken and phonetic variation in Gurindji Kriol as a mixed language.

The methodology describes the four main procedural stages: 1) Non-native speaker IPA (International Phonetic Alphabet) transcriptions of conversational Gurindji

Kriol; 2) The development of visual analogue scales, including how tokens were selected and dichotomies determined for the scales, and how the scales were constructed CHAPTER 2 45

and counter-balanced; 3) Phonological awareness training for the native speakers of

Gurindji Kriol who participated in this study; and 4) Administration of the visual analogue scales to Gurindji Kriol native speakers. Native speaker responses on the scales were compared to the phonetic transcriptions made by non-native transcribers.

Results showed both agreements and disagreements between native speakers and non-native transcribers on different types of judgements in initial, medial and final word positions. For word-initial stop-fricative judgements, native speakers often responded towards the stop where transcribers had judged the segment as a fricative. On medial stop-fricative judgements native speakers sometimes judged the target segment as a stop and sometimes responded along the continuous scale. For voicing judgements native speakers tended to judged Gurindji Kriol segments along the continuous scale in each word position.

In the discussion we explore possible explanations for differences between native speakers and non-native transcribers, including effects of language background and lexical expectation. Findings open up further questions and hypotheses about phonetic variation in maternal Gurindji Kriol speech production.

2.7 Chapter Summary

This chapter provided the empirical and theoretical background of the research presented in this thesis. First I reviewed the literature on child-directed speech and contextualised the concept of maternal speech used here. I also discussed the literature on children’s phonological development as a rationale for investigating phonology of the input to which children are exposed. The following sections contain the aims, research questions and subquestions of the three empirical studies that make up

Chapters 4, 5 and 6, and the significance and outcomes of this thesis. I then discussed exemplar-based models of phonological representations in long-term memory as the CHAPTER 2 46

theoretical context and rationale for the current research. The final section of this chapter provided summaries of the three research studies described in the thesis. The next Chapter (Chapter 3) describes the methodologies used in these studies. The purpose of Chapter 3 is to provide an account of how the project was conducted as a whole, and to detail the collection and processing of the Gurindji Kriol and Katherine

English corpora. The methodology chapter also provides a general overview of the third study’s method. As this study is addressing a methodological question the method is explained in detail in that specific chapter (Chapter 6).

CHAPTER THREE

Methodology 48
CHAPTER 3

The methodology for the thesis reflects its nature as a corpus study of maternal speech involving fieldwork in regional and remote locations, working with speakers of the two language varieties, Gurindji Kriol and Katherine English. Chapters 4, 5 and 6 describe separate research studies that share the overarching aim to examine phonetic variation in maternal speech in Gurindji Kriol and northern Australian English. The specific methods used for each study are described in each of these chapters.

This chapter begins with the timeframes of data collection and a brief description of the activities carried out on the fieldwork for this research. This is followed by four sections that give further detail of the general methods used in this thesis: 1) The Gurindji Kriol corpus collected by Meakins (2011) and our addition of phonetic transcription and analysis, 2) Collection, transcription and analysis of the

Katherine English corpus that was created for this project, 3) Working with speakers of

Gurindji Kriol to get an understanding of the socio-cultural context of the community, discuss informally their opinions, as Gurindji mothers, of speech to children in Gurindji

Kriol, and elicit their implicit linguistic knowledge of Gurindji Kriol phonology using

Visual Analogue Scales, 4) Using the program Phon to phonetically transcribe and analyse naturalistic corpora of maternal speech. All data were analysed quantitatively and the details of analyses are specified in the research papers in Chapters 4 to 6.

3.1 Timeline of Data Collection

Data were collected in a series of four field trips to Katherine and Kalkaringi between September 2009 and June 2011. Figure 3.2 shows the timeline and the main activities conducted in each trip. In the first three trips, approximately six months apart,

I recorded families for the Katherine English corpus when children were aged approximately 1;6, 2;0 and 2;6. In trip 3 I also did some phonological awareness training with community Research Assistants in Kalkaringi in preparation for trip 4. In 49
CHAPTER 3

the fourth trip I revisited the Katherine English families to update them on the research and check with them my description of Katherine English for Chapter 4. I also worked with the community Research Assistants in Kalkaringi on the perceptual tasks described in Chapter 6, to elicit native speaker perceptions of fricative variation in Gurindji Kriol.

The field trips also involved building relationships with the local childcare service providers, consulting with a steering committee and Kalkaringi community leaders and working with cultural mentors and the project liaison. Discussions of informed consent for all participants are ongoing and continued at each trip throughout the project.

A final field trip was undertaken at the end of the project (October 2012). In this final trip we provided more feedback to the participants and communities about the results of the research in the form of presentations, booklets, and classroom materials for the Kalkaringi school.

3.2 Gurindji Kriol Corpus

3.2.1 Design

The Gurindji Kriol recordings used in this project were a subsample selected from an existing audiovisual database created by Meakins (2011) for the Aboriginal

Child Language Acquisition (ACLA-1) project6. The ACLA-1 project was a longitudinal investigation of the input to which children are exposed in three regional/remote northern Australian Indigenous communities. ACLA-1 was a project about input that examined multilingual environments that comprise traditional

Australian Aboriginal languages and a contact variety of English, with mixing between language varieties and speech styles (Simpson & Wigglesworth, 2008). Meakins (2011) created the database of Gurindji Kriol for the Kalkaringi section of the ACLA-1 project























































6 The ACLA-1 project was funded by Australian Research Council grant DP0343189 through the University of Melbourne (C.I. Gillian Wigglesworth, Jane Simpson and Patrick McConvell). http://linguistics.unimelb.edu.au/research/projects/ACLA/ 50
CHAPTER 3

and for her PhD thesis on the development and function of case morphology in Gurindji

Kriol (Meakins, 2007). The map in Figure 1.1 (from Meakins, 2011, p. xxi) is of the

Victoria River District and shows the Kalkaringi and Daguragu communities in relation to the nearest town, Katherine, and the Northern Territory capital city, Darwin.

Meakins recorded and transcribed orthographically (using the CHILDES CHAT format, MacWhinney, 2000) a longitudinal corpus of over sixty hours of naturalistic family interactions. These recordings were of Gurindji Kriol speaking women and their older Gurindji family members, and were made between 2003 and 2007 at approximately 6-month intervals. For the current research we selected a subsample from this database of three speakers at three timepoints approximately six months apart, when the target children were aged approximately 1;6, 2;0 and 2;6. We converted the transcripts to use in the CHILDES-compliant phonological transcription and analysis program Phon (Rose et al., 2007), phonetically transcribed the tokens of interest and their phonological environments using the International Phonetic Alphabet (IPA), and exported the relevant data into SPSS for further analysis. In this section I describe the subsample from the Meakins (2011) corpus of Gurindji Kriol used in the current research as well as our procedure for processing and transcribing the data and consulting community members about the project and our use of the data. 51
CHAPTER 3

Field trip 1 September – October 2009 • Establish relationships with Katherine organisations and childcare centres • Informed consent discussions with Katherine English speakers • Katherine English recordings at child age 1;6 • Work with Aboriginal cultural mentor • Meet with project steering committee

Field trip 2 March – April 2010 • Katherine English recordings at child age 2;0 • Project update meetings with Katherine organisations • Informed consent discussions with all speakers • Discussions with Kalkaringi and Daguragu community leaders and project liaison • Establish relationships with Gurindji Kriol speaking mothers and community Research Assistants

Field trip 3 October – December 2010 • Katherine English recordings at child age 2;6 • Interviews with Katherine English mothers • Project update meetings with Katherine organisations and cultural mentor • Informed consent discussions with all speakers • Discussions with Kalkaringi and Daguragu community leaders and project liaison • Conduct phonological awareness training with community Research Assistants

Field trip 4 May – June 2011 • Revisit Katherine English mothers for project updates and information checking • Project update meetings with Katherine organisations and cultural mentor • Informed consent discussions with all speakers • Discussions with Kalkaringi and Daguragu community leaders and project liaison • Conduct Visual Analogue Scales task with community Research Assistants

Figure 3.1. Stages of fieldwork and timeframes of data collection 52
CHAPTER 3

3.2.2 Participants

Participants in the subsample used in these studies were three Gurindji mothers and their children who are speakers of Gurindji Kriol. The three women were mothers of focus children in ACLA-1 and agreed to participate in the current project. These speakers were selected because there was longitudinal conversational data when their children were aged on average 1;5, 1;11 and 2;5, which are the approximate child ages under investigation in this thesis, and because they were available in the community for consultation and discussions about our use of their recordings and transcripts for this research. The three mothers were recontacted and, in the presence of an interpreter, were invited to give informed consent for us to use their speech data collected by

Meakins (2011) for this project.

The children’s ages in the recordings selected for this research are displayed in

Table 3.1. From this point forward I refer to the three timepoints of data collection in the current subsample as Stages 1, 2 and 3, though note these do not correspond to the field trips in the Meakins dataset (in which there are trips both preceding and following this subsample). The recording length and number of mothers’ utterances at each stage are shown in Table 3.2

Table 3.1 Gurindji Kriol Speaking Mothers and Age of the Focus Children

Child’s age Mother Child’s gender Stage 1 Stage 2 Stage 3

SS male 1;7 2;0 2;6

CE female 1;4 1;9 2;3

AR male 1;4 1;9 2;3

Table 3.2

Total amounts of recording and number of mothers’ utterances in the Gurindji Kriol subsample at each stage

Stage 1 Stage 2 Stage 3 Total

recording recording recording recording N N N N length length length length utterances utterances utterances utterances hh:mm:ss hh:mm:ss hh:mm:ss hh:mm:ss

SS 01:20:13 362 00:54:00 212 01:21:31 482 03:35:44 1,056

CE 00:30:28 327 00:16:33 52 00:22:31 215 01:09:32 594

AR 00:44:13 256 00:52:38 188 00:43:04 22 02:19:55 466

Total 02:34:54 945 02:03:11 452 02:27:06 719 07:05:11 2,116 54
CHAPTER 3

3.2.3 Procedure

The sample selected for the current research comprised audiovisual recordings of naturalistic family interactions in free play or natural outdoor settings. Meakins

(2011) was sometimes present in recording sessions and sometimes left the area to allow speakers to interact as naturally as possible (see Meakins, 2011, pp. 45-54 for further detail on recording methods).

Activities. Common activities in Meakins’s (2011) recordings were playing with toys provided by Meakins, which included dolls, picture books, drawing and puppets.

Recordings also took place outside at the river, which included activities such as swimming, fishing and informal story-telling.

Equipment. Speakers in the Meakins (2011) corpus wore a lapel microphone that recorded at 44Ghz 16 bit, to a minidisc recorder, which was carried by the speaker in a bumbag. A video camera with a shotgun microphone was also used to record sessions.

3.2.4 Phonetic Transcription

The transcription procedure for this project involved converting Meakins’s transcript files from CHAT format to Phon (Rose et al., 2007) with the use of conversion tools created for this purpose7. We then phonetically transcribed in IPA the tokens containing the phonological characteristics being investigated and their phonological environments (see Chapter 3 for further detail), facilitated by a dictionary of common IPA transcriptions of Gurindji Kriol words that was imported into Phon.

7 Chatter converts files from CHAT to xml formats and vice-versa (http://www.talkbank.org/software/chatter.html), and Phontalk converts files from xml to Phon formats and vice-versa (http://phon.ling.mun.ca/phontrac/wiki/PhonTalk). Both tools were created for the Phon project (Rose et al., 2007), which is a part of CHILDES (MacWhinney, 2000), and are freely available.
55
CHAPTER 3

Phon’s search function was then used to export the transcription data into SPSS for further analysis.

3.2.5 Ethical Considerations Relating to the Gurindji Kriol Corpus

There were several ethical considerations involved with building on the data in the Gurindji Kriol corpus for the current project. These included consultation with community leaders and the speakers about the direction of the project, ongoing permission from participants to use recordings and transcripts, who to work with as community Research Assistants, and dissemination of findings. Informed consent was discussed with speakers during each visit to the community in the presence of an interpreter, and throughout the project speakers were given opportunities to discuss and update their decisions about how their data would be used and stored. The consent form for Daguragu mothers in Appendix A shows the options participants had for our use and storage of their data. Participants generally consented to data storage at other facilities

(e.g. the language centre in Katherine, at national archives) if future researchers have to get their permission to access the data. Participants also allowed us to print parts of transcripts in academic publications and to use video clips and photos for conferences and the project website, with varying degrees of permissions. The speakers did not wish to use pseudonyms in academic reports – these women are proud of their language variety and agreed to their names being associated with work on it. Speakers, interpreters, community leaders and other community members involved in the project were all paid for their time and gave informed consent to be a part of the research, where they were able to discuss and choose the levels of anonymity they wished for the various formats and uses of data.

56
CHAPTER 3

3.3 Katherine English Corpus

3.3.1 Design

The Katherine English corpus is a longitudinal audiovisual database of naturalistic northern Australian English maternal speech, recorded in homes and local outdoor settings in family interaction. I recorded this corpus between July 2009 and

December 2010 (see Figure 3.1 for data collection process) in the town Katherine,

Northern Territory, which is about 300km south of Darwin and about 500km northeast of Kalkaringi and Daguragu (Figure 1.1). This corpus was designed to be comparable to the Gurindji Kriol maternal speech sample described above (section 1.5.1), and therefore recordings are of naturalistic family interactions made at approximately the same child ages between 1;6 and 2;6. The Katherine English corpus comprises over 30 hours of audiovisual recordings at three timepoints at about six-month intervals, when the target children were aged approximately 1;6, 2;0 and 2;6. The mothers’ speech was phonetically transcribed and analysed in Phon (Rose et al., 2007). Table 3.3 displays the length of recording and number of mothers’ utterances of each speaker at each stage of recording.

3.3.2 Participants

Participants were recruited through the Katherine childcare centres and other local businesses and were paid for their time. The five mothers who participated were monolingual Australian English speakers who lived and had grown up in Katherine, with the exception of one mother (Erin) who grew up in a small regional town

(population approx. 3,000) in northwest Queensland, Australia. As shown in Table 3.3, four of the mothers had a child aged 1;5-1;7 at Stage 1 of data collection (the target child); the fifth mother’s child (Cullen) was 2;0 at Stage 1. The characteristics of the mothers and the target children are displayed in Table 3.4, followed by a brief 57
CHAPTER 3

description of each dyad and the other people present in recording sessions. Mothers were given the choice of having pseudonyms for themselves and their families in academic manuscripts; thus pseudonyms are used throughout this thesis for the mothers who chose that option.

Table 3.3

Characteristics of Katherine English speaking mothers and the focus children

Child’s Child’s Mother’s Child’s age Mother Child family gender education position Stage 1 Stage 2 Stage 3

Bachelor Cathy Lucy female only child 1;5 1;11 2;5 degree

Secondary youngest Kim George male 1;5 1;11 2;5 school of three

Secondary Kellie Tyson male only child 1;7 2;1 2;7 school

Bachelor oldest of Erin Grace female 1;6 2;0 2;8* degree three

Secondary oldest of Eva Cullen male 2;0 2;6 3;0 school two

* The age difference between Stages 2 and 3 is greater for Grace than the other children because Erin and Grace’s Stage 3 sessions were recorded later due to participant circumstances

Cathy and Lucy. Lucy was Cathy’s only child. Lucy’s father occasionally participated in recordings, but he was not usually home during recording and sessions generally consist of Cathy and Lucy playing together.

Kim and George. George had two older siblings, a boy and girl aged 6;3 and

3;0 respectively at Stage 1. One or both of the siblings participated in most of the sessions. Kim was asked to focus on George in the sessions, so generally the siblings 58
CHAPTER 3

played together while Kim played with George. The children’s father participated occasionally, but usually he was not home when recordings took place.

Kellie and Tyson. Tyson was an only child. Tyson’s grandmother (Kellie’s mother) was present in most of the sessions. Other adult family members were sometimes in the vicinity but did not participate in sessions.

Erin and Grace. Grace had two younger brothers: one aged 0;5 at Stage 1 and another born between Stages 2 and 3. Sessions were generally held when the younger children were sleeping and the majority of interactions were between Erin and Grace.

Two adult males who work for Erin and her husband were occasionally present in some of the sessions.

Eva and Cullen. Cullen was an only child at the first two stages, and then had a younger brother born between Stages 2 and 3. Cullen’s father participated sometimes when he was home during sessions. Cullen refused to wear the microphone in all sessions, accordingly there is only audio of Eva’s speech at each stage. As Cullen was six months older than the other children, Eva’s data were not used for the longitudinal analyses of changes between the three child ages to ensure a simple and complete dataset.

Table 3.4

Total amounts of recording and number of mothers’ utterances in the Katherine English corpus at each stage

Stage 1 Stage 2 Stage 3 Total

recording recording recording recording N N N N length length length length utterances utterances utterances utterances hh:mm:ss hh:mm:ss hh:mm:ss hh:mm:ss

Cathy 02:05:31 1,751 02:00:18 1,863 02:44:58 2,181 06:50:47 5,795

Kim 02:11:37 1,584 01:38:00 1,513 02:28:12 1,844 06:17:49 4,941

Kellie 02:41:54 1,595 03:07:20 1,888 02:24:54 1,515 08:14:08 4,997

Erin 01:39:48 1,041 00:47:53 487 02:06:13 1,506 04:33:54 3,034

Eva 01:37:30 1,388 01:22:20 891 01:57:58 1,316 04:57:48 3,595

Total 10:16:20 7,359 08:55:51 6,642 11:42:15 8,361 30:54:26 22,362

60
CHAPTER 3

3.3.3 Procedure

Mothers were asked to interact with their children as they normally would and to focus on the target child if siblings were present. Settings for the recordings were the family home, both inside and in the backyard, and at a local park, with care taken to choose locations with minimal background noise. Mothers were told that the project was about their everyday speech to children and were debriefed on the details after the final recording session. I made all the Katherine English recordings, and spent time before and after sessions building rapport with the families and familiarising the children with the recording equipment. I was usually present in the sessions operating the video camera, although sometimes when the participants appeared nervous or very aware of my presence I secured the camera to the tripod and left the area.

Activities. The content of the recordings is the mothers and children engaging in free play. They played with their own toys and games as well as those brought by the researcher, and generally the same types of activities were carried out across families and stages of recording: playing with coloured blocks, toy cars, dolls, puzzles and playdough, free drawing and painting, and playing on playground equipment such as swingsets and slides. Originally an effort was made to keep the materials as similar as possible to those in the Gurindji Kriol recordings; however, participants were more at ease choosing their own activities and materials. This was therefore allowed as a more naturalistic speech sample was elicited when participants were comfortable, and as we were interested in the speech the children were typically exposed to it was valid for them to engage in their typical activities. Between one and three sessions were conducted with each family at each stage, generally between half an hour and an hour in length, depending on the participants’ schedules and the children’s compliance (refer to

Table 3.4 for total amounts of recording). 61
CHAPTER 3

Equipment. The microphones were wireless, cardioid, condenser lapel

(Sennheiser ME104) that recorded at 48kHz, 24 bit via bodypack transmitter to a receiver with XLR connection to a hard disk recorder (Edirol R-44). The mother and child wore a microphone each and recorded on separate audio tracks. The mother’s track was used for transcription. The microphone was positioned upwards at the top centre of the mother’s shirt, and the children wore adjustable aprons with the transmitter clipped on the back and the microphone at the top centre. The video camera (Sony

AVCHD HDR XR-500V) filmed in high definition and had an external microphone attached. Figure 3.2 is a photo of a typical session setup, taken at Stage 2.

Figure 3.2. Photo of a typical recording session. The mother and child (2;0) are both wearing lapel microphones and play together in the garden at the child’s grandmother’s house. The grandmother participates in the background and the researcher is present behind the camera.

62
CHAPTER 3

3.3.4 Ethical Considerations Relating to the Katherine English Corpus

Creation of the Katherine English corpus involved several ethical considerations that informed how the research was conducted, so I will briefly discuss these and how they were addressed in the procedure.

Participants were recruited through Katherine childcare services and other local businesses. This recruitment procedure allowed me as the researcher to consult with the community and to build relationships with the local organisations that have an interest in the research. Before starting the Katherine English corpus I attended several childcare centres in operation and observed adult childcare providers interacting with children aged about 1;0 to 5;0, and spoke casually with parents and staff about the experiences of caring for children in Katherine. I also attended staff meetings of an organisation that provides early childhood and parenting programs to communities in and around

Katherine and gained awareness of common issues faced by families in the region.

These organisations also provided a channel for disseminating our findings to the community.

As the research involved collecting, analysing and storing audiovisual recordings of the families’ everyday lives, it was critical to ensure the mothers had an in-depth understanding of what we were doing with the recordings and transcripts, how they would be stored and who could access them, and that we complied with the families’ wishes. I discussed the information sheet and consent forms (Appendix A) in detail with mothers before the first session and during each of the subsequent fieldtrips to check if they had changed their minds on the consent options. These included various options for the storage of data, its current and future use, and levels of de-identification for each of the data formats (audio, video, photos and transcripts). Some mothers chose to give full permission for all options of our use and storage of all forms of their data 63
CHAPTER 3

without any anonymity, for themselves and their children. The other mothers chose de- identified options for online data storage and access (i.e. on the TalkBank website), and for pseudonyms to be used in reports and publications.

Families were sent DVDs compiled of the video recordings overlaid with the audio from the lapel microphones (with assistance from a technical research assistant on the project), as these were the raw data we processed and analysed for the research.

Mothers could then check if there was anything they would like deleted from the stored recordings. This occurred once when a mother requested deletion of a telephone conversation (usually the recording was stopped or the mother’s microphone muted for phone calls but it had kept recording in this case). Participation was voluntary and mothers were reminded at the start of each session that they could stop the recording at any time. They were paid $25 an hour for their time and were aware they would be paid regardless of whether sessions ended early or if they withdrew consent.

3.3.5 Phonetic Transcription

The mothers’ audio tracks were transcribed in the phonetic transcription and analysis program Phon (Rose et al., 2007). Phon was used to segment the tracks into utterances based on conversational turn-taking and pause boundaries (see Table 3.4 for the number of utterances per mother at each child age). Phonetic transcription of the audio recordings was performed by me, a native speaker of Australian English from regional Tasmania. I transcribed mothers’ actual pronunciation in the ‘IPA Actual’ tier in Phon as well as the standard Australian English forms in the ‘IPA Target’ tier. This allowed us to compare the mothers’ pronunciation with the standard forms and to search for phonological processes that involve systematic differences between the two, such as elision (the process investigated in Chapter 4). Further detail of the procedure of using 64
CHAPTER 3

Phon (Rose et al., 2007) for this purpose is in Buchan (2011), reprinted in the last section of this chapter (pp. 59-70).

For the phonetic transcription we used the transcription system recommended by

Harrington, Cox and Evans (1997) for phonetic transcription of Australian English, which is based on acoustic analyses of vowels from a large sample (N=132) of

Australian English speakers recorded as part of the Australian National Database of

Spoken Language (ANDOSL). A second transcriber, who has a Masters in Speech

Science and is experienced in Australian English transcription, separately transcribed a subset of the data for reliability analyses, which showed over 95% agreement.

Disagreements were resolved through consensus discussions between myself and the second transcriber, and where consensus could not be reached a third transcriber, the primary supervisor of this thesis (trained in phonetics and phonology and a researcher in child language), was consulted and a decision was reached.

Transcription was an ongoing process and was done in several settings (field locations and at the university), and the same set of high quality headphones

(Sennheiser HD 280 Pro) was used for all transcription with the second and third transcribers using identical sets. Further detail of the quality of the audio recordings and the reliability analyses are in Chapter 5 (pp. 136-137).

3.4 Working with Gurindji Community Members

Community Research Assistants (RAs) were also employed to contribute to this project. Community RAs were seven young women in Kalkaringi and Daguragu who were native speakers of Gurindji Kriol and recommended by the traditional owners and the community project liaison. Two of the three Gurindji Kriol speakers mentioned in the above description of the Gurindji Kriol subsample were also community RAs. The third speaker was away from the community during the times the research for this thesis 65
CHAPTER 3

was carried out, but we were able to contact her in Katherine and when she visited

Kalkaringi to discuss consent and to ask her permission to use recordings and transcripts of her. The community RAs were involved in a variety of tasks. They familiarised me with the community, the social context and their perspective on the language situation, and taught me some Gurindji Kriol as well as some traditional

Gurindji words. Community RAs also participated in the phonological awareness training and Visual Analogue Scale tasks described in Chapter 4, and through these procedures provided data on their perceptions as native speakers of fricatives in the

Gurindji Kriol corpus.

The study in Chapter 5 involved us developing a new method for using the

Visual Analogue Scales with native speakers of Gurindji Kriol to add to our phonetic transcription of Gurindji Kriol, and so the method is described in some detail in that research article.

3.5 Phonetic Transcription and Analysis in Phon8

3.5.1 Overview

Phon is an open-source program for the transcription and analysis of phonological and phonetic data. It was designed to help systematise research in children’s phonological development, but many functions in Phon, particularly the powerful search function, can be used for a wide range of investigations in phonetics and phonology. Phon is compatible with other language processing programs and is not

8 This section is a refereed published paper: Buchan, H. (2011). Phon: Free software for phonological transcription and analysis. Language Documentation and Conservation, 5 81-87. Acknowledgements: I would like to thank my supervisors Caroline Jones (University of Wollongong) and Felicity Meakins (University of Queensland) for their valuable suggestions. I am also very grateful to Yvan Rose (Memorial University) for his helpful feedback. 66
CHAPTER 3

just limited to English, making it a useful tool for documenting and analysing the phonological system of any spoken language.

The program has a user-friendly interface that makes transcription and analysis a relatively straightforward process. The layout is easily customised, as each major feature opens in a new view panel and so can be displayed as needed. Phon supports blind transcription, allowing multiple transcribers to work on the same file independently, and a validation feature allows the multiple transcriptions to be displayed together for consensus discussions and transcription validation. The media player supports both audio and video files, and media files can be segmented to link to sections of the transcription. Each transcript section is called a record, and once linked to the media the waveform of each record can be displayed. Phon uses Unicode fonts, and there are built-in dictionaries for some languages which facilitates transcription of the target phonological form. Although the built-in dictionaries are only of major languages, there is also a dictionary utility that allows user-defined dictionary input that will show the IPA transcription of target phonological forms. There is also an automatic syllabification feature that can be adapted for other languages. Perhaps most impressive are the program’s search capabilities. The search function permits systematic searches of phones and meta-phones, and the program developers have created PhonEx (Phon

Expressions), a language that allows for complex searches using common linguistic terms. PhonEx is logical and easy to learn from the manual, especially for researchers already familiar with linguistics terminology. The ability to search phonetic transcriptions for feature classes in specific phonological environments makes Phon a very useful program for describing and analysing phonetic features and phonological systems of spoken language. 67
CHAPTER 3

Another very useful feature of Phon is that transcripts can be imported and exported in CSV and XML formats. This makes it easy to archive transcripts, and to use

Phon in conjunction with other language documentation software. For example users can download the program Chatter, which is freely available online, to convert CHAT

(the CHILDES transcription format) transcripts into XML, and then use another freely available script, xml2phon, to import these into Phon for phonetic and phonological analysis9.

Phon is a good primary transcription tool, especially for phonetic transcription as the IPA dictionary function can substantially speed up transcription time. However, transcripts can also be converted to be compatible with programs like Transcriber and

ELAN. This means if a corpus has previously been transcribed in one of these programs it can then be imported into Phon for phonetic transcription and analysis10.

Similarly, search results can be exported in a variety of formats including

Microsoft Excel and Word, OpenOffice Calc and Writer, as well as more general PDF,

HTML and CSV formats, the latter especially suited for transporting Phon data into other applications and statistical analysis programs. In this way Phon’s unique search function can be used in conjunction with functions offered by other programs.

Phon was developed by a team led by Yvan Rose and Greg Hedlund as a part of the PhonBank project, a recent expansion of the CHILDES project (Child Language

Data Exchange System), an international database that supports research in language acquisition. Coordinated by Brian MacWhinney (Carnegie Mellon University), the system facilitates data sharing across the field of child language research. CHILDES is a component of TalkBank, a multilingual corpus containing shared databases from























































9 Chatter and xml2phon can both be downloaded for free from http://talkbank.org/ 10 Prof. Trevor Johnston at Macquarie University identified the potential use of Phon in this way. 68
CHAPTER 3

several subfields of human and animal communication research. Within this larger database, PhonBank offers corpora documenting phonological development in a number of languages. PhonBank data are compatible with the Phon and CLAN software programs.

Good support is available for Phon users. A PDF version of the manual is available online and in the program’s help menu. Users can ask for help and post feature requests in an online forum11 that is checked regularly by the developers, and there is also a Google discussion group for PhonBank12.

Phon has very recently been released in a new version (v1.4). This version is a substantial improvement on Phon 1.3, particularly in the layout and navigation. User feature requests have been incorporated into v1.4, and existing functions have been updated. There is a simple tool within the program for converting project files from v1.3 to v1.4. This review is of the newest version, Phon 1.4b975.

3.5.2 Getting Started

Projects are organised hierarchically, with each project made up of one or more corpora, and each corpus containing one or more sessions. At this stage the user needs to decide whether to work in the Editor’s mode or the Blind mode, the former being the default. Transcription done in Editor’s mode will not appear in the Blind mode, so it is important to make the decision about which mode to work in at the beginning of the project. The Blind mode enables blind transcription from multiple transcribers and should be checked at the start of the project if you would like to compare the phonetic transcription of two or more transcribers. This feature is useful when discussing

11 http://phon.ling.mun.ca/phontrac/discussion 12 [email protected] 69
CHAPTER 3

transcript discrepancies as part of a validation procedure, and also for documentation as it is a way to keep track of changes made through consulting with native speakers.

Transcription is done in the Session Editor. The layout here can be organised according to the user’s needs at the time. The displays (view panels) for different functions (e.g., media segmentation, automatic IPA lookup, search, search results) can be opened and moved around and docked or undocked as needed. Customisation of the layout is one of the strong points of Phon, as you only need to display the functions relevant to the specific task at hand (e.g. you can have a different configuration of view panels for transcribing, validating transcriptions, or searching). Figure 3.3 is an example setup for transcribing in Phon. The Record List on the left makes it easy to navigate between records, which are portions of the transcript that are useful for research, typically consisting of single utterances. The Session Information and Record

Data in the middle display the participants’ details and the record transcript, respectively. The IPA Lookup on the right displays the dictionary IPA transcription of the Record Data.

Figure 3.3. Example of a customised session editor display in Phon. 70
CHAPTER 3

3.5.3 Media

The media player in Phon supports a range of audio and video formats and can handle high quality recordings (I use 24 bit 48 kHz audio files, which work well in this version). Media files can be added by locating the file in the Session Information view panel, and played by opening the player from the Media menu. Media can be segmented to create records (though data records can also be created in the absence of recorded media.) The segmentation feature is simple to use. After entering in the participant information, the user can decide to segment media from the beginning of the recorded media file or from the end of the last record segmented. To segment, simply use either the keyboard shortcuts or click on a participant in media segmentation view panel, as shown in Figure 3.4. This will automatically create a record for that participant that links to the media segment. The default segment window (timeframe of an automatically-created segment) is 3000ms, and can be changed just by typing in another timeframe. I found the default 3000ms window to work generally well for naturalistic interactions between mothers and children. Segment times can also be easily adjusted in the segment tier of each record or by moving the boundaries in the waveform. The segmentation tool is also useful in that it allows the user to export individual media clips as separate files. This is good for storing soundbites, and is also very helpful when audio in the segment is unclear, as the user can then quickly open the individual segment in a speech analysis program such as PRAAT for closer inspection. 71
CHAPTER 3

Figure 3.4. Using the segmentation tool to create a record that is linked to the media.

3.5.4 Transcription

Once a record has been created, the associated utterance(s) can be transcribed.

The Record Data window in the example in Figure 3.5 shows the default transcript tiers. The Tier Management function allows the user to re-order and show or hide tiers, and to create new tiers. There is an IPA chart, available from the Tools menu, for insertion of phonetic characters, suprasegmentals and diacritics.

At this stage the user should decide on how to group words. Word groups in

Phon refer to user-defined boundaries and are useful for defining boundaries of a lexical word, e.g., should, or a prosodic word, e.g. oughta. Word groups are represented by the square brackets and can be created in the Orthography tier, which then automatically creates groups in the IPA Target tier (transcription of the model or standard utterance form) and the IPA Actual tier (transcription of the produced utterance form), and any user-defined tiers that have been created to align with word groups. This feature is a way in which the transcript structure can be adapted to specific requirements of the research as it allows utterances to be segmented at any level, so groups can be made for prosodic phrases, or morphemes, etc., depending on the level of analysis. 72
CHAPTER 3

The IPA target refers to the model phonological form, and the search function in

Phon can make systematic comparisons between target and actual phonological forms.

This is a very useful feature that can be applied, for example, to compare phonemic and phonetic transcriptions, standard and colloquial pronunciations, adult forms with actual child pronunciations, or standard and baby-talk words.

The target phonological forms can be looked up in a built-in dictionary and inserted into the transcript. Dictionaries are already built-in for some major languages

(American English, Catalan, German, French, Icelandic, Dutch, Italian and Spanish), and can be imported for other languages. A fieldworker could therefore create a word list with IPA transcriptions of the most frequent words in a corpus and build it into Phon as a basic dictionary. The phonological word forms can then be displayed in the dictionary look-up view panel and inserted into the transcript simply by clicking on each word form. If a word in the dictionary look-up has more than one IPA entry then the arrow keys can be used to scroll through the possible alternatives (for example you can choose between [ət] and [
t] for the English word it). Combined with the waveform feature, which allows the user to isolate segments in an utterance, this feature is an efficient way for a fieldworker to check transcriptions with native speakers – the user can play a segment and go through potential phonological forms for that segment simultaneously.

Phon also has tiers for target and actual Syllabification and Alignment, which offer syllable-level data annotation as well as systematic phone alignment between the

IPA Target and IPA Actual tiers (see Figure 3.5 for an example). Syllable information is automatically labelled and displayed graphically in the transcript with boxes around each syllable and different colours to represent syllable constituents. The automatic syllabification can be edited manually by simply clicking on a syllable constituent and 73
CHAPTER 3

selecting the correct label from a drop-down list. The syllabification feature does not require the use of stress symbols in transcription, and provides a quick and easy way of seeing how utterances are segmented into syllables and the alignment between the IPA

Target and IPA Actual tiers.

Figure 3.5. Example of Phon’s syllabification and alignment feature.

3.5.5 Search

The search function in Phon is excellent and allows for advanced queries. While this is just an overview of the main search features, searching in Phon can be adapted to be as complex and detailed as necessary. Searches can be performed both within and across transcripts, and results can be filtered by speaker. Tiers can be queried individually using the Data Tiers search (e.g., to search for all instances of a particular word in the Orthography tier), or multiple tiers can be queried at once (e.g., to compare 74
CHAPTER 3

segments in the IPA Target and IPA Actual tiers). Phon also allows for searches of stress patterns and syllable types, i.e. consonant, vowel and glide combinations. The results can be exported in the variety of formats listed above, making it easier for further analysis. For example, CSV report exports are useful for data post-processing in statistical analysis programs.

The search function is so good because it has been specifically designed for phonological analysis. There are tools for searching vowel and consonant harmony and consonant metathesis, in which the user can select the type of harmony and list the features and directionality of the process. As well as using plain text and regular expressions, complex queries can also be made using PhonEx, a language built for Phon using common linguistic terms. The structure of and terms in PhonEx are described in detail in the manual and it does not take long to learn how to create queries. PhonEx allows search expressions to be as broad or specific as needed. For example you can search for all instances of word-initial [k], all vowels in unstressed syllables, dorsal sonorants in primary stressed syllables in onset position, one or more phones that are not in coda position and occur at the end of a word group, etc. The output of searches gives summaries of the total number of instances in a transcript as well as a detailed report with the record numbers for each instance (if you search within a session you can even click in the results box to bring up the records in which each instance occurs). The advanced search system is what really sets Phon apart from other phonetic transcription software. It is not as complicated as it may appear at first, results are easy to navigate, and it allows for a very wide range of complex searches depending on the researcher’s needs.

75
CHAPTER 3

3.5.6 Conclusion

Phon is a very effective program for transcription and analysis of phonological and phonetic data. Transcribing in Phon is straightforward, and the media segmentation and dictionary IPA look-up features help to speed up the transcription process. The powerful search capabilities make it possible to describe the phonological system of a language variety in great detail. Phon would be a good asset to a great variety of investigations in phonetics and phonology. The range of useful features and customisable layout make Phon a very handy tool for fieldworkers interested in describing phonology of underdocumented languages.

Primary function: Transcription and analysis of phonological and phonetic data

Pros: Useful interface for phonetic transcription (includes support

for blind transcriptions) and annotation; support for media

segmentation and transcript linkage; powerful search

function for complex searches of many transcripts at once;

importing and exporting functions makes it compatible with

other transcription and analysis programs; compliance with

Unicode font encoding; good support is available from

program developers

Cons: None significant

Platforms: Mac OS X 10.4 and higher, Windows XP and higher

(Java 1.6 compatibility required on all platforms)

Open Source: Yes, Phon is licensed under the GNU General Public License

Version 2. Phon can be downloaded for free from

http://phon.ling.mun.ca/phontrac/wiki/Downloads 76
CHAPTER 3

Proprietary: Users can use Phon under the GNU-GPL license.

Reviewed version: Phon 1.4b975

Application size: Mac OS X: 41M; Windows XP/Vista/7: 95M (includes VLC

1.10 for the media player)

Documentation: http://phon.ling.mun.ca/phontrac/wiki/

3.6 Chapter Summary

This chapter provided an overview of the methodology used for the empirical research in this thesis. It included descriptions of how the project was conducted, the

Gurindji Kriol and Katherine English corpora, and how data were processed, transcribed and analysed in the program Phon (Rose et al., 2007). There were three main aims of the field trips to Kalkaringi and Katherine in the Northern Territory, Australia, that I undertook between 2009 and 2011: 1) Recontact three Gurindji Kriol mothers whose families were part of the Meakins (2011) corpus and discuss our use of their audiovisual recordings and transcripts for this project, and consult with Kalkaringi community leaders and traditional owners about the project; 2) Create a comparable longitudinal audiovisual corpus of family interactions from native speakers of Australian English in

Katherine; 3) Conduct phonological awareness training with community Research

Assistants, and use Visual Analogue Scales to elicit their native speaker judgements of

Gurindji Kriol fricatives.

Phonological analyses of corpus data conducted at the level in this research are possible due to advanced computer programs for phonetic transcription and analysis.

For this project we used the software Phon (Rose et al., 2007), which is a component of

CHILDES and compatible with the CHAT transcription format (MacWhinney, 2000). 77
CHAPTER 3

The last section in this chapter (previously published, see Buchan [2011]) provided a review of Phon and a guide for using it in the context of this research.

The next three chapters describe three empirical research studies and are written in the format of journal articles. Study 1 (Chapter 4) is an investigation of fricative variation in Gurindji Kriol maternal speech to children at three child ages, approximately 1;6, 2;0 and 2;6. In Study 2 (Chapter 5) we investigated fricative variation in the form of phonological reduction in Katherine English maternal speech to children at approximately the same ages, 1;6, 2;0 and 2;6. Study 3 (Chapter 6) offers a further interpretation of the Study 1 findings, by examining native speaker perceptions of fricative variation in Gurindji Kriol.

CHAPTER FOUR

Fricative Variation in Maternal Gurindji Kriol

CHAPTER 4 79

4.1 Abstract

The majority of traditional Australian languages have relatively few contrastive manners of articulation and lack phonemic fricatives (Butcher, 2006). In Kriol varieties fricative production is highly variable and many Kriol words that have fricatives in their

English cognates can be pronounced variably with a fricative or stop (Sandefur, 1991).

Speakers therefore often have multiple phonetic forms of words to choose from, and in child-directed speech adults may either consciously or unconsciously choose forms that support children’s language acquisition.

This study investigated fricative variation in mothers’ speech in Gurindji Kriol, a mixed language variety in northern Australia that contains lexical forms from both the traditional language Gurindji and Kriol. We phonetically transcribed and analysed data from the Meakins (2011) longitudinal Gurindji Kriol ACLA corpus to examine stop- fricative variation in Kriol-derived words. Speakers were three Gurindji women at three timepoints, when their children were aged approximately 1;6, 2;0 and 2;6. To quantify environments in which fricatives could occur in Kriol-derived words, we searched the transcripts for words that had fricatives in the English cognates. Words were categorised and analysed according to whether they always contained a fricative, were never pronounced with a fricative, or contained stop-fricative variation. We analysed the characteristics of words in each category, and the relationships between fricatives in

English cognates and corresponding pronunciation in Gurindji Kriol. Logistic regression models were performed for word-initial, -medial and -final positions to investigate the relative effects of child age and phonological environment on stop- fricative variation.

Results showed that words containing stop-fricative variation were more likely to be open-class (88%) than closed-class, and the variable segments were most CHAPTER 4 80

frequently word-initial (49%), then word-medial (32%) and word-final (19%). The most frequent places of articulation were labio-dental (44%) and alveolar (32%). Cross- tabulations of fricatives in English cognates and corresponding pronunciation in Kriol- derived Gurindji Kriol words showed variability across place, manner and voicing, and differences by word position. Logistic regression results indicated that in initial and medial positions fricative variation in mothers’ speech changed with child age, taking into account phonological environments and inter-speaker variation. Across all words, word-initial fricatives in mothers’ speech were significantly more likely at child age 2;6 than 1;6. In words containing stop-fricative variation, word-medial fricatives were significantly more likely at 2;6 than 2;0 and 1;6. Results suggest stop-fricative variation changed in Gurindji Kriol maternal speech as children aged, partly due to changes in lexical choice. Findings are discussed in relation to lenition processes and fine-tuning in child-directed speech. CHAPTER 4 81

4.2 Background

Gurindji Kriol is the main language spoken by children and adults under the age of 40 in Kalkaringi, an Aboriginal community in northern Australia, and is the language of acquisition. It is a mixed language that systematically combines grammatical structures and lexical items from the traditional Aboriginal language Gurindji, and from

Kriol, an English-lexifier creole spoken across northern Australia (Meakins, 2011, p.12). The grammar of Gurindji Kriol has been examined in detail (Meakins, 2009,

2010, 2011) and the phonological inventory has been briefly described (Meakins, 2007;

Meakins, in press). The current study contributes to the systematic quantitative description of the phonetics and phonology of a corpus of Gurindji Kriol maternal speech13, with the overarching aim of investigating the phonology of the speech input

Gurindji children are exposed to at Kalkaringi. In this study we analysed segmental fricative variation in family interactions and examined possible factors that may affect fricative variation and whether it changes over time in mothers’ speech as their children get older. Speakers were young Gurindji Kriol speaking women.

Most Australian languages do not have a manner contrast between stops and fricatives and lack phonemic fricatives (Butcher, 2006). Lenis stops will sometimes have fricative allophones, particularly voiced stops following long vowels, and in languages that do contain phonemic fricatives they are generally voiced and contrast with voiceless stops (Dixon, 2002). Phonetic fricatives are often dorsal and/or labial, and common word positions are initial and medial (Dixon, 2002). Traditional Gurindji does not contain a stop-fricative contrast and has allophonic fricatives (e.g. the

13 The Gurindji Kriol corpus analysed in this study was made by Meakins (2004-07) for the Aboriginal Child Language (ACLA-1) project, funded by Australian Research Council grant DP034189 through the University of Melbourne (C.I. Gillian Wigglesworth, Jane Simpson and Patrick McConvell). http://linguistics.unimelb.edu.au/research/projects/ACLA/ CHAPTER 4 82

intervocalic velar stop in tarukap ‘swim’ is sometimes pronounced as a voiced velar fricative). The nature and extent of fricative production in Gurindji and Gurindji- derived words in Gurindji Kriol has not been investigated. The current study focuses on the fricatives in Gurindji Kriol that are in Kriol-derived words. The production of these fricatives is highly variable. They are known to occur at three places of articulation – bilabial, alveolar and palatal – and are used in variation with stops (Meakins, in press).

There are no previous systematic quantitative descriptions of fricatives in Gurindji Kriol or other Australian contact language varieties. In the present study we focus our analyses on fricatives in Kriol-derived words in Gurindji Kriol that have fricatives in their English cognate words, as a basis for quantifying fricative variation.

In this study we examine whether fricative variation in manner of articulation is influenced by lexical and phonetic environments, and also how variation in mothers’ speech changes as their children get older. Evidence from the child-directed speech literature (mostly on standard language varieties) suggests that mothers make segmental modifications in their speech to children, and that modifications may change as children get older and their receptive and productive language develops. Cross-linguistically fricatives tend to be acquired by children later than stops or glides, and it has been suggested that this is due to the greater articulatory difficulty involved in producing fricatives (Dinnsen, 1992, pp.196-199; Kent, 1992, pp.75-76). Thus if mothers make phonetic modifications to ‘fine-tune’ their speech to children’s language development, there may be an effect of child age on fricative variation in mothers’ speech. Qualitative reports from Gurindji women suggest that fine-tuning may contribute to phonetic variation in their speech to children. They have reported using more stops than fricatives when speaking to children as fricatives are considered more difficult sounds

(Jones & Meakins, 2012b). There are, however, no previous quantitative descriptions of CHAPTER 4 83

fricatives in Gurindji Kriol and it is unknown if these impressions are actually realised in mothers’ speech and at what child ages the effect may appear.

The following section provides some background on the context of Gurindji

Kriol, which is followed by more detail on how child age may affect phonetic variation in maternal speech. We then discuss linguistic factors that may contribute to fricative variation in Gurindji Kriol, in particular lenition processes.

4.2.1 Gurindji Kriol

Gurindji Kriol is spoken by Gurindji people in Kalkaringi in northern Australia.

It developed from code-switching between Gurindji and Kriol in the 1970s (McConvell

& Meakins, 2005; McConvell, 2008; Meakins, pp.109-53) and contains a structural split between the noun phrase system, which is mostly from Gurindji, and the verb phrase system, mostly from Kriol (Meakins, 2011, pp.12-13). The lexicon contains nouns and verbs derived from both languages and Gurindji Kriol contains phonology of both

Gurindji and Kriol (Meakins, in press). Gurindji Kriol is the main language of Gurindji people, though the speech community is multilingual. Traditional Gurindji is spoken by some older people, and Kriol is spoken with Kriol-speaking visitors to the community.

The neighbouring language Warlpiri is also spoken by some people. English is the language of the school, government services and media and is spoken with English- speaking visitors. English is also occasionally present in code-switches, particularly in phrases that often occur in the school context (Meakins, 2008; Meakins, 2011, p.65), but Gurindji Kriol is the children’s home language and they generally do not hear much

English in the years before school. See Meakins (2008; 2011, pp.58-70) for more information on the language context and ecology in the region.

The current study is part of a larger project investigating the phonetics and phonology of maternal Gurindji Kriol. The aim of the project is to investigate the sound CHAPTER 4 84

patterns that children are exposed to in their home language Gurindji Kriol, and involves phonetic analyses of a large longitudinal audiovisual corpus made by Meakins between 2004 and 2007 as part of ACLA-1 (see footnote 1). Acoustic analyses of mothers’ speech from this corpus and analyses of minimal pair contrasts in Kriol- derived words suggests Gurindji Kriol may have a 5 vowel system, /ɪ, ɛ, ɐ, ɔ, ʊ/ (Jones,

Meakins & Buchan, 2011; Jones, Meakins & Muawiyath, 2012), and voicing variation in stop consonants depending on word position and place of articulation (Jones &

Meakins, 2012a). Fricatives in Gurindji Kriol have not previously been studied in detail.

The fricatives under investigation in this study are in Kriol-derived words.

Research on the phonology of Kriol has found that fricative production is highly variable. Sandefur (1991) suggests that when Kriol was developing, speakers produced

English-derived words with traditional language phonology, which did not contain fricatives, and as speakers remained in contact with English some Kriol words started to be pronounced with the original English sounds. He also suggests that the newer

English-like pronunciations were produced in addition to the older pronunciations rather than replacing them, which is why Kriol words often have several pronunciations

(Sandefur, 1991) and accounts for the high variability in fricatives, which are present in the phonology of English but not traditional languages in the region.

Impressionistically, Kriol-derived words in Gurindji Kriol are also highly variable, and

Meakins (2007, p.358) notes that there is variation in manner of articulation between stops and fricatives.

Fricative variability in Gurindji Kriol may be affected by a number of factors, including situational contexts and linguistic factors such as word position and phonological environment. The settings of the current dataset are family interactions, as the broad aim is to investigate the phonology of the input that children acquiring CHAPTER 4 85

Gurindji Kriol are exposed to. Thus the data in the current study is maternal speech in

Gurindji Kriol. Maternal speech is defined here as mothers’ speech spoken in the presence of their children and includes child-directed speech, that is speech directed specifically to the child (see Soderstrom, 2007, for a discussion of maternal speech as linguistic input to children). In the next section we discuss previous research on phonetic variation in child-directed speech and how fricative variation in maternal

Gurindji Kriol may be affected by child age.

4.2.2 Phonetic Variation in Child-Directed Speech

Caregiver speech to children (termed infant- or child-directed speech) contains phonological differences to interadult speech, including slower tempo, more variability in pitch contours, longer pauses, and higher fundamental frequency (Snow, 1977, p.36;

Fernald et al., 1989; Katz, Cohn & Moore, 1996). Cross-linguistic evidence suggests that phonetic variation in speech input to children may be ‘fine-tuned’ (consciously or not) to children’s receptive and productive language development.

Longitudinal studies show that the phonological modifications mothers make to their speech change as children get older. For example, in Mandarin acoustic exaggeration in vowels, i.e. larger vowel space, longer durations, and increased tone range, was greater in adult speech to younger children (aged 0;7-1;0) than to older children (aged 5;0), and vowel exaggeration in speech to the older children was still greater than in adult-directed speech (Liu, Tsau & Kuhl, 2009). Similarly in a regional variety of English from northern England, segmental phonetic variants associated with non-standard or vernacular speech were found to gradually increase in child-directed speech between child ages 2;0 to 4;0 (Foulkes, Docherty & Watt, 2005). Phonological reduction in mothers’ speech related to casual speech processes has also been found to change with a child's age in northern Australian English, with mothers proportionately CHAPTER 4 86

increasing segmental deletion as children aged 1;6 to 2;0 and then decreasing deletion between 2;0 and 2;6 (Buchan & Jones, in press), suggesting mothers may fine-tune their speech differentially as their children age, possibly to children’s receptive and productive language development.

Mothers’ fricative production may also vary between child- and adult-directed speech, due to either lexical choice or mothers fine-tuning to children’s productive language. Distributional differences between child-directed and adult-directed speech in manner of articulation have been found in English (Lee, Davis & MacNeilage, 2010) and Korean (Lee & Davis, 2008). English child-directed speech to children aged 1;0 contained proportionately more stops and glides, and fewer fricatives, affricates, liquids and nasals than adult-directed speech, partly due to different lexical items used in each register. In Korean fricatives were also significantly less frequent in child-directed speech compared with interadult speech. These results were interpreted to suggest that mothers may ‘match’ their speech to their child’s language capabilities, scaffolding children according to the sounds they are acquiring (Lee et al., 2010). The difference in lexical items used in child-directed and adult-directed speech further suggests that the difference in fricative distributions is at least partly due to mothers avoiding words containing fricatives and/or favouring words containing stops and glides.

Fricatives are more difficult to produce and tend to be acquired later than stops and glides (Dinnsen, 1992, pp.196-199; Kent, 1992, pp.75-76), therefore if mothers modify their speech according to the sounds children are acquiring (Lee & Davis, 2010) we would expect a relationship between child age and fricative variation in mothers’ speech. Gurindji Kriol is an ideal variety in which to investigate fricative variability as it occurs both within and between lexical forms. Speakers may have a choice of CHAPTER 4 87

alternate phonetic forms and opportunity to avoid fricatives in speech to children and/or promote ‘easier’ manners of articulation such as stops and glides.

The current study is an investigation of input children are exposed to, thus the data consists of maternal speech, which is a broader term than child-directed speech

(Soderstrom, 2007) that includes mothers’ speech in family interactions when the child is present rather than only speech directed to the child. Studies on English suggest that a substantial proportion of speech young children are exposed to is not necessarily directed specifically to them (van de Weijer, 2002) and that exposure to third party conversations may facilitate children’s lexical development (Akhtar, 2005). Gurindji children are part of large extended family groups and the speech input they are exposed to consists of general family interactions. Child age is one factor that may contribute to fricative variation in maternal Gurindji Kriol, but variability is also high in interadult speech and in this study we also examined linguistic factors that potentially affect fricative variation.

4.2.3 Lenition Processes and Fricative Variation

The focus of this study is on fricative manner variability, particularly stop- fricative variation. Spirantization, a lenition process whereby the manner of a sound changes into a fricative (e.g. [t]  [θ]), may play a role in this type of variation.

Lenition is a sound change process involving consonant ‘weakening’, but exactly what constitutes weakening is debated in the literature (e.g. Bauer, 2008; Honeybone, 2008;

Kingston, 2008; Kirchner, 2005). Traditional definitions are based on hierarchies or trajectories of sounds from strong to weak (e.g. Escure, 1977; Hock, 1986), and focus on the acoustic and articulatory features of the sound change such as sonority (i.e. changing from less sonorous to more sonorous), or a decrease in resistance to airflow between articulators. CHAPTER 4 88

One explanation of lenition is that it is a reduction in the speaker’s articulatory effort whereby the phonetic target is not reached (e.g. Kirchner, 2004). In spirantization the reduced effort leads to frication of stops, that is, allowing turbulent airflow through the articulators rather than stopping it (Bauer, 2008). Kirchner (2004, p.316) suggests certain contexts are more susceptible to lenition, in particular fast/casual speech and intervocalic segments, because in these contexts greater articulatory effort is required to produce a constriction. Stop-fricative variability may therefore be affected by articulatory effort, which is in turn influenced by the speech context and phonological environment.

The reduced articulatory effort explanation is debatable. While fricatives are considered ‘weaker’ sounds than stops in that there is more airflow between the articulators, they are also more difficult to articulate in that producing fricatives requires more precision and muscular control than stops (Ladefoged & Maddieson, 1996, p.137).

As mentioned in the section above, the articulatory difficulty of fricatives is reflected in children’s acquisition as they are generally acquired later than ‘easier’ manners of articulation (Dinnsen, 1992, pp.196-199; Kent, 1992, pp.75-76). Further, difficulty of articulation may vary with the position of the segment within a word. Sole’s (2010) aerodynamic study of English speakers found that coda fricatives were articulatorily more difficult to produce than onset fricatives, and suggests this may account for word- final fricative weakening or elision of syllable-final fricatives that is found across many languages.

Lenition processes may be further influenced by perceptual factors. Kingston

(2008) found that the immediate phonological environment, that is segments either side of the lenited phone, influenced consonant lenition. He posited lenition may be driven by the speaker modifying the perceptual saliency of the segment and those surrounding CHAPTER 4 89

it. A perceptual explanation of lenition was also examined by Kaplan (2011), who suggests one of the speaker’s goals in lenition is minimal perceptual salience of the sound change. Kaplan (2011) found that intervocalically the distinction between voiced and voiceless stops was more perceptually salient than the distinction between voiced stops and voiced fricatives, suggesting listeners may be less sensitive to intervocalic stops being lenited to fricatives. The magnitude of the effect, however, was smaller for labials than dorsals and coronals, despite labial being a common place of articulation for frication across languages (Kaplan, 2011). Perceptual factors may therefore account for some but not all lenition processes, and both perceptual and articulatory factors may contribute to some phonetic variation.

Among Australia languages lenition processes can explain some continuant-stop variation, for example in Yolngu varieties Gaalpu (Chong, 2011), Djambarrpuyngu

(Wilkinson, 1991) and Djapu (Morphy, 1983). In a recent study of Gaalpu, Chong

(2011) found phonological context was a factor in the conditioning of lenition. Suffix- initial stops were lenited to semi-vowels preceding a vowel and following a liquid, semi-vowel or vowel.

4.2.4 The Current Study

The high variation in fricatives in Gurindji Kriol may be due to a number of potentially interacting factors. In addition to lenition processes, the input that children are exposed to is dynamic and may also contain variability as a result of mothers avoiding or promoting fricatives differentially over time as they fine-tune to children’s productive and receptive acquisition. The aim of this study is to quantitatively describe stop-fricative variation in Gurindji Kriol maternal speech and to examine the nature of variation, particularly how it changes over time in mothers’ speech as their children age.

The current study investigated three research questions: 1) What is the nature and extent CHAPTER 4 90

of fricative variation within Gurindji Kriol lexical forms in maternal speech? 2) Within lexical forms containing stop-fricative variation, what are the relationships between fricatives in English cognates and their pronunciation in Gurindji Kriol words? 3) How does fricative variation within and between Gurindji Kriol lexical forms change in mothers’ speech as children get older, taking into account phonological environment and word position?

4.3 Method

4.3.1 Design

The data in this study is a subsample from a longitudinal corpus of family interactions in Gurindji Kriol collected by Meakins (2011, pp. 45-47) for the Aboriginal

Child Language (ACLA-1) project. From the subsample we selected all tokens of words that could contain a fricative and analysed the IPA transcription of these tokens and their phonological environments.

To analyse fricative variation in naturalistic data we first had to determine the environments in which fricatives could occur. As there are no ‘standard’ pronunciations or known phonological targets in Kriol words and therefore no model forms to which actual pronunciation could be compared, fricatives in English cognates were used as a basis for quantifying where fricatives could potentially occur in Kriol based words in

Gurindji Kriol. From hereon we use the term ‘potential fricative’ to refer to segments in

Gurindji Kriol words that are pronounced as fricatives in the English cognates and can therefore be pronounced as fricatives or as something else in Gurindji Kriol. This approach allowed us to examine not only the environments where fricatives and stop- fricative variation occurred, but also the environments where fricatives did not occur, i.e. where there were only stops despite the English cognate containing a fricative, providing further insight into the factors that may be influencing variation. The current CHAPTER 4 91

dataset contains tokens of ‘potential fricatives’. The dependent variable was manner of articulation of the potential fricative: fricative or stop.

4.3.2 Speakers

Speakers in the subsample used in this study were three young Gurindji mothers who are native speakers of Gurindji Kriol. We selected recordings from the larger audiovisual database (Meakins, 2011, pp.45-47) that were made at three timepoints: when the mothers’ children were aged on average 1;6, 2;0 and 2;6. The children were two boys and one girl.

4.3.3 Recordings

The audiovisual recordings contain conversational maternal speech. Recordings are of family interactions in naturalistic settings, that is in natural outdoor settings or engaging in free play. Speakers wore a lapel microphone that recorded at 44 kHz 16 bit to a minidisc recorder that was worn by the speaker. Sessions were also recorded with a video camera that had a shotgun microphone. Further detail on the corpus and recording procedures are in Meakins (2011, pp.45-49).

4.3.4 Analysis

The mothers’ speech was phonetically transcribed in Phon (Rose, Hedlund,

Byrne, Wareham & MacWhinney, 2007) using IPA. Phonetic transcription was done by the first and second authors and two research assistants, all native speakers of

Australian English trained in phonetic transcription. We then queried Phon for the lexical items whose English cognate words contain fricatives and exported all tokens of these words spoken by the three mothers along with the immediate phonological environments and indexical information.

A total of 2,070 tokens were exported from Phon. Words containing [h] in their

English cognates were then excluded (N=474), as in Australian languages [h] behaves CHAPTER 4 92

differently to other fricatives and would have skewed the results: in the current dataset only 9 of the potential [h] tokens (0.02%) were transcribed as being pronounced with a

[h]. For the purpose of analysing variation between stops and fricatives we also excluded cases where the potential fricative was pronounced as a glide (N=92), most of which were word-initial (N=72). Excluded glides generally occurred intervocalically in unstressed words, for example dat ‘the,

4.3.5 Coding

Characteristics of the potential fricatives in tokens were coded for further analysis. Data included the IPA transcription, corresponding sound in the English cognate, word position, within-word phonological environment in IPA, and flanking segments from adjacent words (i.e. the phones preceding word-initial potential fricatives and the phones following word-final ones). Binary logistic regression analyses were performed separately for each word position, initial, medial and final. For all three models the actual pronunciation of the potential fricatives was dichotomized into two categories, fricative or stop. Phonological environments of segments were coded separately for each word position according to their sound class.

There were a small number of cases (3 word-medially, 12 word-finally) where phones in the phonological environment were ‘absent’ as compared to the English cognate. These are all cases of consonant clusters in the cognate where the ‘potential fricative’ in Gurindji Kriol is pronounced as a fricative but the other consonant is absent. Where this occurred the whole cluster was represented as the cognate segment.

This was done so that differences in consonant clusters between the cognate and CHAPTER 4 93

Gurindji Kriol pronunciation could be examined while the actual pronunciation of the phonological environment was still accurately represented.

Some potential fricative phones were also coded as ‘absent’ for the purpose of comparison with fricatives in English cognates. Because target forms in Gurindji Kriol are unknown, ‘absent’ in this case does not refer to elided segments but rather phones that are not present in relation to the corresponding segment in the English cognate, particularly where a consonant cluster in the cognate is pronounced as a single consonant in the Gurindji Kriol form. For example the word-initial cluster [st] is generally pronounced as a cluster in some Gurindji Kriol words, e.g. stori ‘story’ [stɔɹɪ] and as the alveolar stop in other words, e.g. dop ‘stop’ [dɔp]. In the latter case the potential fricative initial [s] was coded as absent for comparison purposes.

4.3.6 Transcription Reliability

To check transcription reliability a second transcriber who is experienced with

Kriol blind-transcribed 10% (N=202) of tokens and the phonological environments of these tokens. Tokens in the reliability analysis were a stratified sample of the dataset, randomly selected from each word position and place of articulation of potential fricatives. The second transcriber listened to the whole utterance and transcribed in IPA the token indicated on a spreadsheet, as well as the preceding phone for word-initial potential fricatives and the following phone for word-final ones. Percentage agreement between the two transcriptions was calculated for features of the potential fricatives and class of preceding and following phonological environments.

The breakdown of agreements and disagreements for transcription of potential fricatives is in Table 4.1. There was 92% agreement (N=185/202) on whether a phone was a stop or a fricative. In four of the 17 stop-fricative disagreements the second transcriber had noted uncertainty about whether the phone was a stop or a fricative. The CHAPTER 4 94

other manner of articulation discrepancies were between palatal stops and affricates, intervocalic stops and glides, and word-initial deletions, which generally reflected differences in transcribing assimilation with the preceding phone. ‘Full feature agreement’ (63%) refers to when both transcribers used the same IPA symbol for the potential fricative, that is both transcribers agreed completely on all features of the phone including voicing, manner and place.

There was also high agreement on transcription of phonological environments:

93% agreement on the preceding phone and 98% on the following. Phonological environment disagreements generally involved pause boundaries, that is, whether the environment was coded as a pause or a phone. There was one stop-fricative disagreement, for a following phone.

Table 4.1 Number and percentage of feature agreement on ‘potential fricative’ segments between two transcribers

% N agree N disagree agreement Full feature agreement 128 74 63%

Voicing 186 16 92%

Place 196 6 97%

Manner 153 49 76%

stop-fricative 185 17 92%

palatal stop-affricate - 20 -

stop-glide-deletion - 12 -

Unclear token - 3 -

CHAPTER 4 95

4.4 Results

4.4.1 Analysis 1: Fricative Variation Within Gurindji Kriol Lexical Forms

To examine the nature and extent of fricative variation within Gurindji Kriol lexical forms, we analysed the Kriol-derived words in the sample that contained fricative variation in manner of articulation (i.e. variability between fricatives and stops or affricates) and compared them to the words where the potential fricative was not pronounced with stop-fricative variation. Words were categorised into four groups according to the Gurindji Kriol pronunciation of the potential fricative: 1) only pronounced as a stop or affricate, 2) only pronounced as a fricative, 3) pronounced variably as a stop/affricate or a fricative, 4) only one token in the dataset and variability could not be determined. The latter group (N=57) were excluded from the current analysis. The list of words in each of the three groups analysed is in the appendix.

Tokens that were English code-switches were included in these analyses as they form part of the linguistic input children are exposed to. Words involving code-switches are in bold in the appendix, which shows that code-switched words containing potential fricatives were nearly always pronounced with a fricative.

Words containing multiple potential fricatives were coded for each word- position. For example dij ‘this’ was coded twice, for word-initial and -final positions.

There were 28 types of ‘all stop’ words where the potential fricative was always pronounced as a stop or affricate in the sample, 41 types of ‘all fricative’ words where it was always pronounced as a fricative, and 58 types of ‘variable’ words containing stop- fricative variation. These three lexical form categories were then analysed by word class, word position and place of articulation of the potential fricative, and by overall word frequency. In these categories some cells had few (< 5) observations, therefore inferential statistics were not run and these results are presented descriptively. CHAPTER 4 96

Word class. Words were broadly classified into open-class and closed-classed words. Figure 4.1 shows the proportions of open and closed class words within the ‘all stops’, ‘all fricatives’ and ‘stop-fricative variation’ words. In words where the potential fricative was always a stop or affricate, just over half of them were closed-class words

(54%), such as dat ‘the

The majority of words containing stop-fricative variation were open-class words (88%) such as baldan ‘fall’and juka ‘sugar’.

100%

90%

80%

70%

60%

50%

40% closed-class open-class 30%

20%

Percentage pronounced in each word class 10%

0% all stops all fricatives stop-fric variation Pronunciation of 'potential fricative' within lexical forms

Figure 4.1. Proportions of closed-class and open-class words within lexical form categories

CHAPTER 4 97

Word position. Figure 4.2 shows the proportions of the word position of the potential fricatives in words where it was always pronounced as a stop, always as a fricative, and words where there was stop-fricative variation. In words that were never pronounced with fricatives, i.e. the potential fricative was always a stop or affricate,

32% of potential fricatives were word-initial, 39% word-medial, and 29% word-final. In words that were always pronounced with a fricative, the fricative was most frequently word-initial (62%), followed by word-medial (24%), and then word-final (14%). In words containing stop-fricative variation, 49% of the potential fricatives were in word- initial position, 32% were word-medial, and 19% were word-final. Thus stop-fricative variation within lexical forms occurs in all word positions, though it appears to be most frequent word-initially and least frequent word-finally.

100%

90%

80%

70%

60%

50% final 40% medial 30% initial

20%

Percentage pronounced in each word position 10%

0% all stops all fricatives stop-fric variation Pronunciation of 'potential fricative' within lexical forms

Figure 4.2. Proportions of word positions of potential fricatives within lexical form categories CHAPTER 4 98

Place of articulation. The proportions of places of articulation of the fricatives in the English cognates across the three categories of lexical items are shown in Figure

4.3. This description is of the place of the fricative in the cognates, as several of the variable words involved place of articulation variation. More detail on the place of articulation of fricatives pronounced in Gurindji Kriol is given in Analysis 2. In words where the potential fricative was always pronounced as a stop or affricate, the cognate place of articulation was most frequently dental (46%) e.g. the medial stop in jamting

‘something’, and labio-dental (36%) e.g. abim ‘have’, with some alveolars (18%) e.g. mawuji ‘mouse’ and no palato-alveolars. In words where the potential fricative was always pronounced as a fricative, the fricatives in the cognates were most frequently alveolars (67%) e.g. stori ‘story’, followed by labio-dentals (21%) e.g. fait ‘fight’, with few palato-alveolars (10%) e.g. shade ‘shade’ and just one dental (2%), three, which is an English code-switch. Words containing stop-fricative variation had the highest proportion of labio-dentals (44%), e.g. baind ‘find’. Within word stop-fricative variation was also frequent in sounds that in cognates were alveolars (32%) e.g. medial consonant in dijei ‘this way’, and also occurred in palato-alveolars (17%) e.g. medial consonant in machine and dentals (7%), e.g. medial consonant in najan ‘another one’.

Note that fricatives in Gurindji Kriol are not always at the same place of articulation as in the English cognate (see Analysis 2).

Overall, stop-fricative variation within words occurred most frequently when the fricative in the cognate was labio-dental and least frequently when it was dental.

Cognate dentals, which were present due to their occurrence in words like that and this, were most likely to be pronounced invariably as a stop or affricate within Gurindji Kriol words, while alveolars were more likely to always be pronounced as a fricative within words and palato-alveolars were either always a fricative or varied with a stop. CHAPTER 4 99

100%

90%

80%

70%

60%

50% palato-alveolar 40% alveolar dental 30% labio-dental 20%

10% Percentage pronounced in each place of articulaton 0% all stops all fricatives stop-fric variation Pronunciation of 'potential fricative' within lexical forms

Figure 4.3. Proportions of places of articulation of fricatives in English cognates within Gurindji Kriol lexical form categories

Analysis 1 provided information about the lexical forms in which stop-fricative variation was occurring and the nature of these forms. The lexical forms analysed were selected based on the occurrence of fricatives in their English cognates. In the following analyses we use the term ‘variable words’ to refer to the Gurindji Kriol words that contained stop-fricative variation in the current dataset. ‘All stops’ refers to the words that have fricatives in their English cognates but were always pronounced with a stop of affricate in the data, and ‘all fricatives’ are the words that were always pronounced with a fricative. In the next analysis we examined all tokens of words containing stop- fricative variation, i.e. the ‘variable words’, and analysed the relationships between the

Gurindji Kriol pronunciation and fricatives in the English cognates. CHAPTER 4 100

Analysis 2: Relationships Between Fricatives in English Cognates and

Pronunciation in Gurindji Kriol Words that Contain Stop-Fricative Variation

The following analyses are of tokens the ‘variable words’, where the potential fricative was pronounced variably as a fricative or stop in the sample, and their relationships with the corresponding fricatives in the English cognate words. Cross- tabulations are presented of the fricatives in English cognates and the corresponding segments in Gurindji Kriol in word-initial, -medial and -final positions. To reiterate the terminology used for the purpose of this study: the Gurindji Kriol phones that are the focus of investigation are referred to as ‘potential fricatives’, and these are in Kriol- derived words whose English cognates contain fricatives. The corresponding segments of Gurindji Kriol potential fricatives in the English cognate words are referred to here as

‘cognate fricatives’, and are part of the analyses as a basis for calculating proportional variation.

Word-initial. Table 4.2 shows the frequency counts of fricatives in English cognates and corresponding segments in Gurindji Kriol variable words in word-initial position. Gurindji Kriol potential fricatives in word-initial position most frequently corresponded to labio-dental fricatives [f, v] in their English cognates (43%,

N=156/366), for example the initial consonant in bijin ‘fish,

Gurindji Kriol potential fricatives also frequently corresponded to word-initial alveolar fricative [s] in English cognate words (28%, N=101/366), for example ji ‘see’. These were pronounced as palatal stops [c, ɟ] (41%, N=41/101), affricates [ʧ, ʤ] (12%,

N=12/101) and were also often pronounced as the fricative [s] in Gurindji Kriol words

(36%, N=36). ‘Absent’ [s] is represented by the symbol [Ø] (N=9), and word-initially refers to segments that are the cluster [st] in English words pronounced as a stop CHAPTER 4 101

without the fricative in Gurindji Kriol words (e.g. Gurindji Kriol top 'to be', English cognate stop).

The frequency of cognate dental fricatives is due to their presence in English closed-class words, particularly demonstratives e.g. dat ‘the,

English cognates have a range of pronunciations in Gurindji Kriol – most frequently these words are pronounced with a palatal stop [c, ɟ], the voiced alveolar stop [d], or an affricate [ʧ, ʤ]. Word-initial dental fricatives in English cognates were rarely pronounced as fricatives in Gurindji Kriol variable words (N=3/366). The voiceless palato-alveolar fricative [ʃ] in cognate words is also pronounced with a range of phones in corresponding Gurindji Kriol words: as palatal stops and affricates (N=9/19) and as palatal fricatives [ʃ, ʒ] (N=10/19), for example juka (sugar).

Overall in Kriol-derived words that contained stop-fricative variation, 113 out of

366 (31%) word-initial potential fricatives were pronounced as fricatives in Gurindji

Kriol. The most frequent pronounced fricatives were [f] (N=44/113), [s] (N=36/113) and [v] (N=19/113). Palatal fricatives also occurred (N=11), and there were some rare cases of dental fricatives [θ, ð] (N=3).

Word-medial. Table 4.3 shows that most word-medial fricatives in English cognate words are pronounced with a voiced palatal stop [ɟ] in corresponding Gurindji

Kriol variable words (N=62/143). These included words where the cognate fricative was alveolar, e.g. dijei (this way), dental e.g. nojing (nothing) and palatal e.g. wajing

(washing). The most frequent corresponding phones in English cognates of medial

Gurindji Kriol potential fricatives were [s] (27%, N=39/143), [ð] (24%, N=34/143), and

[ʃ] (19%, N=27/143). The alveolar cognate fricative [s] was pronounced variably as a stop (N=22/39) and a fricative (N=14/39). Dental cognate fricatives [θ, ð] were mostly CHAPTER 4 102

pronounced as stops (N=36/49) in the corresponding Gurindji Kriol words, and when it was a fricative it was mostly voiced palatal [ʒ], e.g. nojing (nothing) [nɐʒŋ]. The voiceless palatal fricative [ʃ] in cognates was pronounced most frequently as the voiced palatal fricative [ʒ] (56%, N=15/27), for example majin (machine) [mɐʒn]. Labio- dental fricatives in cognate words, for example libim (leave) were more often pronounced medially as stops (N=18/28) than fricatives (N=8/28) in Gurindji Kriol variable words.

Overall, 48 out of the 143 (34%) potential word-medial fricatives in Gurindji

Kriol words containing stop-fricative variation were pronounced as fricatives, with the most frequent being palatal [ʒ] (69%, N=33/48).

Word-final. Frequencies of word-final fricatives in English cognates and pronunciation in corresponding Gurindji Kriol words with stop-fricative variation are shown in Table 4.4. The most frequent cognate fricative, word-final alveolar [s] e.g. pleis (place) was most frequently pronounced in Gurindji Kriol as a voiceless palatal stop [c] (48%, N=21/44), followed by other stops or affricates (N=18), and then alveolar fricatives [s, z] (N=5). A similar pattern was found for cognate palatal fricatives [ʃ], e.g. word-final consonant in binij ‘finish’, which was also most often pronounced with a palatal stop [c, ɟ] in Gurindji Kriol (75%, N=18/24), with a few cases of palatal fricatives (17%, N=4/24). Word-final phones that correspond to labio-dental fricatives [f, v] in English cognates were usually pronounced with bilabial stops in

Gurindji Kriol variable words (78%, N=18/23). There were only 4 cases of final [st] clusters, but these were pronounced variably as stops [c] in bej ‘first’ and bust-im-ap

‘bust’, a fricative [s] in bust, and the cluster [st] in bej. CHAPTER 4 103

Overall, in word-final position potential fricatives in Gurindji Kriol were rarer than in initial and medial positions, and they were also less frequently pronounced as fricatives (N=17/103, 17%).

Summary. Analysis 2 examined the relationships between potential fricatives in

Gurindji Kriol words that contain stop-fricative and the corresponding fricatives in their

English cognate words. Results gave further insight into the characteristics of segments that vary in the words described in Analysis 1. Findings of the first two analyses open up the question of what factors might influence whether a phone is pronounced as a fricative or a stop. To address this question, in the Analysis 3 we examined the relative contributions of phonological environment and child age to fricative variation in maternal Gurindji Kriol, and also described inter-speaker differences.

The next analyses were performed separately on tokens of words that contain stop-fricative variation and on all tokens in the dataset. In both cases the potential fricatives were coded according to whether they were pronounced as a fricative or other sound. To provide information on the characteristics of all potential fricatives (not just those in variable words as in Analysis 2), Figure 4.4 shows the proportions of potential fricatives pronounced as fricatives and other phones. The figure displays the differences in word positions, particularly that fricatives were more likely word-initially. It also shows that alveolar and palato-alveolar fricatives, and to a lesser extent voiceless labio- dentals [f], in English cognates were more likely to be pronounced as fricatives in

Gurindji Kriol than other places of articulation. Conversely, dental fricatives in English cognates were the least likely to be fricatives in Gurindji Kriol words, though as shown in Analysis 1 this was mostly lexically driven.

Table 4.2

Frequencies of word-initial fricatives in English cognates and corresponding phones in Gurindji Kriol variable words

Corresponding phone as pronounced in Gurindji Kriol words Phone in English b ɟ c f s p d v ʤ t Ø ʃ ʧ ʒ ɡ j w θ ð Total cognate f 59 - - 44 - 27 - 19 ------2 - - 151

s - 29 12 - 36 - - - 10 - 9 - 2 1 2 - - - - 101

ð - 22 31 - - - 5 - 2 1 - - - - - 2 - - 1 64

θ ------14 - 1 9 ------2 - 26

ʃ - 2 3 ------8 4 2 - - - - - 19

v 3 - - - - 2 ------5

Total 62 53 46 44 36 29 19 19 13 10 9 8 6 3 2 2 2 2 1 366

Table 4.3

Frequencies of word-medial fricatives in English cognates and corresponding phones in Gurindji Kriol variable words

Corresponding phone as pronounced in Gurindji Kriol words Phone in English ɟ ʒ b ʤ s f v Ø p j w d ð t Total cognate s 16 8 - 4 6 - - 3 - - - 1 - 1 39

ð 23 9 ------2 - - - - 34

ʃ 10 15 - 2 ------27

f - - 7 - - 4 1 - 3 - - - - - 15

θ 13 1 ------1 - 15

v - - 8 - - - 3 - - - 2 - - - 13

Total 62 33 15 6 6 4 4 3 3 2 2 1 1 1 143

Table 4.4

Frequencies of word-final fricatives in English cognates and corresponding phones in Gurindji Kriol variable words

Corresponding phone as pronounced in Gurindji Kriol words Phone in English c p t ɟ b s ʧ ʤ ʃ f z Ø st v ʒ Total cognate s 21 - 9 4 - 4 4 1 - - 1 - - - - 44

ʃ 14 - - 4 - - - 2 3 - - - - - 1 24

v - 3 - - 7 - - - - 1 - - - 1 - 12

f - 8 ------2 - 1 - - - 11

ðz 3 ------1 - - - - 4

st 2 - - - - 1 ------1 - - 4

ð 1 ------1 - - - - 2

Total 41 11 9 8 7 5 4 3 3 3 3 1 1 1 1 103

100%

90%

80%

70%

60%

50%

40% stop 30% fricative

20%

10%

0% Percentage pronounced as a fricative or stop in Gurindji Kriol Percentage pronounced as final final final final final final final initial initial initial initial initial initial initial medial medial medial medial medial medial medial

s z f v ð θ ʃ Phone in English cognates Figure 4.4. Relationships between fricatives in English cognates and corresponding pronunciation in Gurindji Kriol words.

CHAPTER 4 108

4.4.3 Analysis 3: Factors affecting fricative variation in maternal Gurindji Kriol

This analysis was performed to investigate factors influencing fricative variation in maternal Gurindji Kriol. The relative contributions of child age and phonological environments on fricative variation were analysed by performing binary logistic regression models for each word position. The outcome variable was pronunciation of

‘potential fricatives’ in Gurindji Kriol (fricative, stops) and the predictor variables were approximate child age (1;6, 2;0, 2;6) and phonological environment. For each word position analyses were first performed on all tokens in the dataset to analyse the global effect of child age on whether or not mothers pronounced ‘potential fricatives’ as fricatives or stops. The same model was then run just on the tokens in the words that were found to contain stop-fricative variation in the current dataset (see Analysis 1), to investigate whether within these words there was an effect of child age on the form used

(i.e. with the fricative or stop). Cell counts were too low to include speaker as a predictor, thus speaker was included as a random factor in all analyses and inter-speaker differences are described separately.

Word-initial regression. Word-initial phonological environment refers to the phone preceding word-initial potential fricatives. These were classified into fricatives, stops, nasals, liquids, vowels, and utterance boundary or pause. There were no preceding glides in this data. Two cases were excluded from the following analysis because the preceding phone was ambiguous.

All tokens. Table 4.5 shows the results of the logistic regression for word-initial position. Results indicate a significant effect of child age 2;6, Wald χ2 (1) = 6.897, p=0.009. The likelihood of a fricative occurring in mothers’ speech is significantly higher at child age 2;6 than 1;6 (odds ratio 0.630). There was also a significant effect of phonological environment when the preceding phone was a fricative, Wald χ2 (1) = CHAPTER 4 109

5.210, p=0.022, vowel, Wald χ2 (1) = 17.075, p<0.001, and pause or utterance boundary,

Wald χ2 (1) = 6.109, p=0.013. The likelihood of a potential fricative being pronounced

as a fricative rather than a stop increased when the preceding phone was a fricative

(odds ratio 0.185), vowel (odds ratio 0.385), and pause or utterance boundary (odds

ratio 0.551), compared to when the preceding phone was a stop.

Table 4.5 Logistic regression results for word-initial fricative variation in all words

95% C. I. for Exp (B) B s.e. Wald df p Exp(B) Lower Upper

Child Age

2;0 0.301 0.219 1.882 1 0.170 1.351 0.879 2.076

2;6 -0.462 0.176 6.897 1 0.009 0.630 0.447 0.889

Preceding Phone

Fricative -1.688 0.739 5.210 1 0.022 0.185 0.043 0.788

Nasal 0.125 0.275 0.208 1 0.648 1.134 0.661 1.943

Liquid -0.927 0.564 2.703 1 0.100 0.396 0.131 1.195

Vowel -0.954 0.231 17.075 1 0.000 0.385 0.245 0.606

Pause or -0.597 0.242 6.109 1 0.013 0.551 0.343 0.884 Boundary The reference group for Child Age is 1;6 and for Preceding Phone is Stop

Tokens in variable words. A logistic regression model with the same variables

as used in the above model was also performed on word-initial potential fricatives that

only occurred in variable words. Results are shown in Table 4.6. The were no CHAPTER 4 110

significant effects of child age, although the pattern was consistent with the analysis

above as the odds ratios were in the same direction (compared to age 1;6 odds ratio

1.585 for age 2;0, 0.713 for age 2;6). Results for phonological environment were also

consistent; however the only significant effect was when the preceding phone was a

vowel, Wald χ2 (1) = 6.185, p=0.013. Compared to when the preceding phone was a

stop, a fricative was more likely when the preceding phone was a vowel (odds ratio

0.363).

Table 4.6 Logistic regression results for word-initial fricative variation in variable words

95% C. I. for Exp (B) B s.e. Wald df p Exp(B) Lower Upper

Child Age

2;0 0.461 0.404 1.304 1 0.253 1.585 0.719 3.496

2;6 -0.338 0.269 1.576 1 0.209 0.713 0.421 1.209

Preceding Phone

Fricative -0.508 1.296 0.153 1 0.695 0.602 0.047 7.637

Nasal 0.219 0.452 0.234 1 0.629 1.245 0.513 3.020

Liquid -1.079 0.899 1.439 1 0.230 0.340 0.058 1.981

Vowel -1.014 0.408 6.185 1 0.013 0.363 0.163 0.807

Pause or -0.560 0.434 1.665 1 0.197 0.571 0.244 1.337 Boundary The reference group for Child Age is 1;6 and for Preceding Phone is Stop

CHAPTER 4 111

Word-medial regression. Word-medial phonological environments were

categorised into four groups: intervocalic (V_V) e.g. pijinglain 'fishingline', consonant

cluster preceded by a vowel (V_C) e.g. laswan 'last', consonant cluster followed by a

vowel (C_V) e.g. insaid 'inside', and triple consonant cluster (C_C) e.g. dragonflai

'dragonfly'. To test the hypothesis that word-medial fricatives would be more frequent

in intervocalic environments, V_C was used as the reference group, as fricatives and

other phones occurred almost equally in this environment (N= 26 fricatives, 27 other).

All tokens. The results for the word-medial logistic regression are shown in

Table 4.7. There were no significant effects of child age or phonological environment

on medial stop-fricative variation in all tokens.

Table 4.7 Logistic regression results for word-medial fricative variation in all words

95% C. I. for Exp (B) B s.e. Wald df p Exp(B) Lower Upper

Child Age

2;0 0.149 0.361 0.171 1 0.679 1.161 0.572 2.355

2;6 -0.408 0.292 1.951 1 0.163 0.665 0.375 1.179

Environment

V_V 0.535 0.329 2.647 1 0.104 1.707 0.896 3.250

C_V -0.776 0.574 1.826 1 0.177 0.460 0.149 1.418

C_C -0.589 0.770 0.585 1 0.444 0.555 0.123 2.509 The reference group for Child Age is 1;6 and for Environment is V_C

CHAPTER 4 112

Tokens in variable words. The results for the word-medial logistic regression of

tokens in variable words are shown in Table 4.8. Results indicate a significant effect of

child ages 2;0 and 2;6 compared to age 1;6, Wald χ2 (1) = 5.646, p=0.018 and Wald χ2

(1) = 4.530, p=0.033, respectively. The likelihood of a fricative occurring over a stop

medially in variable words was higher at child age 2;0 (odds ratio 0.250) and 2;6 (odds

ratio 0.368) than 1;6. There was no significant effect of phonological environment

medially in tokens of variable words (note there were not enough C_C cases in variable

words to include the C_C environment in this model).

Table 4.8 Logistic regression results for word-medial fricative variation in variable words

95% C. I. for Exp (B) B s.e. Wald df p Exp(B) Lower Upper

Child Age

2;0 -1.388 0.584 5.646 1 0.018 0.250 0.079 0.784

2;6 -0.999 0.469 4.530 1 0.033 0.368 0.147 0.924

Environment

V_V -0.740 0.676 1.198 1 0.274 0.477 0.127 1.795

C_V -1.419 1.369 1.073 1 0.300 0.242 0.017 3.544 The reference group for Child Age is 1;6 and for Environment is V_C

Word-final regression. Word-final phonological environment refers to the

phone following word-final potential fricatives. These were classified into fricatives,

stops, nasals, glides and liquids, vowels, and utterance boundary or pause. CHAPTER 4 113

All tokens. Table 4.9 shows the results of the logistic regression for word-final

position. There were no significant effects of child age or phonological environment on

word-final fricatives in all tokens. Phonological environment approached significance

for nasals, Wald χ2 (1) = 3.593, p=0.058, indicating fricatives decreased in likelihood

word-finally when the following phone was a nasal (odds ratio 3.094).

Table 4.9 Logistic regression results for word-final fricative variation in all words

95% C. I. for Exp (B) B s.e. Wald df p Exp(B) Lower Upper

Child Age

2;0 0.211 0.520 0.164 1 0.685 1.235 0.446 3.421

2;6 0.043 0.461 0.009 1 0.927 1.043 0.423 2.577

Following Phone

Fricative -0.291 0.812 0.128 1 0.721 0.748 0.152 3.675

Nasal 1.129 0.596 3.593 1 0.058 3.094 0.962 9.946

Liquid -1.635 0.968 2.855 1 0.091 0.195 0.029 1.299 or Glide

Vowel 0.337 0.753 0.200 1 0.654 1.401 0.320 6.130

Pause or 0.132 0.476 0.077 1 0.782 1.141 0.448 2.902 Boundary The reference group for Child Age is 1;6 and for Following Phone is Stop

Tokens in variable words. Table 4.10 shows the results of the logistic regression

for word-final position. There were no significant effects of child age or phonological

environment on word-final stop-fricative variation in variable words. CHAPTER 4 114

Table 4.10 Logistic regression results for word-final fricative variation in variable words

95% C. I. for Exp (B) B s.e. Wald df p Exp(B) Lower Upper

Child Age

2;0 -0.418 0.764 0.300 1 0.584 0.658 0.147 2.940

2;6 -0.310 0.692 0.201 1 0.654 0.733 0.189 2.848

Following Phone

Fricative -1.420 1.472 0.930 1 0.335 0.242 0.013 4.332

Nasal 0.100 0.828 0.015 1 0.904 1.105 0.218 5.602

Liquid -1.904 1.589 1.435 1 0.231 0.149 0.007 3.356 or Glide

Vowel -0.668 1.026 0.424 1 0.515 0.513 0.069 3.827

Pause or -0.529 0.757 0.488 1 0.485 0.589 0.134 2.599 Boundary The reference group for Child Age is 1;6 and for Following Phone is Stop

Inter-speaker differences. The proportions of potential fricatives pronounced

as fricatives for each speaker across child ages are displayed in Figure 4.5. This

included all tokens and indicates individual speaker variability in fricative production.

The same analysis performed just on tokens of words containing stop-fricative variation

showed the same pattern for each speaker, indicating the interspeaker differences in

Figure 4.7 are the same for both within-word and between-word stop-fricative variation.

The increase in fricatives when children are 2;6 appears to be driven by SS and CE, as

AR’s fricatives decrease from 2;0 to 2;6 (although note there were fewer tokens across

all ages for AR than the other two speakers). From child ages 1;6 to 2;0 CE’s use of CHAPTER 4 115

fricatives decreases, while SS’s remains about the same and AR’s increases. At age 2;6

SS and CE are using proportionately similar amounts of fricatives, while AR is using less.

0.40

0.35

0.30

0.25

0.20 SS 0.15 CE AR 0.10 Proportion of fricatives to stops 0.05

0.00 1;6 2;0 2;6 Average child age

Figure 4.5. Proportions of 'potential fricatives' pronounced as fricatives across child ages by each speaker in all tokens

The previous analyses revealed an effect of child age on fricative variation, accounting for phonological environment and word position. Results suggested that word-initially there is an effect of child age on all potential fricatives as well as just in those in words that contain stop-fricative variation in the current dataset. Word-medially the child age effect was only found in words containing variation. In the fourth and final analysis we further investigated the change in variation across child age by examining proportions of each word category as well as the proportions of pronounced fricatives in mothers’ speech across child ages. CHAPTER 4 116

4.4.4 Analysis 4: Effect of Child Age

First we examined the proportions of each word category described in Analysis

1 across child ages, displayed in Figure 4.6. Between child ages 1;6 and 2;0 the words with stop-fricative variation (‘variable words’) and the words invariably pronounced with fricatives (‘all fricatives’) proportionately decreased, while words where the potential fricative was always pronounced with a stop or affricate (‘all stops’) proportionately increased. Between child ages 2;0 and 2;6 the ‘all fricatives’ words remained proportionately low, while the proportions of the ‘all stops’ words decreased and the ‘variable words’ increased. Thus mothers were using relatively more words containing stop-fricative variation when children were 2;6 than when they were 2;0 or

1;6.

0.70

0.60

0.50

0.40 all stops 0.30 all fricatives variable words 0.20

Proportions of each word category 0.10

0.00 1;6 2;0 2;6 Average child age

Figure 4.6. Proportions of word categories at each child age

CHAPTER 4 117

Figure 4.7 further illuminates the change in variation as children aged, showing the proportions of potential fricatives pronounced as fricatives across child ages in tokens of words containing stop-fricative variation, tokens of all other words, and all tokens together (note the y-axis is of a different scale to that in Figure 4.6). Between child ages 1;6 and 2;0 the overall proportions of fricatives decreased, however the proportions of fricatives within words containing stop-fricative variation increased.

Taken with the proportional decrease in the frequency of variable words between 1;6 and 2;0 shown in Figure 4.5, this indicates that although mothers use relatively fewer variable words when children are 2;0 compared with 1;6, they are more likely to pronounce the variable words with fricatives at 2;0 than at 1;6.

0.40

0.35

0.30

0.25 tokens of variable 0.20 words tokens of other words 0.15 all tokens 0.10 Proportion of fricatives to stops

0.05

0.00 1;6 2;0 2;6 Average child age

Figure 4.7. Proportions of 'potential fricatives' pronounced as fricatives across child ages in variable words, other words, and the total sample

CHAPTER 4 118

Between child ages 2;0 and 2;6, Figure 4.7 shows a proportional increase in fricatives in both variable words and in other words. Thus when children are 2;6, mothers are both more likely to use words containing stop-fricative variation and are more likely to pronounce those words with fricatives compared to when children are 2;0 or 1;6. Fricatives in other words are also relatively more frequent at age 2;6 than 2;0 and

1;6. This is not due to an increase in words that are always pronounced with fricatives, as Figure 4.6 indicates these words remain proportionately low across all three time points. Rather, it is due to an increase in words pronounced with fricatives that were not coded as either variable words or ‘all fricatives’ words as there was only one token in the sample (and were also low frequency).

4.5 Discussion

Fricative production in Gurindji Kriol is highly variable. In addition to variation from the variable lexical forms of Kriol-based words, children are exposed to stop- fricative variability that changes differentially in mothers’ speech as children age from approximately 1;6 to 2;0 to 2;6, due to changes in mothers’ choice of both lexical forms and phonetic forms of variable words.

In the first analysis we investigated how much stop-fricative variation occurs within lexical forms and the nature of this variation. Words with multiple tokens in the dataset were grouped according to whether the potential fricative was always a fricative, always a stop or affricate, or pronounced variably as a stop or fricative. Findings showed that each group of words had different characteristics. Overall the ‘all stops’ words tended to be relatively high frequency with a near even split between open and closed class words. The ‘potential fricative’ (the corresponding segment in the English cognate word is a fricative) in these words was fairly evenly distributed across word CHAPTER 4 119

positions and the highest proportion of them were dentals, followed by labio-dentals and then palato-alveolars. The ‘all-stops’ group of words contained most of the Kriol- derived closed-class words that have fricatives in their English cognates, including the high frequency articles and demonstratives. It remains unclear, however, exactly which factors or interactions contribute to lexical items always being pronounced with a stop rather than a fricative, as it is not possible to isolate each factor. For example, dentals are rare in open-class words but have very high frequency due to the high frequency of closed-class words, so we cannot separate effects of place of articulation and word frequency. It is interesting to note that some of these words contained two potential fricatives where one was always pronounced as a stop and the other was variable, for example the initial and medial consonants in dijan 'this': the word-initial phone was always pronounced with a stop, while the word-medial phone was variable between a stop and a fricative. However this did not occur the other way round (i.e. the initial phone variable and the medial phone always a stop) so it could also be a word-position effect. Whether potential fricatives are always pronounced with stops or contain stop- fricative variation is also likely related to the length of time that they have been in the

Gurindji Kriol lexicon and the time it takes for conventional pronunciations of words to develop.

In contrast, words where fricatives were not variable with stops or affricates were nearly all open-class words. Fricatives occurred in all word positions, most frequently initial and least frequently final. They occurred in all places of articulation, although the majority were alveolar. Word-initial [s] thus occurred in a relatively large proportion of the ‘all fricatives’ words. Most of the words in this group (see appendix) were nouns and included some consonant clusters, e.g. flai ‘fly’, verbs e.g. stirimraun CHAPTER 4 120

‘stir’, recent borrowings e.g. fens ‘fence’, and some English code-switches e.g. swim

‘swim’.

Words containing stop-fricative variation tended to fall between the ‘all stops’ and ‘all fricatives’ words in most categories. Variable words were mostly open-class words, though there were more closed-class words in the variable category than the all fricatives category. The proportions of word-position of the potential fricative for the words containing variation were also in-between those for the ‘all stops’ and ‘all fricatives’ words, with the majority of stop-fricative variation occurring word-initially.

For place of articulation there were more labio-dentals and palato-alveolars in the words containing variation than in the non-variable (all stops or all fricatives) words. Words in the variable group contained relatively common nouns such as bijinlain ‘fishingline’ and jinek ‘snake’, verbs, e.g. libim ‘leave’, lijan ‘listen’, and closed-class words, e.g. dijei ‘this way’, abta ‘after’.

Impressionistically it is possible that these findings are associated with the time at which words were adopted into Gurindji Kriol, with target phone in early adopted words more likely to always pronounced with a stop, in the more recent words to always be a fricative, and in the variable words to be in-between, but the data does not exist to investigate this possibility systematically.

The second analysis addressed the relationship between fricatives in English cognates and pronunciation in Gurindji Kriol. Overall results demonstrated there is no one-to-one correspondence between Gurindji Kriol pronunciation and fricatives in cognates, accounting for word position. Analysis 2 findings suggested that while some words may be pronounced with the English manner of articulation it is not necessarily same phone in the cognate word that is produced. For example the medial sound in najan ‘another’ is a dental fricative in the English cognate. In Gurindji Kriol it is CHAPTER 4 121

pronounced variably with a palatal stop [nɐɟæn], but when it is a fricative it is palato- alveolar [nɐʒæn] rather than dental. These findings open up avenues of further research to analyse fricatives in Kriol and phonetic (allophonic) fricatives in Gurindji, to examine if and how traditional Gurindji phonology may be influencing the pronunciation of fricatives in Kriol-based words in Gurindji Kriol.

In analyses three and four we analysed the effect of child age on fricative variation in mothers’ speech. Accounting for phonological environment and interspeaker variation, the effect of child age was greatest in word-initial position, and the effect was also significant word-medially in tokens of words that contain stop- fricative variation. It is unknown whether the differences between word positions is a real effect, for example mothers modifying word-initial segments due to the greater perceptual saliency of segments in this position for acquisition (Bent, Bradlow & Smith,

2007; Vihman, Nakai, DePaolis & Hallé, 2004), or simply an analysis artefact, as there were many more tokens with potential fricatives in word-initial position than in medial or final positions. With a much larger sample it may be possible to randomly sample a comparable number of target segments from each word position, however, in the current corpus this was not possible and we would still not be able to make direct comparisons of word position because of the differing effects of phonological environment.

In these results we found fricative variation in mothers’ speech changed according to the children’s age14. This was partly due to lexical choice. Tokens of words in which the potential fricative was always pronounced as a fricative remained proportionately low across the three child ages, which was not surprising given these

14 The apparent effect of change over time may be within lexical items or due to the mothers' use of different items as the children get older. This is an analysis which we will include in the published version of the article. Interested readers are encouraged to look up and should cite the published version of this paper as a journal article, rather than this chapter. CHAPTER 4 122

words had a lower than average frequency. In contrast, tokens of words containing no fricatives and stop-fricative variation did not change much in proportion between child ages 1;6 and 2;0, with a greater proportion of ‘all stops’ words in mothers’ speech at both ages, while between ages 2;0 and 2;6 the proportion of ‘all stops’ words decreased as the words containing variation increased. Thus compared to age 1;6 and 2;0, when children were 2;6 mothers were using proportionately more words that could be pronounced variably with a stop or a fricative. The change in lexical choice may be driven by differences in discourse topic. A recent study of frequency variation in manner of articulation in English child-directed speech suggested segmental frequency variation is largely due to topic differences (Daland, 2013).

The total proportions of fricatives to stops in mothers’ speech also changed over time but followed a different pattern to the change in use of words containing stop- fricative variation, suggesting the effect of child age is only partly due to changes in the types of words mothers used. At 1;6 the proportions of fricatives to stops was about the same in tokens of variable words and other words. Between child ages 2;0 and 2;6 the proportion of fricatives to stops increased within 29 variable words, as well as in other words. In the other words the increase was likely due to the inclusion of words that only had one token in the dataset as many of these were low frequency words that were pronounced with a fricative. The increase of fricatives in the variable words indicates that not only were mothers using more words containing stop-fricative variation at 2;6 than at the earlier child ages, but also that they were more likely to use a fricative in these variable words at 2;6.

In terms of the fine-tuning concept, these results suggest mothers make differential segmental modifications according to child age and the change between approximate ages 1;6 to 2;6 may be in response to the child’s own language CHAPTER 4 123

development. Previous research on English varieties has found segmental phonetic variation in child-directed speech gradually changes over time to become like adult- directed speech, possibly to introduce children to the variation typical in adult speech in the community (Foulkes et al., 2005; Smith, Durham & Fortune, 2009). Recent studies have found variation in mothers’ speech may change differentially over time, with mothers possibly fine-tuning their speech to the child’s age and language abilities

(Buchan & Jones, in press; Ko, 2012). This is a plausible explanation for the child age effect in the current study, which found that word-initially and -medially the likelihood of mothers using fricatives over stops increased as children got older. These findings support previous studies on English and Korean child-directed speech (Lee & Davis,

2008; Lee et al., 2010), which found proportionately more stops and fewer fricatives in child-directed speech compared with adult-directed speech. Cross-linguistically fricatives tend to be acquired later by children than stops, as fricatives are relatively difficulty to articulate and require precise motor coordination (Dinnsen, 1992; Kent,

1992). Thus when children were younger mothers may have chosen (consciously or not) to increase their use of phonetic forms containing fricatives as children got older and their language abilities developed.

The interpretations here are tentative, as there are limitations to this study and further research is needed to find out what is causing these effects. The current findings are only based on data from three speakers and there was interspeaker variation, although this is consistent with the child-directed speech literature showing relatively large individual differences in language development (e.g. Ko, 2012). It would be useful for future studies to compare segmental distributions in adult-directed speech with those in child-directed speech at different child ages. This is difficult in naturalistic samples involving family groups; however, it would provide further insight into whether the CHAPTER 4 124

effect of change over time in mothers’ speech found here represents a change toward adult-directed speech, thus serving the function of exposing children to the type of fricative variation typical in the speech community.

CHAPTER FIVE

Fricative Reduction in Maternal Northern

Australian English15

15 This chapter has been published as a journal article. To cite, please look up and use the following citation: Buchan, H. & Jones, C. (2013). Phonological reduction in maternal speech in northern Australian English: Change over time. Journal of Child Language, doi: 10.1017/S0305000913000123. All material in this chapter (Chapter 5) is copyright of Cambridge University Press. CHAPTER 5 126

5.1 Abstract

Segmental variation in maternal speech to children changes over time. This study investigated variation in non-citation speech processes in a longitudinal, 26-hour corpus of maternal northern Australian English. Recordings were naturalistic parent- child interactions when children (N=4) were 1;6, 2;0 and 2;6. The mothers’ speech was phonetically transcribed and analysed. Based on previous sociophonetic research showing proportional changes in speech variants in maternal speech as children get older, it was predicted that deletion of word-initial /h/ and word-final /v/, processes common in non-citation speech, would increase over time. Instead results showed a non-linear change in deletion within a stable set of lexical items. Deletion proportionately increased between 1;6 and 2;0 and decreased between 2;0 and 2;6.

Further analysis indicated that increased deletion was not accounted for by changes in speech rate, which only marginally increased over time. Findings suggest mothers fine- tune differentially over time as children’s receptive and productive language knowledge develops.

CHAPTER 5 127

5.2 Background

Children’s phonological knowledge is shaped by their exposure to spoken language, and their phonological development includes learning the systematic phonetic variation that is a part of everyday speech. Variation in the speech input that children are often exposed to, known as infant- or child-directed speech, differs from that in interadult speech (adult speech directed to other adults) in lexical and syntactic structures as well as in phonetic features such as tempo, pitch and frequency (Fernald,

Taeschner, Dunn, Papousek, de Boysson-Bardies & Fukui,1989). Phonetic variation in infant-directed speech changes systematically over time, as caregivers’ speech to their children becomes more like interadult speech as children get older. For example it has been found cross-linguistically that mothers tend to use hyperarticulated vowels when talking to their infants aged 0;2-0;5 compared with their speech to other adults (Kuhl,

Andruski, Chistovich, Chistovich, Kozhevnikova, Ryskina, Stolyarova, Sundberg &

Lacerda, 1997). A longitudinal study on child-directed speech found that, compared with interadult speech, Mandarin-speaking mothers used hyperarticulated vowels when children were prelinguistic, aged 0;7-1;0, and at age 5;0-5;9 (Liu, Tsao & Kuhl, 2009).

Further, vowels were more hyperarticulated to children at 0;7-1;0 than at 5;0-5;9, in that they were longer, had a higher pitch, and the vowel space was larger so vowels were more distinct from one another, suggesting mothers modify their speech differently as their children get older (Liu et al., 2009).

Few studies have examined segmental variation and changes in maternal speech after infancy and even fewer have used longitudinal data. Much of the literature on speech input to children has focussed on the first year of life, as this is a crucial period for children’s language development (Jusczyk, Cutler & Redanz, 1993; Jusczyk,

Friederici, Wessels, Svenkerud & Jusczyk, 1993; Jusczyk, Luce & Charles-Luce, 1994;

CHAPTER 5 128

Kuhl, Williams, Lacerda, Stevens & Lindblom, 1992). Research has found differences in segmental distributions between mothers’ speech to children aged 1;0 and to adults

(Lee & Davis, 2010; Lee, Davis & MacNeilage, 2008), indicating segment patterns in maternal speech change sometime after the first year of life to eventually develop into adult-directed speech.

Further research is needed in order to understand how child-directed speech develops into adult-directed and at what ages the changes take place. Early research on segmental differences between child- and adult-directed speech suggests that the period between child ages 1;0 and 2;6 is a time of extensive variability in the input, and longitudinal analysis could be useful in elucidating how variability in mothers’ speech changes around these ages. Findings of these early studies were mixed in regards to the type of variation and the nature of the phonological differences in speech directed to children and to adults. However, the age of the children varied between studies and relatively large age ranges were used within studies. Bernstein-Ratner (1984a, 1984b) for example found more vowel lengthening, which is a feature of hyperarticulation (Liu et al., 2009), in child-directed speech to children aged 0;9-2;3 than in adult-directed speech, and more consonant reduction in adult-directed than child-directed speech.

Vowel hyperarticulation is a clarification process while consonant reduction refers to reduced or deleted consonants, so these studies suggested mothers modify their speech to pronounce segments more carefully and clearly when speaking to children.

Other early research, however, found that reduction was more frequent in speech to children than speech to adults (Shockey & Bond, 1980), and that isolated segments were less intelligible in child-directed than interadult speech (Bard & Anderson, 1983).

A discussion of these early studies by Cruttenden (1994) argued that children’s own language development may be a mediating factor, as the clarification processes were

CHAPTER 5 129

found in speech to children who were prelinguistic or starting to produce single words while more reduction and less intelligibility than interadult speech was found in speech to older children, generally over 2;0. A recent study on speaking rate in child-directed speech also found that age 2;0 is associated with a shift in mothers’ speech. Ko (2012) analysed mothers’ speech rate in longitudinal data from 25 speakers in the CHILDES database (MacWhinney, 2000). Ko (2012) performed a break-point analysis and found that in general a non-linear function fit the data significantly better than a linear model, and that the break-point tended to occur around age 2;0 to 2;5. Results showed large individual differences in the direction of the pattern; however the non-linear shift in mothers’ speech around child age 2;0 was consistently significant. It is possible that the mixed findings in the early child-directed speech studies are due to linear or non-linear changes in mothers’ speech over time, as mothers change their speech depending on their children’s age and/or communicative development.

The phonology of adult speech to children also varies as a function of speech style. In a corpus of caregiver-child speech recorded in Scotland, Smith, Durham and

Fortune (2007) found that caregivers’ use of variable phonetic forms changed depending on their speech style. The timing of children’s acquisition of stylistic variation in relation to their acquisition of linguistic (phonological and grammatical) constraints appears to vary depending on the type of variable being acquired (Labov,

1989; Roberts, 1997; Youssef, 1991). Kerswill and Williams (2000) suggest that children gradually acquire stylistic variation over time as they gain sociolinguistic competence, and Labov (2001) suggests that stylistic variation in caregiver speech plays a key role in children’s acquisition of variable forms.

Cruttenden (1994) suggests phonetic modifications made in speech to children may serve different functions at different child ages and stages of language

CHAPTER 5 130

development. For example exaggerated prosodic features in mothers’ speech peak in the first six months (Stern, Spieker & Barnett, 1983), and this may function to help infants attend to the speech signal and develop sensitivity to affect expressed in speech

(Cruttenden, 1994). Similarly the apparent peak in hyperarticulation when children start to produce intelligible words may function to clarify segments to ease perception

(Cruttenden, 1994). This explanation relates to the concept of fine-tuning: that mothers modify their speech according to their perceptions of their child’s receptive and productive language.

In addition to fine-tuning to children’s linguistic development, there is evidence that mothers may also fine-tune to children’s understanding of social categories indexed in speech, such as age, sex, region and socioeconomic status. Variation in phonetic segments due to socioindexical variables is termed sociophonetic variation (Foulkes &

Docherty, 2006), and in speech to children this type of variation introduces children to phonetic forms associated with speakers’ identities and social groups in the speech community. The few studies that have examined sociophonetic variation in caregiver speech suggest that it does change as children age. Smith, Durham and Fortune (2009) analysed mothers’ speech to children aged 2;11 to 3;11 in a Scottish dialect of English.

They examined t/d deletion in word-final consonant clusters, a reduction process common in non-citation inter-adult speech across English dialects that has been studied widely as a sociolinguistic marker (for a review see Tagliamonte & Temple, 2005).

Findings showed that rates of t/d deletion tended to be higher in speech to the older than the younger children, suggesting sociophonetic modifications in maternal speech may also vary with child age.

Sociophonetic variation associated with regional vernacular has also been shown to change over time in mothers’ speech to children after infancy. Foulkes, Docherty and

CHAPTER 5 131

Watt (2005) analysed a cross-sectional sample of mothers’ speech to children aged between 2;0 and 4;0 in a non-standard variety of English from Tyneside in northeastern

England. The phonetic segments analysed were variants of /t/ in word-medial intersonorant and word-final prevocalic positions, as the realisation of /t/ in these contexts is a sociolinguistic marker in Tyneside English (Foulkes et al., 2005).

Pronunciation of the segment as [t] was considered standard, while [
] and glottals in these contexts were considered non-standard variants that are characteristic of the region and may not be held in high regard outside the speech community. Foulkes et al.

(2005) found that overall, child-directed speech contained a higher proportion of the standard [t] variant compared with interadult speech. Standard [t] was relatively less frequent in speech to older than younger children while the non-standard variants were relatively more frequent in speech to the older children. It is suggested that phonetic variation in mothers’ speech changes over time to gradually familiarise children with the alternate phonetic forms common in interadult non-citation speech, and that socioindexical variation functions to facilitate children’s learning of the social meanings associated with phonetic variants as well as phonological category boundaries (Foulkes

& Docherty, 2006; Foulkes et al., 2005). These studies on sociophonetic variation in speech to children indicate that mothers may fine-tune their speech to their perceptions of children’s socioindexical as well as linguistic knowledge.

An explanation of how children may learn systematic phonetic variation from speech input can be found in exemplar-based models of phonological development, such as Pierrehumbert’s (2003) model. According to this approach all tokens contain both linguistic (e.g. acoustic) and socioindexical information, which is encoded concurrently for each exemplar and associated with frequency distributions (Foulkes &

Docherty, 2006; Pierrehumbert, 2003). Munson, Edwards and Beckman (in press)

CHAPTER 5 132

suggest that learning to map sociophonetic variation involves developing another level of representation for socioindexical information. This level may provide another source of top-down information for children to draw on when interpreting phonetic variability in the speech signal (Munson et al., in press). The strength of representations may depend on the frequency and recency of exemplars (Pierrehumbert, 2003), that is the stored memories of tokens, and so variable input over time during early childhood could shape representations at different levels of mapping. Usage-based models of phonological acquisition are supported by studies on children’s acquisition of liaison consonants in French. Liaison is a phonological process that occurs variably in French connected speech, depending on linguistic factors and social economic status (SES)

(Chevrot, Dugua & Fayol, 2009; Chevrot, Nardy & Barbu, 2011). SES differences in children’s productions of liaison consonants reflect those in adult speech and increase as children age from 2;0 to 6;0 (Chevrot et al., 2011). Chevrot et al. (2009) suggest that children acquire variation by first encoding multiple exemplars of the variable lexical forms in the input and then abstracting information from the exemplar store to form schemas.

In addition to social meanings of segmental variation, children are exposed to variation associated with speech acts and social interactions. Phonetic variation can be associated with style shifting, which is learned throughout childhood and related to caregiver usage (Kerswill & Williams, 2000). Mothers’ use of either standard or local vernacular variants has been found to be constrained by situation in a Scottish dialect of

English, with more vernacular variants used in the informal contexts of play and routine speech than the formal contexts of discipline and teaching (Smith et al., 2007).

Introducing children to situational variation may be another function of modifications made in speech to children.

CHAPTER 5 133

In summary, children are exposed to systematic phonetic variation from early on in life. Phonetic modifications made in maternal speech change over time as child- directed speech develops into interadult speech, but there is relatively little literature on how and when this happens as children get older. There is some evidence that infant- and child-directed speech has different functions depending on the age and linguistic development of the child, and this may drive changes in variation in the input over time as mothers fine-tune their speech to their perceptions of children’s receptive and productive language. The few studies that have examined variation in maternal speech to children after infancy suggest mothers may also shape modifications in their speech to their perceptions of children’s knowledge of socioindexical factors, and there are also likely to be situational constraints on mothers’ use of non-standard phonetic forms. In the current study we investigated how variation in non-citation speech changes in mothers’ speech to children from ages 1;6 to 2;6 in a regional variety of Australian

English. To get a clearer understanding of when mothers might start using more processes in non-citation speech, this study started with children of a younger age than those in the Foulkes et al. (2005) study, in which children were aged 2;0 to 4;0, and the

Smith et al. (2009) study, in which children were aged between 2;11 and 3;11. In the current study we started with children aged 1;6.

5.2.1 Phonetic Variation in Australian English

The phonological system of standard Australian English is well established, although regional dialects have generally not been studied in detail. Standard Australian

English differs from other standard English varieties in realisation of many of the vowels, prosodic patterns, lexical items and paralinguistic characteristics (Harrington,

Cox & Evans, 1997). In a recent description of the main phonological features of

Australian English, Standard Australian English is described as, ‘the dominant dialect

CHAPTER 5 134

used by the vast majority of speakers’ (Cox & Palethorpe, 2007, p. 341) in Australia today. Being a multicultural country there are many other dialects spoken, which have previously been broadly grouped into ‘Aboriginal English’ and ‘Ethnocultural

Australian English’ categories (Cox & Palethorpe, 2007). There may also be varieties differing on other dimensions such as geographic region, but this has not been studied in depth owing to a historical perception that there is relatively low regional variation in

Australian English.

However, recent research has found some regionally distributed vowel characteristics (Cox & Palethorpe, 2004) as well as variation in vowel ‘broadness’ by region and other socioeconomic factors (Cox & Palethorpe, 2010). Relatively less research into Australian English has focused on consonants (Ingram, 1989; Tollfree,

2001). Sociolinguistic studies of consonant variation include Horvath (1985) and

Ingram (1989); both found significant effects of sociolinguistic variables on consonant processes. Horvath’s (1985) study included ‘h dropping’ (/h/ deletion) and Ingram’s

(1989) study investigated connected speech processes. In Australian English, /h/ deletion is considered a sociolinguistic feature which correlates with socioeconomic status and gender; connected speech processes (assimilation, elision) are also socially correlated and were one of five variables differentiating the ‘classical’ accent classes in

Australian English: broad, general, and cultivated (Mitchell & Delbridge, 1965).

Horvath (1985) analysed consonant variants in Australian English speech from

117 speakers in Sydney, New South Wales. Data were from the Sydney Urban Dialect

Survey, a sociolinguistic survey which involved interviews conducted with a stratified sample of Australian English speakers from Sydney, including speakers of English as a second language. Horvath (1985) conducted principal components analyses to identify four main ‘sociolects’ based on sex, age, socioeconomic status, and ethnicity, and

CHAPTER 5 135

analysed the frequency of six consonant processes to compare the incidence of these processes in the ‘sociolect’ subgroups. One consonant process she examined was word- initial /h/ deletion (1).

(1) go get him → go get ‘im

Along with other consonant processes considered typical of ‘broad’ Australian English speakers, word-initial /h/ deletion was generally more frequent in males, teenagers, and speakers from ‘lower working class’ backgrounds (Horvath, 1985). In dialects of

English in England, ‘h dropping’ is also a known sociolinguistic marker (Tollfree, 1999;

Wells, 1982; Williams & Kerswill, 1999).

Ingram (1989) examined connected speech processes according to socioeconomic status in adolescent speakers from Brisbane, Queensland. One type of connected speech process is the articulatory simplification that often occurs in fast non- citation speech, leading to phonetic assimilation and reduction (2). The frequency of connected speech processes account for much of the phonological differences between citation and non-citation speech, and vary with speech rate, register, and ‘broadness’

(Ingram, 1989).

(2) I don’t know about him → I dunno ‘bout ‘im

One consonant process that Ingram (1989) analysed was cluster reduction, and it was noted that cluster reduction largely affected word-final /nt/, /nd/ and /st/ clusters.

Ingram (1989) found that the frequency of consonant cluster reduction was relatively high in both working class and middle class adolescents, and there was no significant difference between the two groups. There was however a significant difference between working class and middle class adolescents in the frequency of other speech processes.

Deletion of /h/ occurred significantly more often in speech of the working class than the middle class participants, and was most common in third person pronoun forms. The

CHAPTER 5 136

working class group were also significantly more likely to delete schwas (e.g. ‘bout for about) and word-initial interdental fricatives (e.g. ‘em for them). These findings show sociolinguistic factors affect connected speech processes among Australian English speakers. In the present study we consider word-initial /h/ deletion and the speech process word-final /v/ deletion in a little described variety of northern Australian

English. In the absence of sociolinguistic evidence for this dialect, we refer to these as non-citation speech processes henceforth.

5.2.2 Katherine English

Katherine is a small regional town in the Northern Territory, Australia, located about 317 kilometres south of Darwin, the Northern Territory capital. The English spoken in Katherine, which we term Katherine English, was sampled for the current study as a variety of regional Australian English. This language variety has been the subject of just one previous study (Jones, Meakins, & Buchan, 2011), and no sociolinguistic research, so we offer comments from speakers to characterise the sociolinguistic setting. Terms that speakers used to describe Katherine English included

‘very casual’, ‘relaxed’, ‘slang’, ‘country-sounding’, ‘not as refined’, ‘rough’ and

‘ocker’ (slang for a stereotypically uncultivated Australian), Speakers also say – we lack objective data – that their words tend to be simpler and not as long, and that the speech of Southerners (people from outside the Northern Territory or city people) is ‘fuller’ in that they use longer words, more words, and fewer shortened versions. Overall speakers suggest that Katherine English may be considered a casual or relaxed variety of

Australian English with a relatively high use of vernacular in everyday non-citation speech.

CHAPTER 5 137

5.2.3 The Current Study

The aim of this study was to examine how phonetic variation related to non- citation speech processes in regional Australian English maternal speech changes as children age. We collected and analysed a corpus of naturalistic speech from mothers when their children were aged 1;6, 2;0 and 2;6. This study is part of a larger project on fricative variation in maternal speech in northern Australia (Buchan, in prep.), which involves a comparison of the similarities and differences in fricative variation in

Katherine English and an Aboriginal contact language variety in the region, Gurindji

Kriol. Gurindji Kriol contains extensive fricative variation in Kriol-derived words, i.e. of English historical origin (Meakins, 2011). Because fricative variation is the focus of the larger project, in this study we examined variants in fricatives that are related to processes in non-citation speech. Word position effects are also being investigated in the Aboriginal language variety, and so in this study we analysed a word-initial and a word-final fricative.

The phonetic variants we chose to measure were word-initial /h/ deletion (3a, b) based on the previous findings from Horvath (1985) and Ingram (1989), and word-final

/v/ deletion (4a, b), which are regarded as processes that occur in non-citation speech

(or lexical alternatives, depending on the analysis adopted): standard form

(3a) Go get her!

(4a) Get out of there. containing deletion

(3b) Go get ‘er!

(4b) Get outta there.

CHAPTER 5 138

We analysed deletion across function and content words. Previous research has found that phonological information is important for classifying word class, and that phonological reduction is more likely to occur in function than content words (Cutler,

1993; Monaghan, Chater & Christiansen, 2005). It was therefore expected that overall deletion would be more frequent in function words.

In this study maternal speech is the main source of speech input for the target children and is defined as all speech spoken by the mother around the target child. This is based on Soderstrom’s (2007) definition of speech input to children and studies that have found children learn from overhearing interadult speech as well as speech directed specifically to them (Akhtar, 2005; Au, Knightly, Jun & Oh, 2002; Martinez-Sussman,

Akhtar, Diesendruck & Markson, 2011).

Based on recent studies showing child-directed speech to older children gradually changes to be more like interadult speech over time (Foulkes et al., 2005;

Smith et al., 2009), it was predicted that mothers would be more likely to use the reduced variants as children got older.

5.3 Method

5.3.1 Participants

Four mothers and their children participated in the recording sessions, which were conducted at three stages six months apart when the children were aged approximately 1;6, 2;0 and 2;6. Exact ages are shown in Table 5.1. A fifth mother participated but her child was six months older than the others so her data are not included in this study, as she was only recorded at two of the timepoints under investigation here and therefore cannot be analysed longitudinally.

Mothers were monolingual Australian English speakers living in Katherine,

Northern Territory, Australia. Three mothers had grown up in Katherine, while the

CHAPTER 5 139

fourth had grown up in a small regional town in northwestern Queensland, Australia.

The highest educational attainment was a Bachelor degree (two mothers) and secondary school (two mothers). Mothers were recruited through local childcare centres and other local businesses, and were paid for their time.

Two of the target children had siblings, and there was also some participation in the sessions from other family members and visitors. Mothers were instructed to focus on the target child, so mothers’ interactions with other people were generally minimal.

Table 5.1 shows the target children’s sex, their ages at each stage and the mothers’ education levels, and is followed by a brief description of the dyads and other people present in recording sessions. Only the mothers’ speech was analysed for this study.

Informed consent was obtained from or on the behalf of all participants.

Cathy and Lucy. Lucy was Cathy’s only child. Lucy’s father occasionally participated in recordings, but in the majority of the sessions Cathy and Lucy were the only participants.

Kim and George. George had two older siblings, a boy and girl aged 6;3 and

3;0 respectively at Stage 1. One or both of the siblings participated in most of the sessions. Kim was asked to focus on George in the sessions, so generally the siblings played together while Kim played with George. The children’s father participated occasionally.

Kellie and Tyson. Tyson was an only child. Tyson’s grandmother was present in most of the sessions. Other adult family members were sometimes in the vicinity but did not participate in sessions.

Erin and Grace. Grace had two younger brothers: one aged 0;5 at Stage 1, and another born between Stages 2 and 3. Sessions were held when the younger children were mostly sleeping, so the majority of interactions were between Erin and Grace. Two

CHAPTER 5 140

adult males who work for Erin and her husband were occasionally present in some of the sessions.

Table 5.1

Participant Characteristics

Target Target Mother’s Family Stage 1 Stage 2 Stage 3 Mother child child’s sex education position child age child age child age

Bachelor only Cathy* Lucy* female 1;5 1;11 2;5 degree child

Secondary youngest Kim George male 1;5 1;11 2;5 school of three

Secondary only Kellie Tyson male 1;7 2;1 2;7 school child

Bachelor oldest of Erin* Grace* female 1;6 2;0 2;8† degree three * pseudonyms used at mother’s discretion

† the age difference between Stage 2 and 3 is greater for Grace than the other target children because these sessions were recorded later due to participant circumstances

5.3.2 Procedure

Speech samples were recorded to make a corpus of naturalistic maternal speech for the analysis of sound patterns that the children are typically exposed to. Audiovisual recordings were made in a natural setting, with each mother and child recorded at their home or at a local park, with care taken to choose quiet environments with minimal background noise. All the recordings were made by the first author, who spent time before and after each recording session talking with the participants about both the research project and unrelated subjects to build rapport. Mothers were aware that the project was about their speech, and were debriefed on the details after the final recording session.

CHAPTER 5 141

Before each recording mothers were asked to do free play with their children and reminded that they could stop the session at any time. Participants played with materials brought by the researcher and their own toys and games. The most common activities across children and stages were drawing and painting, and playing with coloured blocks, toy cars, dolls, and playdough. Other activities included swingsets and other playground equipment, and general outside activities such as feeding pets.

The mother and target child each wore a wireless, cardioid, condenser lapel microphone (Sennheiser ME104) that recorded high quality audio (at 48kHz, 24 bit) via bodypack transmitter to a receiver with XLR connection to a hard disk recorder (Edirol

R-44), in separate audio tracks for mother and child. The mother’s audio track was used for transcription. The acoustic quality of the recordings was high, and with the cardioid lapel microphone positioned upwards at the top centre of the mother’s shirt her speech was recorded clearly on her audio track. As recordings were made in busy naturalistic environments, occasionally a background noise was too loud to clearly hear the speech.

In these cases speech was marked as unclear in the transcripts and was not analysed. A spectrogram of an utterance from a typical recording is shown in Figure 5.1. This example was recorded outdoors in a park by a river and has a dynamic range of 52 dB.

The researcher also filmed sessions with a video camera (Sony AVCHD HDR XR-

500V) that had a camera-mounted microphone. Several recording sessions were conducted with each mother at each stage over one to three weeks, to produce in total between 8 and 11 hours of recording time at each stage as detailed in Table 5.2. The amount of recording depended on mothers’ availability and the children’s compliance.

Most sessions went for half an hour to an hour in length, and usually ended when the child or children got tired.

CHAPTER 5 142

Figure 5.1. Spectrogram of utterance "There we go, we'll just leave him there, shall

we?"

Table 5.2

Total and Mean Lengths of Recording Sessions, hh:mm:ss

Mother child age 1;6 child age 2;0 child age 2;6 Total

Cathy 02:05:31 02:00:18 02:44:58 06:50:47

Kim 02:11:37 01:38:00 02:28:12 06:17:49

Kellie 02:41:54 03:07:20 02:24:54 08:14:08

Erin 01:39:48 00:47:53 02:06:13 04:33:54

Total 08:38:50 07:33:31 09:44:17 25:56:38

Mean 02:09:43 1:53:23 02:26:05

CHAPTER 5 143

Mothers’ speech was transcribed from the audio using the phonological analysis and transcription software Phon (Rose, Hedlund, Byrne, Wareham & MacWhinney,

2007). Phonetic transcriptions were made of the mothers’ actual pronunciation as well as the standard Australian English form. Transcription was done by the first author, a native Australian English speaker, using the transcription system recommended by

Harrington et al. (1997) for phonetic transcription of Australian English. To check transcription reliability, a second transcriber separately transcribed one hour of recording at Stage 1 and one hour at Stage 2, made up of a random selection from each session. The second transcriber was a native Australian English speaker who has a

Masters degree in speech science and is trained in phonetic transcription of Australian

English. Reliability analyses showed 95.63% agreement between transcribers at Stage 1 and 96.88% at Stage 2. The majority of disagreements were about vowel quality, in particular between the schwa and full vowels (making up 45.93% of total disagreements at Stage 1 and 49.24% at Stage 2). The two transcribers held consensus discussions to make decisions about how to transcribe disagreements. Where consensus could not be reached, a third transcriber (the second author) was consulted to make a decision.

5.3.3 Analysis

The search tool in Phon was used to perform frequency counts of fricatives by word position. Consonant deletion was quantified by searching for differences in the standard and actual IPA transcriptions. The transcripts were then searched manually to determine whether the differences between fricatives in the IPA Actual and IPA Target tiers were due to speech errors or the mother using a different lexical form (e.g. the child-directed speech form horsie for horse would show up as a difference in the Target and Actual word-medial and word-final [s]), or whether it was an apparent systematic process in the sample. The exclamation hey was excluded from the /h/ deletion analyses.

CHAPTER 5 144

To test whether the frequency of specific speech processes increased over time, the processes word-initial /h/ deletion and word-final /v/ deletion were analysed further in

SPSS using binary logistic regression. To explore local effects on the processes we analysed the number and types of lexical items in which deletion occurred, in both function and content words, using frequency counts at each child age.

Speech rate was measured to investigate whether changes in rates of deletion over time were related to changes in speech rate. Speech rate in words per second was analysed for the entire sample at each timepoint, following the procedure outlined and validated by Ko (2012). A more detailed analysis of speech rate was also performed on a subset of the data by calculating syllables per second in a random selection of sixty utterances from each mother at each child age: thirty in which deletion occurred and thirty in which it did not.

5.4 Results

5.4.1 Analysis 1: Incidence of Word-Initial /h/ and Word-Final /v/ Deletion in

Maternal Speech Over Time

Word-initial /h/ deletion. It was hypothesised that the proportion of word- initial /h/ deletion in maternal speech would increase at each time point. Table 5.3 shows the frequencies of word-initial /h/ deletion out of possible contexts at each average child age, and percentages of deletion for each speaker are in Figure 5.2.

Descriptively there was a proportional increase in deletion between child ages 1;6

(24.65%) and 2;0 (35.11%), and a proportional decrease between 2;0 and 2;6 (24.62%).

CHAPTER 5 145

Table 5.3

Raw Frequencies of Word-Initial /h/ Deletion Out of Possible Contexts by Each Speaker at Each Average Child Age

Proportion of deletion (N deleted/N contexts)

1;6 2;0 2;6 Total % deletion

Erin 18/169 24/112 20/162 14.00

Cathy 56/306 83/327 42/340 18.60

Kim 86/211 87/252 91/210 39.23

Kellie 102/377 201/434 105/336 35.57

Total N 262/1063 395/1125 258/1048 915/3236

Total % deletion 24.65 35.11 24.62 28.28

CHAPTER 5 146

Figure 5.2. Percentage word-initial /h/ deletion out of word-initial /h/ contexts by child age and speaker

To test the hypothesis that word-initial /h/ deletion is influenced by child age we performed a binary logistic regression. The outcome variable was word-initial /h/ deletion and the predictor variables were child age and speaker. Child age (1;6, 2;0 and

2;6) was entered as a repeated contrast making age 2;0 the reference group. Table 5.4 shows the results of logistic regression analysis. The model has good fit to the data χ2

(11)=295.36, p<.001, and accounts for some of the variance (Nagelkerke R2 = 0.10).

Results indicate a significant effect of child age, Wald χ2 (2) = 16.08, p <.001. The odds of /h/ deletion are lower at child age 1;6 than 2;0 (odds ratio 0.67), and higher at 2;0 than 2;6 (odds ratio 1.42). The hypothesis that deletion would increase as child got older

CHAPTER 5 147

was supported between child ages 1;6 and 2;0, while the decrease in deletion between

ages 2;0 and 2;6 was unexpected.

Table 5.4

Logistic regression for word-initial /h/ deletion: Child age and speaker

B s.e. Wald df p Exp(B) 95% C. I. for Exp (B)

Lower Upper

Child Age

1;6 -0.40 0.11 13.15 1 <0.001 0.67 0.54 0.83

2;6 0.35 0.11 9.91 1 0.002 1.42 1.14 1.76

Speaker

Erin -1.34 0.15 81.36 1 <0.001 0.26 0.20 0.35

Cathy -1.07 0.10 111.45 1 <0.001 0.34 0.28 0.42

Kim 0.02 0.09 0.05 1 0.82 1.02 0.85 1.23

Child Age x Speaker

1.6 x Erin -0.02 0.36 0.00 1 0.96 0.98 0.48 1.99

2.6 x Erin 0.15 0.35 0.17 1 0.68 1.16 0.58 2.32

1.6 x Cathy 0.63 0.23 7.30 1 0.01 1.88 1.19 2.97

2.6 x Cathy -0.03 0.25 0.01 1 0.91 0.97 0.60 1.58

1.6 x Kim 1.22 0.23 29.08 1 <0.001 3.39 2.18 5.29

2.6 x Kim -1.15 0.23 25.59 1 <0.001 0.32 0.20 0.50

The reference group for Child Age is 2;0 (repeated contrast) and for Speaker is Kellie.

There is also a significant effect of speaker on /h/ deletion, Wald χ2 (3) = 191.21,

p <.001. The speaker Kellie was the reference group (total deletion 35.57%), and results

CHAPTER 5 148

show that the odds of deletion were lower for Erin (odds ratio 0.26) and Cathy (odds ratio 0.34). There was no significant difference in the odds of deletion for Kim (odds ratio 1.02) compared with Kellie.

Interactions between child age and speaker were tested in the logistic regression model to test whether the main effect of age was found across all speakers. Results show a significant interaction effect, Wald χ2 (6) = 46.80, p <.001. As shown in Table

5.4 there were three significant interactions: Cathy by age 1;6, Kim by age 1;6, and Kim by age 2;6. Figure 5.2 shows that Kim displays a v-pattern of deletion over time, a decrease in deletion from 1;6 to 2;0 followed by an increase from 2;0 to 2;6, while

Kellie, Cathy and Erin display an inverted v-pattern, that is an increase from 1;6 to 2;0 followed by a decrease from 2;0 to 2;6. Between child ages 1;6 and 2;0 /h/ deletion was significantly more likely to increase for the reference group Kellie than for Kim (odds ratio 3.39). Between child ages 2;0 and 2;6 the odds of /h/ deletion were significantly more likely to decrease for Kellie than Kim (odds ratio 0.32). The interaction of Cathy by child age 1;6 shows that the increase in the odds of /h/ deletion between ages 1;6 and

2;0 was significantly higher for Kellie than Cathy (odds ratio 1.88).

Word-final /v/ deletion. It was hypothesised that the speech process word-final

/v/ deletion would proportionately increase in mothers’ speech at each time point. Table

5.5 shows the frequencies of word-final /v/ deletion out of possible contexts at each average child age. The percentage of word-final /v/ deletion for each speaker is shown in Figure 5.3. This process was not as frequent as word-initial /h/ deletion, so there were fewer possible contexts and fewer tokens in which the process occurred16. Overall there was a slight increase in word-final /v/ deletion between child ages 1;6 and 2;0, from

16 Searches were performed in Phon for word-final [v] contexts that precede a word-initial labio-dental fricative [v, f] in the following word. This environment is extremely rare in the corpus, with one case at child age 1;6 (Cathy), none at 2;0, and four at 2;6 (1 Cathy, 3 Kim).

CHAPTER 5 149

16.15% to 21.95%, and a decrease between ages 2;0 and 2;6 to 16.74%. This inverted v- pattern resembles that for word-initial /h/ deletion and was partially consistent with the hypothesis: word-final /v/ deletion changed in mothers’ speech over time but not in the expected direction as children aged from 2;0 to 2;6.

Table 5.5

Raw Frequencies of Word-Final /v/ Deletion Out of Possible Contexts by Each Speaker at Each Average Child Age

Proportion of deletion (N deleted/N contexts)

1;6 2;0 2;6 Total % deletion

Erin 5/70 7/50 8/74 10.31

Cathy 22/96 32/132 33/178 21.43

Kim 13/49 9/42 19/82 23.70

Kellie 7/76 24/104 15/114 15.65

Total N 47/291 72/328 75/448 194/1067

Total % deletion 16.15 21.95 16.74 18.18

CHAPTER 5 150

Figure 5.3. Percentage word-final /v/ deletion out of word-final /v/ contexts by child

age and speaker

To test the hypothesis that word-final /v/ deletion is influenced by child age we performed another binary logistic regression. Table 5.6 shows the results of the analysis using child age and speaker as predictors of word-final /v/ deletion. The model fit to the data approached significance, χ2 (11) = 18.66, p = 0.067 (Nagelkerke R2 = 0.02).

Results indicate no significant effect of child age, Wald χ2 (2) = 2.18, p = .336. The hypothesis that word-final /v/ deletion would increase as children got older was not supported. In this model the effect of speaker approached significance, Wald χ2 (3) =

7.75, p = .052. The speaker with the highest total /v/ deletion was the reference group

(Kim, total /v/ deletion 23.70%) and results show that the odds of deletion were marginally lower for Erin (odds ratio 0.64).

CHAPTER 5 151

Table 5.6

Logistic Regression for Word-Final /v/ Deletion: Child Age and Speaker

B s.e. Wald df p Exp(B) 95% C. I. for Exp (B)

Lower Upper

Child Age

1;6 -0.34 0.23 2.08 1 0.15 0.71 0.45 1.13

2;6 0.09 0.21 0.17 1 0.68 1.09 0.73 1.63

Speaker

Erin -0.44 0.30 2.24 1 0.14 0.64 0.36 1.15

Kellie 0.27 0.25 1.18 1 0.28 1.31 0.81 2.14

Cathy 0.23 0.21 1.22 1 0.27 1.26 0.83 1.91

Child Age x Speaker

1.6 x Erin 0.40 0.76 0.28 1 0.60 1.50 0.34 6.58

2.6 x Erin -0.06 0.65 0.01 1 0.93 0.94 0.26 3.36

1.6 x Kellie 1.57 0.65 5.90 1 0.02 4.81 1.36 17.14

2.6 x Kellie -0.71 0.56 1.59 1 0.21 0.49 0.17 1.48

1.6 x Cathy 1.04 0.54 3.68 1 0.06 2.81 0.98 8.12

2.6 x Cathy -0.09 0.44 0.04 1 0.84 0.91 0.38 2.18

The reference group for Child Age is 2;0 (repeated contrast) and for Speaker is Kim.

There was no significant interaction effect, Wald χ2 (6) = 7.48, p = .279. Figure

5.3 suggests, however, that the pattern of /v/ deletion over time depends on speaker.

Kellie, Cathy and Erin display an inverted v-pattern as they are less likely to delete at

child age 2;0 than at 1;6 or 2;6, while Kim displays the opposite pattern as she is more

likely to delete at 2;0 than 1;6 or 2;6. The interspeaker patterns for word-final /v/

CHAPTER 5 152

deletion over time resemble those found for word-initial /h/ deletion and show that within speakers there was consistency across both processes.

Analysis 1 discussion. It was predicted that the likelihood of deletion in mothers’ speech would increase over time to reflect a gradual change from child- directed speech to interadult speech. The hypothesis was partially supported: word- initial /h/ and word-final /v/ deletion were more likely to occur in mothers’ speech when children were 2;0 than 1;6, but were unexpectedly less likely at 2;6 than 2;0. The change over time was significant for /h/ deletion but not for /v/ deletion, likely because word-final /v/ contexts were rarer in the corpus. Results also show interspeaker differences, which is consistent with Ko’s (2012) finding of individual variation in developmental paths of speech rate in child-directed speech. In this study the presence of siblings may have contributed to interspeaker variation, as the mother who did not have the inverted-v pattern of deletion also had two older children who were present in many of the sessions, while with the other three mothers the target child was generally the only child in the recordings. The inverted-v effect was also stronger in one of the speakers, Kellie, than in the other two speakers who displayed the pattern. Future research on more speakers would be beneficial to examine possible causes of interspeaker variation: a possible contributing factor is the gender of the child, as Erin and Cathy, who were mothers of girls, displayed similar patterns of a relatively shallow inverted-v while Kim and Kellie were mothers of boys and displayed greater variation.

Possible effects of the children’s gender is discussed further in Analysis 2.

It was found that the two mothers with less formal education, Kim and Kellie, displayed the highest rates of /h/ deletion at all timepoints. We cannot draw any conclusions here due to the small number of speakers, but it would be interesting to further explore this effect in future research.

CHAPTER 5 153

The finding of an inverted-v pattern in three of the four speakers is surprising in light of previous studies that have found consonant deletion in clusters is more likely in mothers’ speech to children aged 3;11 than 2;11 (Smith et al., 2009), and that non- standard variants become more frequent as children age from 2;0 to 4;0 (Foulkes et al.,

2005). This could be due to the different indexical natures of the speech variables investigated in these studies, i.e. cluster reduction in the Smith et al. (2009) study, [t] variants in Foulkes et al. (2005), and /h/ and /v/ deletion in the current study. A possible interpretation of our findings in Analysis 1 is that /h/ and /v/ deletion are non-citation speech processes in Katherine English, and mothers increase their use of these processes between child ages 1;6 and 2;0 and decrease them as children age from 2;0 to 2;6, perhaps in relation to the children’s speech production. To explore this interpretation further analyses are needed to arrive at a more detailed, lexically-specific understanding of the deletion processes.

Another possible explanation for the findings in Analysis 1 is that there is a change over time in the distribution of lexical items in which deletion occurred, and/or in the distribution of function words, which are more susceptible to reduction processes

(Cutler, 1993) than content words. It raises the question, are mothers changing their pronunciation in a specific set of words? If so then is the frequency of these words changing over time, resulting in an apparent effect of change in deletion as children get older? Thus we conducted a more fine-grained second analysis to investigate the question of how widespread /h/ and /v/ deletion are across lexical items and word class.

In Analysis 2 we calculated frequencies and percentages of word-initial /h/ and word- final /v/ deletion in all the lexical items in which it occurred in the corpus.

CHAPTER 5 154

5.4.2 Analysis 2: Exploration of Local Effects on Casual Speech Processes

Deletion in lexical items. Table 5.7 shows the proportion of word-initial /h/ and word-final /v/ deletion across ages for each word that displayed the process. Words are listed in order of highest to lowest frequency of deletion (see ‘Total n deleted/contexts’ column in Table 5.7). Word-initial /h/ deletion occurred across a range of words at all timepoints, while word-final /v/ deletion only occurred in of, have and give. The /h/ items are separated by word class to examine the distributions of deletion in function and content words. At all child ages deletion was more frequent in function than content words (30.18% and 5.91% total deletion respectively). The inverted-v pattern of deletion over time is apparent in the function words for /h/ deletion. In content words /h/ deletion decreased slightly between timepoints, and /v/ deletion only occurred in three lexical items. Note that in both /h/ deletion in content words and /v/ deletion the N for deleted consonants in most lexical items is very low. In particular /h/ deletion was frequent in pronouns, the adverb here, and the verb have (as a main verb and an auxiliary verb). The words displaying the highest frequencies of deletion are relatively frequent overall at each time point, indicating that changes in rates of deletion are not due to the mothers introducing or removing words from their speech as children get older.

Within Table 5.7 it is clear that there are approximately seven words that are high frequency and involve considerable word-initial /h/ deletion and that the inverted- pattern of deletion over time is mostly driven by function words rather than content words. The seven words are the same words at each child age and account for 94.53% of /h/ deletion at age 1;6, 94.09% at age 2;0, and 94.57% at age 2;6. It is therefore of interest to examine how /h/ deletion changes over time for these words. Figure 5.4 shows the percentage of /h/ deletion over time for each word. Five of the seven words

CHAPTER 5 155

display an increase in deletion between 1;6 and 2;0 and a decrease between 2;0 and 2;6 and are therefore contributing to the inverted v-pattern of deletion found in Analysis 1.

The two words that do not follow this pattern are him and have. Deletion in him is the highest of all words, between 80.00% and 83.17%, and the lack of change may be due to a ceiling effect. /h/ deletion in have increases slightly at each timepoint.

The three words containing word-final /v/ deletion are of, have and give.

Deletion in of is relatively stable over time, between 30.00% and 31.45% while /v/ deletion in have displays the inverted v-pattern. Deletion in give displays the opposite pattern, although note that deletion in give has a very low number of occurrences at each time point and must be interpreted with caution. The inverted-v pattern of word- final /v/ deletion therefore appears to be driven primarily by the lexical item have.

Figure 5.4. Most frequent lexical items containing word-initial /h/ deletion at each

timepoint

CHAPTER 5 156

Table 5.7

Lexical Items Containing Word-Initial /h/ and Word-Final /v/ Deletion at Each Stage 1;6 2;0 2;6 TOTAL n % n % n % n deleted/ % deleted/ deletion deleted/ deletion deleted/ deletion contexts deletion contexts contexts contexts /h/ deletion: function words he 52/178 29.21 73/144 50.69 55/113 48.67 180/435 41.38 him 84/101 83.17 55/65 84.62 41/51 80.39 180/217 82.95 have 25/145 17.24 37/158 23.42 53/208 25.48 115/511 22.50 he’s 23/121 19.01 67/154 43.51 22/78 28.21 112/353 31.73 here 21/260 8.08 52/259 20.08 30/250 12.00 103/769 13.39 his 28/44 63.64 39/52 75.00 35/71 49.30 102/167 61.08 her 15/35 42.86 49/92 53.26 8/36 22.22 72/163 44.17 has 0/9 0.00 5/21 23.81 4/8 50.00 9/38 23.68 haven’t 2/11 18.18 4/10 40.00 2/10 20.00 8/31 25.81 himself 3/4 75.00 3/3 100.00 0/4 0.00 6/11 54.55 he’ll 2/6 33.33 1/7 14.29 0/1 0.00 3/14 21.43 how 0/46 0.00 2/67 2.99 1/81 1.23 3/194 1.55 had 1/12 8.33 0/13 0.00 1/24 4.17 2/49 4.08 hasn’t 0/2 0.00 2/4 50.00 0/1 0.00 2/7 28.57 herself - - 2/3 66.67 - - 2/3 66.67 who - - - - 1/20 5.00 1/20 5.00 Total 256/974 26.28 391/1052 37.17 253/956 26.46 900/2982 30.18 /h/ deletion: content words hang 1/30 3.33 1/32 3.13 1/33 3.03 3/95 3.16 head 1/14 7.14 0/21 0.00 2/17 11.76 3/52 5.77 hop 2/37 5.41 1/4 25.00 0/20 0.00 3/61 4.92 Heather 2/3 66.67 0/11 0.00 0/7 0.00 2/21 9.52 hit 0/5 0.00 2/5 40.00 - - 2/10 20.00 hold - - - - 2/15 13.33 2/15 13.33 Total 6/89 6.74 4/73 5.48 5/92 5.43 15/254 5.91 /v/ deletion of 35/112 31.25 39/124 31.45 66/220 30.00 140/456 30.70 have 8/145 5.52 31/158 19.62 7/208 3.37 46/511 9.00 give 4/34 11.76 2/46 4.35 2/20 10.00 8/100 8.00 Total 47/291 16.15 72/328 21.95 75/448 16.74 194/1067 18.18

CHAPTER 5 157

Gender differences in lexical items. Gender differences in the use of male and

female pronouns were also analysed. Table 5.8 shows the number of tokens and

percentage of deletion in male and female pronouns used by caregivers of boys (N=2)

and of girls (N=2). Relevant male pronouns are much more frequent overall than female

pronouns. Male pronouns in mothers’ speech to boys had a higher rate of deletion than

in speech to girls (52.30% and 40.06% deletion respectively). However, for male

pronouns, the inverted-v pattern of /h/ deletion occurred in speech to both boys and

girls. Female pronouns were more frequent in speech to girls but had a higher rate of

deletion in speech to boys (58.49%) than to girls (38.05%). The inverted-v pattern of /h/

deletion occurred in female pronouns in speech to girls, while in speech to boys the

opposite pattern occurred in female pronouns (though note the small number of tokens

at child ages 1;6 and 2;6).

Table 5.8

/h/ Deletion in Male and Female Pronouns by Mothers of Boys (N=2) and Girls (N=2)

at Each Child Age

1;6 2;0 2;6 Total n n n n deleted/ % deleted/ % deleted/ % deleted/ % contexts contexts contexts contexts Speech to

Boys Male 137/306 44.77 192/325 59.08 126/239 52.72 455/870 52.30 pronouns Female 12/16 75.00 15/32 46.88 4/5 80.00 31/53 58.49 pronouns Speech to

Girls Male 58/148 39.19 46/100 46.00 27/79 34.18 131/327 40.06 pronouns Female 3/19 15.79 36/63 57.14 4/31 12.90 43/113 38.05 pronouns

CHAPTER 5 158

Analysis 2 discussion. In Analysis 1 we found an inverted v-pattern of /h/ and

/v/ deletion as children aged from 1;6 to 2;0 to 2;6, and Analysis 2 was performed to examine /h/ and /v/ deletion in lexically specific contexts. Results showed changes in deletion occurred across a specific set of function words. The words containing the highest frequencies of deletion were function words and were the same across each child age. These words accounted for most of the deletion (94%) at each timepoint. This finding suggests overall changes in deletion were not due to mothers introducing new words or removing other words from their speech as children aged.

Analysis 2 also indicates most deletion occurred in function words. In English conversational speech function words tend to be unstressed and contain reduced vowel forms compared with content words (Cutler, 1993), so it is unsurprising that consonants are also reduced, or in this case deleted, in function words. Function words are also high frequency in child-directed speech relative to content words (Shi, Werker & Cutler,

2006), and the finding that phonetic variation is high in these words has implications for children’s word learning that will be returned to in the general discussion.

Possible gender differences were also revealed. Overall there was proportionately more /h/ deletion in mothers’ speech to boys than to girls in both male and female pronouns. The inverted-v pattern of deletion over time was found in male pronouns in all speakers, but in female pronouns it occurred in speech to girls but not boys. With only four speakers these findings cannot be generalised, but they do appear to support the Tyneside English study (Foulkes et al., 2005) which found that caregiver speech to boys contained more vernacular variants of [t] than speech to girls, which contained more standard variants. Foulkes et al. (2005) suggested that these gender differences may be due to mothers fine-tuning their speech to children’s developing gender identities, as in adult speech males tend to use more non-standard phonological

CHAPTER 5 159

variants than females. Horvath’s (1985) study on adult Australian English speakers from Sydney did find that /h/ deletion generally occurred more in male than female speech so the gender identity explanation could apply to the current findings, although more data are needed to explore this further.

5.3.3 Analysis 3: Effect of Speech Rate

A second possible explanation for the finding of change in deletion over time is that it is a function of speech rate, and the rate of deletion changes because mothers modify their rate of speech as children age. If this were the case we could expect to find an association between more deletion and faster speech, and that mothers’ speech rate across child age follows the same pattern as deletion, i.e. an increase between ages 1;6 and 2;0 and a decrease between 2;0 and 2;6.

To investigate this we calculated speech rate in words per second in each utterance, following the method used and validated by Ko (2012). Transcripts were segmented by utterance in Phon, where utterances were units of speech with pause boundaries at each end, and speech rate was calculated from the segmentation time stamps. Following the procedure of Ko (2012), paralinguistic notations (e.g. crying, laughter) were omitted and utterances were excluded when the duration was longer than

10 seconds (N=12), as these were likely to contain pauses that had not been segmented, and where they contained unintelligible speech and phonological fragments (N=396).

This left 5,856 utterances at age 1;6, 5,613 at 2;0 and 6,685 at 2;6.

A one-way ANOVA indicated a significant effect of child age on speech rate,

F(2, 18,151)=61.35, p<.001. Games-Howell post-hoc comparisons revealed mothers’ speech rate at child age 2;6 was significantly higher than 1;6 (M=0.20, 95% CI [0.15,

0.25]) p<.001), and 2;0 (M=0.18, 95% CI [0.13, 0.23]), p<.001. There was no significant difference between 1;6 and 2;0 (M=0.02, 95% CI [-0.03, 0.07]), p=.557.

CHAPTER 5 160

Thus, mothers’ speech rate did not significantly change as children aged from 1;6 to

2;0, and significantly increased between 2;0 and 2;6.

A more detailed analysis of speech rate in syllables per second was performed on a subset of 682 utterances in which word-initial /h/ and word-final /v/ deletion occurred. These are referred to here as deletion utterances (deletion occurred) and non- deletion utterances (containing word-initial /h/ or word-final /v/ contexts but with no deletion occurring). Overall there was little difference in mean speech rates between utterances containing deletion and no deletion, M = 3.26 and 3.06 syllables per second respectively. An independent samples t-test confirmed that the difference was not significant, t(681) = -1.354, p = .176.

In this subset of utterances, mean speech rate increased slightly at each timepoint (child age 1;6 M = 2.88, child age 2;0 M = 3.11 child age 2;6 M = 3.43). A one-way ANOVA showed a significant difference between child age for adults’ speech rate, F(2, 680) = 9.58, p<.001, adjusted R2 = 0.025, indicating child age contributed to only 2.5% of the variance in speech rate. A REGWQ post-hoc test indicated that speech rate was significantly faster at child age 2;6 than at 1;6 (p<.001) and 2;0 (p = .031). The difference between 1;6 and 2;0 was not significant (p = .156). Thus the results from the more fine-grained analysis of the subset were the same as those found in the analysis of speech rate in the whole sample: mothers’ speech rate did not significantly change as children aged from 1;6 to 2;0, and got slightly faster between 2;0 and 2;6.

5.5 General Discussion

The present study is the first to examine phonetic variation in maternal speech in a longitudinal corpus of Australian English. Two non-citation speech processes were analysed in mothers’ speech when their children were aged 1;6, 2;0 and 2;6. Based on the few studies that have investigated phonological variation in input to children after

CHAPTER 5 161

the first year of life, it was expected that variants containing consonant deletion would become more frequent in mothers’ speech as their children got older. Overall findings suggest that word-initial /h/ and word-final /v/ deletion are phonological processes in

Katherine English that occur across a set of lexical items. Mothers increase deletion as children age from 1;6 to 2;0 and decrease deletion as children age from 2;0 to 2;6, and these changes do not appear to be driven by speech rate. Results also showed interspeaker variation: the inverted-v effect over time appeared in three of the four speakers and was stronger in one of the speakers who displayed the pattern relative to the other two. The effect is therefore a preliminary finding and will need to be interpreted with caution.

The increase in the proportion of /h/ and /v/ deletion between child ages 1;6 and

2;0 is consistent with literature suggesting mothers use more non-citation speech processes and correspondingly fewer standard phonetic variants as children get older.

There was, however, an unexpected decrease in both /h/ and /v/ deletion in mothers’ speech between 2;0 and 2;6, although the results for /v/ deletion were not statistically significant at either interval. Previous research suggests that phonetic variation in speech input to children exposes them to socioindexical variation, and change over time is due to mothers’ gradually shifting their speech style to become more like interadult speech (Foulkes et al., 2005). While this may explain the current finding of an increase in deletion between 1;6 and 2;0, the question remains of why did deletion proportionately decrease in maternal speech between 2;0 and 2;6? The current findings suggest that mothers change their speech style as children get older, using more non- citation speech processes between 1;6 and 2;0 and then shifting back to using fewer processes at 2;6. Before interpreting these results in reference to mothers’ style-shifting,

CHAPTER 5 162

however, it was important to consider whether changes in deletion over time were due to changing lexical distributions in the corpus, or were a function of speech rate.

When we examined the distribution of /h/ and /v/ deletion across lexical items we found that deletion occurred in the same set of words at each timepoint, and that most of these words displayed the inverted-v pattern of deletion over time. The results provide evidence that /h/ and /v/ deletion are phonological processes occurring in a set of words, or one particular item (have) in the case of /v/ deletion, and not simply due to changing distributions of words with word-initial /h/ and word-final /v/ contexts.

Interestingly this analysis also showed that deletion was most frequent in function words rather than content words, and that the significant changes in /h/ deletion over child age were due to changes in function words. This is consistent with the literature on phonological differences in content and function words in English. At the prosodic level function words are frequently unstressed and at the segmental level they are likely to be realised with weak vowels, usually schwa (Cutler, 1993). The current study suggests that in addition to vowel reduction, function words contain high variation in consonants, relative to content words. Function words also have a high token frequency relative to content words in English, in both interadult and child- directed speech (see Shi et al. [2006] for calculations of type-token ratios in function and content words in corpora from CHILDES [MacWhinney, 2000]).

According to exemplar-based models it is possible that children may have more robust information about high frequency words in long-term memory, as a greater number of perceived tokens may result in a child having more distributional information about the word. It is, however, not known whether the frequency of tokens in the input is proportionately represented in the exemplar store. Phonological information in high frequency words in the input may in fact be less likely to be perceived and thus

CHAPTER 5 163

represented in memory because they are more phonetically reduced more often and contain fewer phonological cues than low frequency words (Cutler, 1993; Monaghan et al., 2005). Low frequency words might also have an advantage in novelty and saliency of new information.17 It would be interesting for future studies to examine longitudinally children’s perceptions and use of phonological cues in high and low frequency items to investigate any relationship with the current finding of variation in the input at different child ages.

We tested another alternative explanation for change in deletion in maternal speech over time by analysing the mean speech rate of each speaker at each time point.

It is well-known that speech rate is related to segmental deletion, with deletion more likely in faster speech (e.g. Koreman, 2006). It is therefore possible that the inverted-v pattern of deletion found in the current study was a result of mothers using faster speech at child age 2;0 than at 1;6 or 2;6. Speech rate may also play a subtler role as a component of style. Both high speech rate and segmental deletion are features of hypoarticulation, or non-citation speech, where reduction is the result of a trade-off between production ease for the speaker and perception ease for the listener (Lindblom,

1990). If change in speech rate was accounting for the change in deletion found in the current study then we expected speech rate would display a similar inverted-v pattern over the three timepoints. Instead we found that overall speech rate did not significantly differ between1;6 and 2;0 and increased between 2;0 and 2;6, although the latter effect was small at less than half a syllable per second difference. Further, there was no significant difference in speech rate between utterances containing /h/ and /v/ deletion and utterances where /h/ and /v/ were fully realised. These findings indicate changes in deletion over time were not due to changes in speech rate, as deletion and speech rate

17 We thank an anonymous reviewer for raising these issues for us.

CHAPTER 5 164

both displayed different patterns of change over the three timepoints. The lack of difference in speech rate between utterances with and without /h/ and /v/ deletion provided further evidence that in this sample deletion is not, or at least not primarily, driven by speech rate changes.

It appears that the change in deletion over time found in this study was not an artefact of changing distributions of lexical items, nor could it be accounted for by mothers’ speech rate. We thus return to the possibility that rates of /h/ and /v/ deletion in maternal speech change with the age of the child. There are at least two plausible explanations for this: 1) that mothers are modelling forms differentially as children age, and 2) that deletion rates vary with types of speech acts (e.g. questions, recasts, and repetitions, which could be frequent as children become more verbal) and social interactions (e.g. play, teaching, behaviour management) and the distributions of these shift with child age. Note that these are not mutually exclusive explanations and both may be components of fine-tuning – mothers (consciously or not) adjust phonetic variation over time as children develop knowledge of the phonology of their home language as well as across the different speech acts and social contexts in which variation is used. Because there is so little previous research on sociolinguistic markers in regional Australian English, it is not clear whether /h/ and /v/ deletion contain socioindexical information beyond being non-citation speech processes, and further research on this is needed to draw any strong conclusions on the indexical information children may acquire from being exposed to this variation.

Recent studies on phonetic variation in caregiver speech examined different indexical variables to those in the current study. They found a slow linear increase in mothers’ use of non-standard variants over time as their speech gradually became more like interadult speech (Foulkes et al., 2005; Smith et al., 2009). However, the mixed

CHAPTER 5 165

findings in older studies on clarification and reduction processes in child-directed speech did suggest that the different findings may be due to the different ages of the children in the studies (Cruttenden, 1994). Vowel clarification was found mostly in speech to children who were pre-linguistic or producing single words (Bernstein-Ratner

1984a), while speech to older children was found to contain more consonant reduction and unintelligible segments than speech to adults (Bard & Anderson, 1983; Shockey &

Bond, 1980). The non-linear change over time found in the present study is consistent with Cruttenden’s (1994) argument that the apparent mixed findings of these early studies may be explained by the different functions of child-directed speech at different ages. The current finding of an inverted-v pattern in deletion may indicate that mothers follow clarification modifications with a period of reduction modifications as children start producing multi-word utterances (by around 2 years of age), and then revert to clarification (by 2;6). Further research is needed to investigate whether the current finding of the inverted-v pattern over time is specific to the phonological processes of

/h/ and /v/ deletion or whether it is also found in other phonological variables in speech to children ages 1;6 to 2;6.

If phonetic variability in speech to children serves different functions at different times then we might expect mothers to fine-tune differentially as their child’s language develops. Mothers’ use of more standard clarified forms may be particularly important when frequency distributions of exemplars are relatively newly formed in children’s memories, and use of deleted or reduced forms may be more important when children are learning the alternate phonetic forms common in non-citation interadult speech and the socioindexical associations. Full standard forms may then become important again as children start producing multi-word utterances and mothers become aware of their child’s pronunciation. This may explain the decrease in deletion after 2;0, as this is

CHAPTER 5 166

when most of the children start producing multi-word utterances and could indicate that mothers are starting to fine-tune to children’s productive language.

Another possibility is that mothers’ use of phonetic variation is related to types of speech acts and social interactions, and that there was less deletion in more formal contexts. Previous research has found relationships between mothers’ use of sociolinguistic markers and speech style, with standard phonetic variants more frequent in formal contexts and the local vernacular variants more frequent in informal contexts

(Smith et al., 2007). It would be of interest to do further research with the current corpus to investigate whether /h/ and /v/ deletion is related to speech style, interaction contexts and sentence types, and whether the distributions of these change in mothers’ speech as children get older.

The results of this study reveal a non-linear change in deletion in maternal speech over time; further research is needed to understand why mothers make differential segmental modifications as their children age. The current study is limited by the relatively small number of participants, and studies with more speakers are needed in order to understand the nature and extent of individual differences in phonological variation. These findings open several avenues for further investigation.

For example future studies could examine the phonological modifications mothers make when they are specifically modelling speech for children, such as through repetition and explicit teaching which may involve hyperarticulation, and how this changes as children progress from producing single-word to multi-word utterances. Further research could also investigate the relationship between deletion and speech rate in maternal speech, as the current finding that speech rate increases as deletion decreases between 2;0 and 2;6 in mothers’ speech was unexpected. In particular it would be useful to have more timepoints to examine how speech rate and deletion change over time and whether

CHAPTER 5 167

deletion ‘catches up’ to speech rate, or vice-versa, in relation to children’s productions.

Further timepoints and more speakers are also needed in order to understand interspeaker variation and whether there are different trajectories of phonological modifications in maternal speech over time, and again how these may relate to the child’s language development. Further investigation of children’s productions over time is needed to test different theoretical explanations for the pattern of adult speech found here. Experimental investigation of children’s receptive knowledge of phonetic variation at different ages would also be relevant to help elucidate how their phonological representations are shaped by phonetic variation in the input.

CHAPTER SIX

Investigating Native Speaker Judgements of Phonetic Variation in Gurindji Kriol Using Visual Analogue Scales CHAPTER 6 170

6.1 Abstract

Phonetic transcription is primarily a perceptual task and can be subject to perceptual biases. Native speakers bring implicit linguistic knowledge that can be useful for understanding the sound system of the language, and for checking the reliability of phonetic transcription. In field linguistics, however, it is often not feasible to train native speakers to the technical level required for IPA transcription. In this study we investigated the use of Visual Analogue Scales (VAS) for eliciting perceptual judgements on phones from native speakers of Gurindji Kriol, an Australian Aboriginal contact language variety.

Recordings of conversational Gurindji Kriol made by Meakins (2011) for the

ACLA-1 project were phonetically transcribed by non-native speaking transcribers.

Tokens for this study were selected from words that potentially contained a fricative or affricate across a range of places of articulation in initial, medial and final word positions. Tokens were selected where the non-native transcribers had both agreed on the target segment and disagreed and/or marked the target segment as ambiguous. 73 visual analogue scales were constructed in total for the target segments in each token.

Community research assistants (RAs) who were native speakers of Gurindji Kriol participated in phonological awareness training and completed the VAS task.

RA responses on the scales were compared to the non-native transcriber IPA transcriptions for each token. Results showed both agreements and discrepancies between RAs and transcribers on different judgements in each word position. These findings provided possible explanations for ambiguities in the phonetic transcription and suggest several hypotheses for further research on phonetic variation in Gurindji

Kriol. CHAPTER 6 171

6.2 Introduction

Phonetic transcription is a useful tool clinically and in the field, however it does have limitations, particularly when the transcribers are non-native speakers of the language being analysed. Native speakers have implicit linguistic knowledge that is beneficial to analyses and interpretation of phonetic variation in the language, but often in field linguistics it is not practical to train speakers in IPA to the level required to contribute to the phonetic transcriptions. IPA transcription also requires categorical judgements of speech sounds, which may limit our understanding of phonetic variation that is potentially continuous. The purpose of this study was to gain a deeper understanding of phonetic variation in a language that is an L2 for the analysts by eliciting native speaker judgements of phonetic variability using a continuous scale.

Visual analogue scales have been used in previous studies to elicit continuous perceptual judgements on speech segments for the purpose of examining fine phonetic detail. For example this method has been used to investigate adults’ sensitivity to within-category variation in stops and fricatives (Skorniakova & Ito, 2011) and stop voicing contrasts (Kong & Edwards, 2011), as well as adult perceptions of covert contrasts and phonetic accuracy in children’s speech (Julien, Munson & Edwards, 2012;

Munson, Edwards, Schellinger, Beckman & Meyer, 2010; Schellinger, 2008; Urberg-

Carlson, Munson & Kaiser, 2009). In the current study we explore the use of visual analogue scales in fieldwork to obtain native speaker judgements of speech segments in

Gurindji Kriol, a phonetically and phonologically underdescribed mixed language spoken by Gurindji people in northern Australia. The purpose of eliciting native speaker perceptions was to add information to our phonetic transcription in order to gain a deeper understanding of phonetic variation in adult conversational speech. CHAPTER 6 172

The current research is part of a larger investigation of the phonology of maternal speech in Gurindji Kriol. Gurindji Kriol is the everyday language of traditionally Gurindji speaking communities in the Northern Territory, Australia, and an overarching aim of the project is to examine phonetic variation in the input that Gurindji children are exposed to. One particular kind of variation we are investigating is in fricatives, which contain variability in voicing, manner and place of articulation. The data for the larger project came from an existing longitudinal corpus of conversational

Gurindji Kriol maternal speech (Meakins, 2011, pp. 45-47), to which we added phonetic transcription and analysis. Phonetic transcription was conducted by native Australian

English speaking linguists because it was not practical to train native speakers of

Gurindji Kriol to the technical level required for IPA transcription. Consideration of native speaker phonological judgements was an important part of our methodology and theoretical understanding of variation in Gurindji Kriol, and this formed the basis of the present study.

The background section is structured as follows. First we discuss previous studies that have used visual analogue scales to measure perception of speech sounds.

We then consider some limitations of phonetic transcription and potential biases, including transcriber language background. We finish up this section with more detail about the current applied context and the rationale for examining fricative variation in

Gurindji Kriol.

6.3 Background

6.3.1 Measuring Speech with Visual Analogue Scales

Eliciting perceptual judgements using visual analogue scales involves asking listeners to indicate their perception of a speech stimulus on a horizontal line with two different sounds at the endpoints. Differentiation of the endpoint sounds depends on the CHAPTER 6 173

process being investigated; for example the endpoints can differ acoustically within a phonological category to test listener sensitivity to within-category detail (e.g. Kong &

Edwards, 2011; Skorniakova & Ito, 2011), or they can represent different phonological categories to test continuous variation between segments or intermediate productions

(e.g. Munson et al., 2010).

The visual analogue scale is useful as a methodological tool because it is appropriate to use with naïve listeners. Recent research on children’s speech productions found it to be an effective way to measure naïve listeners’ perceptions along a continuous scale. Urberg-Carlson et al. (2008) found that adult listener ratings of children’s productions using visual analogue scales were generally well correlated with acoustic measures, and Munson et al. (2010) found naïve listeners reliably distinguished between correct and intermediate productions and between correct productions and substitutions by responding on the scales for /s/-/θ/ and /d/-/ɡ/ contrasts.

It is important to note that phonetic training and experience is still advantageous to the transcription process. A recent study found that experienced speech-language pathologists had greater reliability and sensitivity to acoustic properties of segments than naïve listeners in their visual analogue scale judgements of children’s productions

(Munson et al., 2012).

Visual analogue scale responses are reliable measures of perceptions of gradient voicing changes. Kong and Edwards (2011) used visual analogue scales and eye- tracking to examine changes to stop consonant voicing, and found that adult listeners were sensitive to gradient changes in VOT and fundamental frequency in stop consonants (Kong & Edwards, 2011). Within-category variation in stop consonant voicing has also been analysed using the scales in a goodness-rating task, which found listeners made intermediate judgements of stop and fricative VOT contrasts when CHAPTER 6 174

stimuli were acoustically intermediate, though overall fricatives were perceived more continuously than stops (Skorniakova & Ito, 2011).

Visual analogue scales were used in the current study for two main reasons: first, they allow for continuous judgements of speech sounds, which may provide further insight into variation of fricatives in Gurindji Kriol; second, they are effective in eliciting judgements of speech sounds from naïve listeners who do not necessarily have high literacy or technical knowledge of IPA transcription. Using visual analogue scales with native speakers may add another dimension of information to that gained through phonetic transcription and analysis.

6.3.2 Limitations of Phonetic Transcription

Phonetic transcriptions are records of a transcriber’s perception of auditory events and involves the transcriber identifying speech segments and assigning each perceived phone a category using IPA notation. It has long been recognised that there are many different factors that can compromise validity and reliability in phonetic transcription. Early studies in the field have shown that validity can be affected by the transcriber’s language background and transcribing experience, as both of these influence perceptual expectation effects (Oller & Eilers, 1975). The effect is not necessarily negative – Oller and Eilers (1975) found that transcription accuracy was generally improved when the transcriber knew the ‘correct’ meaning of the utterances and suggested that knowing the target form (i.e. what to expect perceptually) may help transcribers to direct their attention to relevant phonetic features.

By the very nature of phonetic transcription IPA judgements are categorical, although fine phonetic detail can be represented in IPA with the use of diacritics. If variation in production is potentially continuous, for example as voicing (Jones &

Meakins, 2012a) and possibly stop-fricative continuancy variation may be in Gurindji CHAPTER 6 175

Kriol, then transcribing using IPA categories could be somewhat artificial. The potential artificiality of transcription underlies some of the issues in discrepancies between multiple transcribers. Reasons for transcription discrepancies include systematic perceptual differences and ambiguous productions, which may be influenced by the transcriber’s background. Transcription discrepancies thus raise the question of whether disagreements are due to errors or whether the segment is an intermediate production and influenced by the transcriber’s own expectations and phonological categories, in turn leading to the question of whether native speakers categorise these segments more reliably (Edwards & Beckman, 2008).

Reliability in phonetic transcription refers to the consistency with which IPA categories are applied, both within and between transcribers. It is commonly measured by checking agreement between multiple blind transcriptions. In a review study of phonetic transcription reliability, Shriberg and Lof (1991) analysed phonetic transcriptions of five teams of trained transcribers and described a heuristic of four main sources of variance that can be addressed to improve reliability: 1) Participants factors, or characteristics of the speaker. These include the amount and type of speech errors made (Kearns & Simmons, 1988), intelligibility (Shriberg & Lof, 1991), and canonicity of speech sounds (Ramsdell et al., 2007). 2) Analysis factors, such as the level of detail in the transcript, specific agreement criteria and type of agreement (Cucchiarini, 1996;

Shriberg & Lof, 1991). 3) Linguistic contexts, including structural, grammatical and stress forms (Shriberg & Lof, 1991), stylistic factors (Goddijn & Binnenpoorte, 2003), and immediate phonetic environment and word-position (Hardison, 2003). 4) Phonetic and phonological units such as class, features, sounds and diacritics (Shriberg & Lof,

1991). CHAPTER 6 176

Although there are a range of potentially interacting factors that can affect reliability and validity, phonetic transcription remains a useful tool clinically and in the field. Reliability can be tested by assessing intertranscriber agreement, where a second person transcribes a portion of the recording and the percentage of segments agreed on is calculated. Cucchiarini (1996) points out that this approach assumes that all disagreements are equal, i.e. a disagreement based on one feature is treated the same as a disagreement based on several features. Another concern is with how disagreements are treated – whether they are excluded or discussed among transcribers to reach a consensus, or whether they represent informative systematic differences (Edwards &

Beckman, 2008). In addition, high reliability does not indicate high validity. Even if transcribing blind, that is without knowledge of each other’s transcriptions, multiple transcribers may be subject to the same perceptual biases, resulting in a transcription that has high intertranscriber reliability but that is not necessarily an accurate representation of the speech data.

6.3.3 Language Background Affects Phonetic Transcription

Phonetic transcription reflects the transcriber’s perception of auditory events and involves the transcriber identifying categorical speech segments. Variation in speech production, however, is gradient and categorical perceptual judgements depend to some extent on the listener’s own language-specific phonological categories (e.g. Coussé et al., 2004; Hardison, 2003; Strange & Jenkins, 1978). Native speaker transcriptions are therefore considered valid, as phones are judged using the phonological categories of the speech community (Edwards & Beckman, 2008). Where phonological contrasts of a language are undetermined, as is often the case with underdocumented languages, phonetic transcription may be somewhat artificial and cannot necessarily be interpreted in terms of native speaker phonological categories. CHAPTER 6 177

Factors that influence auditory perception can influence categorical perception of speech segments and thus also phonetic transcription, for example linguistic context

(Hardison, 2003), sentence context (Warren & Warren, 1970), and listener expectations

(Oller & Eilers, 1975; Ralston & Johnson, 1990). The listener’s phonological categories, determined by his or her language background, also influence these factors, so language background potentially has both direct and indirect biases on phonetic transcription.

Indeed, research shows that a transcriber’s own language background and experience can affect phonetic transcription (Coussé et al., 2004), and that it takes training and practice to become sensitive to non-native contrasts (Pollock & Hinton,

2001). For example, Pruitt et al. (2006) compared native American English speakers’ and native Japanese speakers’ perceptions of dental and retroflex stop consonants in

Hindi. These sounds do not phonemically contrast in either language; however Japanese has a similar contrast between /d/ and flapped /r/, which is sometimes produced as a retroflex. it was found that Japanese speakers identified the Hindi contrast more accurately than American English speakers (Pruitt et al., 2006). These findings suggest that distinguishing non-native contrasts may be easier when the categories are comparable to those in the listener’s native language.

The effect of language background can also be seen by analysing inter- transcriber agreement for phonetic transcription of native and non-native speech.

Ramsdell et al. (2007) compared phonetic transcription agreement of infant vocalisations across eight transcribers who were native American English speakers.

Agreement was significantly higher (r=0.64) for sounds found in American English

(including allophones) than for sounds outside the phonetic range of American English

(r=0.44). Thus reliability was lower for transcriptions of non-native than native segments, demonstrating the effect of phonetic familiarity on transcription reliability. CHAPTER 6 178

Language background can also affect transcription through lexical expectation effects, as knowing the target word causes the listener to perceive the phonetic features they expect to hear, even in some contexts where those features are not acoustically present

(Louko & Edwards, 2001; Oller & Eilers, 1975).

In summary, although phonetic transcription in IPA is useful for many purposes, it does have limitations and should be supplemented with other types of analysis where possible. Visual analogue scales have been used in the child language literature to gain further understanding of continuous variation in child speech and adult sensitivity to covert contrasts. In the current study we tested whether visual analogue scales could be used to record native speaker judgements of fricatives in Gurindji Kriol, and examined how these data could further our analyses and interpretation of our phonetic transcription data.

6.3.4 Current Applied Context

The current study is part of a larger research project on maternal speech in

Gurindji Kriol, a mixed language spoken by Australian Aboriginal people in Kalkaringi and Daguragu communities in northern Australia. Gurindji Kriol is a relatively new language, emerging in the 1970s from code-switching (McConvell & Meakins, 2005).

Gurindji Kriol speakers are fluent and literate in varying degrees in English, which is the language of the local school and administration in the community. Many speakers are also fluent in Kriol and other Aboriginal languages spoken in the region.

Gurindji Kriol combines lexical items and grammatical structures of Gurindji, which is a traditional Australian Aboriginal language, and Kriol, which is an English- lexifier creole spoken across northern Australia. Lexically, 36.6% of vocabulary in

Gurindji Kriol is derived from Kriol, 35% from Gurindji, and the remaining 28.4% are CHAPTER 6 179

synonymous forms from both languages (Meakins, 2011, p.19)18. There are both similarities and differences in sounds compared to Australian English, which is the researchers’ language background. The phonology of Gurindji is similar to many traditional Australian languages, in that the consonant system contains many contrasting places of articulation and relatively few contrasting manners (Butcher, 2006). For example there are palatal and retroflex places of articulation for stops, but no contrast between stops and fricatives. There also does not appear to be evidence of a voicing contrast in Gurindji Kriol stops (Jones & Meakins, 2012a). Traditional Gurindji does not contain phonemic fricatives, although stops may be realised as fricatives in some contexts due to lenition processes. In Gurindji Kriol fricatives are highly variable in

Kriol-derived words, both within fricatives (voicing and place of articulation variation) and with stops. In Kriol-derived words most fricatives are labio-dental and alveolar, and occur in word-initial position, followed by medial with proportionately fewest in word- final position (Buchan et al. in preparation). These observations are consistent with the consonant inventory of Gurindji Kriol proposed in Jones and Meakins (2012a), adapted from Meakins (in press) and shown in Table 6.1.

18 This count is based on a 200 word Swadesh list. CHAPTER 6 180

Table 6.1

Proposed Gurindji Kriol consonant inventory (Jones & Meakins, 2012b; Meakins, in press)

Bilabial Apical Apical Alveo- Velar alveolar post- palatal alveolar Stop p t ʈ ɟ k

Fricative f s ʃ

Nasal m n ɳ ɲ ŋ

Lateral l ɭ ʎ

Rhotic r ɹ ɻ

Glide j w

The phonetic transcribers of the data in this study are native Australian

English speakers with varying degrees of first-hand experience with Gurindji Kriol. In the current study one transcriber (the first author) has visited the community several times, and the second transcriber was not familiar with Gurindji Kriol before starting transcription. We refer to these transcribers here as ‘non-native transcribers’. The community Research Assistants employed to work on the project are native Gurindji

Kriol speakers and are also fluent English speakers. The community RAs do not have technical expertise in IPA transcription and their phonological awareness is likely to be variable but relatively low at the phoneme level, that is, the ability to segment phones in words. The specific focus for this study is on the transcription of fricatives in maternal

Gurindji Kriol.

CHAPTER 6 181

The aim of the current study was to investigate native speaker perceptions of fricatives in conversational Gurindji Kriol maternal speech by eliciting native speaker judgements on continuous visual analogue scales. Research questions were:

• Methodologically, how can we use visual analogue scales in the field

to record native speaker perceptions of fricatives in Gurindji Kriol?

• How do native speaker judgements relate to non-native transcriber

categorical judgements, and do they differ depending on word position

and type of judgement?

6.4 Method

6.4.1 Non-Native Speaker IPA Transcriptions

The corpus. The recordings were from a longitudinal corpus of conversational maternal Gurindji Kriol made by Meakins (2003-2007) for the Aboriginal Child

Language Acquisition (ACLA-1) project (Meakins, 2011, pp. 45-47). For the current project a subset of the Meakins corpus was selected. These were recordings of three mothers, SS, AR and CE, at three time-points when their children were aged around 1;6,

2;0 and 2;6.

Non-native transcribers. Phonetic transcriptions were made by the first author and a Research Assistant who has formal training in phonetic transcription. Both transcribers are native Australian English speakers. The first author has had direct experience with Gurindji Kriol.

Procedure. First, orthographic transcripts made by Meakins were imported into the phonetic transcription and analysis program Phon (Rose, Hedlund, Byrne, Wareham

& MacWhinney, 2007). Phon was used to search the orthography of all transcripts for words that may be pronounced with a fricative in Gurindji Kriol, that is Kriol words that have a fricative in the English cognate word. The orthographic transcription of Gurindji CHAPTER 6 182

Kriol is not phonetic. The orthography was developed over a long period of time in close consultation with community members. It uses English letters and does not have one-to-one correspondence with Gurindji Kriol phonology. Thus the orthography was searched by the authors, who had prior knowledge of Kriol words and their English meanings, rather than searching for letters or letter combinations in the orthographic transcription. All tokens of these words were transcribed phonetically by the first author and the Research Assistant. These transcriptions were used to select the tokens and judgements for the visual analogue scales.

6.4.2 Visual Analogue Scale Development

Token selection. A sample of transcribed tokens was selected for checking with

Gurindji Kriol native speakers. The sample comprised tokens that contained the target segment (fricative or potential fricative) at four places of articulation (labial, dental, alveolar, palato-alveolar) in each word position (initial, medial, final). Also included were word-initial and word-medial affricates, and word-initial /h/. Two tokens in each of these conditions were randomly selected for each of the three speakers: one where the two initial non-native transcribers agreed on the phonetic transcription, and one where there was a discrepancy. The tokens selected for checking were rated for difficulty on a scale of 1 (easy) to 3 (hard). Some positional segments are rarer than others and were not found for some speakers, for example word-final th. In total 73 tokens were selected to validate with native speakers.

Judgements. Generally, uncertainties and discrepancies in the phonetic transcription involved ambiguity between two sound categories, for example [v] and [b].

Visual Analogue Scales were set up so that native speakers could judge where the target segment lies on a continuum between two sound categories. A scale was created for each target segment of the 73 tokens with each one having two sound categories as CHAPTER 6 183

anchors, one at each extreme of the scale. For the tokens where the two non-native transcribers disagreed on the phonetic transcription of the target segment these anchors were the IPA transcriptions of each on-native transcriber. For example, for a token of the word binij 'that's it!

Table 6.2

Number of each type of judgements in the Visual Analogue Scales (N = 73)

judgement type example n

stop – fricative [c] – [ʃ] 32

voiced – not voiced [z] – [s] 17

affricate – stop [dʒ] – [ɟ] 12

affricate – fricative [tʃ] – [ʃ] 2

other [ð] – [z] 10

Table 6.2 shows that the largest number of judgements were stop – fricative, followed by fricative voicing judgements. Included in the ‘other’ category were place of articulation judgements, sound pairings that differed by more than one feature (e.g. [ʃ] – CHAPTER 6 184

[ɟ]), and the word-initial [h] judgements. With the latter the judgement was whether [h] was present or not; in these scales the alternative anchor to [h] was the first vowel sound in the corresponding token. The positioning of each type of sound on either the left or the right of the scale was counter-balanced to avoid any right-side bias. For example in the twelve affricate-stop judgements the affricate appeared six times on the left of the scale and six times on the right. The specific judgements and frequencies for each judgement type are shown in Table 6.3 for word-initial sounds, Table 6.4 for word- medial and Table 6.5 for word-final. See Figures 6.2, 6.3 and 6.4 for the lexical items associated with each judgement.

Table 6.3

Word-initial judgement type frequencies

affricate- stop-fricative voicing affricate-stop other fricative

judgement n judgement n judgement n judgement n judgement n

b – v 2 f – v 4 c – tʃ 1 tʃ – s 1 θ – s 1

c – s 1 θ – ð 1 ɟ – dʒ 5 ʃ – dʒ 1 h – Ø 6

d – ð 2 z – s 2

ʃ – c 2

ʃ – ɟ 1

t – s 1

θ – c 1

θ – t 1

Total 11 7 6 2 7

CHAPTER 6 185

Table 6.4

Word-medial judgement type frequencies

stop-fricative voicing affricate-stop other

judgement n judgement n judgement n judgement n

ɟ – ʒ 1 z – s 3 dʒ – ɟ 6 z – ɟ 1

ɟ – ð 2 ʃ – s 2

d – ð 1

ʃ – ɟ 2

v – b 5

Total 12 3 6 3

Table 6.5

Word-final judgement type frequencies

stop-fricative voicing

judgement n judgement n

b – v 1 f – v 2

c – ʃ 5 z – s 5

ɟ – ʃ 1

d – ð 1

f – p 2

Total 10 7

CHAPTER 6 186

Scale order. First, tokens were grouped into three blocks by word position.

Tokens with the target segment in word-initial position appeared first, as word-initial sounds were easier for RAs to isolate. The second block contained word-final target segments, and the third word-medial. Within each block the scales were ordered so that the same sound pairings were together. Within pairings they were then ordered by the difficulty ratings that were assigned during token selection.

Scale construction. The scales were constructed in PowerPoint. Audio tokens of each IPA target phone were downloaded from the UCLA phonetics website

(Ladefoged, last accessed 12/12/2011) and embedded into the PowerPoint slides as anchors. Each of the 73 scales were on separate slides. Each sound was also associated with a colour, as in the phonological awareness training Research Assistants were trained to isolate sounds in words by identifying each sound with a colour. Using colours in combination with the audio files also helped the researcher avoid both spelling the sounds and producing the sounds in her Australian English accent. Sounds could instead be referred to by the associated colour, e.g. [b] was ‘the yellow sound’, and kept constant between presentations. Colours were displayed as a gradient to represent the potential continuum between sound categories. Each scale had a light 7- point graph-type background to assist coding and analysis of responses. An example scale is shown in Figure 6.1. The anchors on this scale were [b] (yellow) and [v] (red).

The audio files for these sounds were played by clicking on the speaker symbols. CHAPTER 6 187

Figure 6.1. Example Visual Analogue Scale for [b] and [v]

6.4.3 Phonological Awareness Training

Research Assistants. Three women in the community participated in phonological awareness training. The women were recommended by traditional owners and the community project liaison, and were paid as research assistants. Two of the

Research Assistants, Kirsty and Trisha, were young women (age 20 and 22 years respectively) and the third, Rosemary, was an older woman who works as an interpreter in the community. Rosemary had taken several interpreter training courses and has higher literacy than the other two women.

Materials. A set of 108 picture cards measuring 3”x4” were used as stimuli. The set contained commonly used words and most had Gurindji counterparts. English words for the pictures contained CVC minimal pairs, words with consonant digraphs, words with consonant clusters, words with vowel digraphs, and multisyllabic words. Other CHAPTER 6 188

materials used were a set of small coloured blocks. There were about one hundred blocks in total of five different colours that were used to represent segments in words.

Procedure. Training was conducted over two sessions, each three to four hours in length. Session 1 consisted of an introductory discussion about sounds in speech, syllable segmentation, sound isolation, and sound matching tasks. In Session 2 the sound isolation tasks were extended using coloured blocks to represent different sounds in words. For example bag was represented as red-green-yellow and bat as red-green- blue. Using minimal CVC pairs this task helped RAs to separate sounds from spelling and to attend to and discriminate different sounds. Difficulty progressed from English

CVC words to words with consonant clusters and vowel diagraphs, and Gurindji words.

These tasks were precursors to the Visual Analogue Scale task, where sounds from conversational recordings of Gurindji Kriol were represented using colours.

6.4.4 Scale Administration

Research Assistants. The visual analogue scale task was administered in a separate trip following the one in which the phonological awareness training was conducted. Three young Gurindji women were paid as RAs to do the task although they did not undertake the training. Rosemary, the interpreter who participated in the training, did not participate in the task. Thus in total five young women who were native speakers of Gurindji Kriol participated in the visual analogue scale task. The women who participated in the phonological awareness training assisted the women who did not; although it is a limitation of the procedure that not everybody did the training.

Procedure. The PowerPoint slides were printed in colour and bound into booklets, with one scale per page. The Research Assistants were given a booklet each while the researcher had the PowerPoint version open on a laptop with external speakers.

Before each word position block, Research Assistants were reminded how to isolate CHAPTER 6 189

sounds in that word position, as they were taught in the earlier phonological awareness training. RAs who participated in the training helped the three who did not. For each scale the audio files of the anchors were played. To parallel the IPA transcription context, the utterance containing the corresponding token was then played after RAs were told which word to listen for in the utterance. Anchors and utterances were played as many times as needed. RAs were instructed to mark on the scale in their booklets where they thought the target segment lay between the two sounds on the scale. They were asked to tell the researcher if they thought the target segment was something different to the two sounds in the scale, in which case the researcher noted the alternative sound. RAs were also asked to say if they thought a scale was too hard or if the target segment was unclear.

RAs did the task over several sessions in groups of two to five. Often there was discussion among the group and some group decisions were made, however RAs also made individual responses, which are apparent in between-RA response discrepancies.

Analysis. RA responses were measured from the 1-7 scale to two decimal places.

Box and whisker plots were made to display collated responses to each judgement type in each word position. These were then overlaid with the scale anchors and the non- native transcriber transcriptions to compare the RAs’ continuous responses on the scales with the transcribers’ categorical IPA judgements.

6.5 Results

Data were split by word position and analysed separately for initial, medial and final positions. Based on discussions during the visual analogue scale judgment sessions with the RAs, mid-point responses are interpreted to mean that RAs perceived the sound as midway between the two anchors; ‘other’ and ‘don’t know’ responses were recorded separately. Figures 6.2, 6.3 and 6.4 were constructed to show the RA responses and the CHAPTER 6 190

non-native speakers’ IPA transcriptions (henceforth ‘non-native transcriptions’). The box and whisker plots in the figures represent the pooled RA responses to each judgement pair, and the dots represent the categorical judgement of each sound as transcribed in IPA by the non-native listeners. Green dots show that both non-native listeners agreed on the transcription, and purple dots represent non-native transcriber disagreement or ambiguity (i.e. one or both transcribers were uncertain).

6.5.1 Word-Initial

Word-initial judgements are shown in Figure 6.2. The first eight judgements along the x-axis are the stop-fricative judgements, and the next three are voicing, followed by four stop-affricate judgements, and then [h] present or absent and one fricative place judgement. For the first two pairings, [b, v] and [c, s], there was close agreement among RAs who all judged the sounds as stops, and disagreement between

RA judgements and the non-native transcriptions. The purple dots indicate there was also disagreement between the two non-native transcribers on both of the fricative transcriptions, [v] and [s] (baldan ‘fall’ and jeya ‘there’, respectively). For [d, ð] RA judgements agreed with those of the non-native transcribers. For [c, ʃ] RAs tended to agree with non-native transcribers on the palatal stop although their responses were continuous, indicating that RAs did not judge the sound as being entirely like the stop endpoint. RAs also gave continuous responses for the [ʃ, ɟ] sound. RAs responded closer to the stop than the fricative as they also did for the initial sound in jouim; non- native transcribers judged it as a fricative although also noted it as ambiguous.

The next judgement involved the [st, t] judgement. There was non-native transcriber disagreement about whether the initial sound was [st] in the word top ‘be,

‘whatchamacallit,

The next three word-initial judgements involve voicing distinctions. For the [f, v] pairing RAs gave both categorical (indicated by the whiskers extending to the endpoints) and continuous responses. In judging dental fricative voicing, [θ, ð], non- native transcribers disagreed and RAs all gave mid-point responses. For [s, z] non- native transcribers both judged the fricatives as voiceless, while RAs judged them as voiced.

RAs perceived stops for the voiced stop-affricate judgements [ɟ, ʤ] with one midpoint response, disagreeing with non-native transcribers who perceived the English- based affricate. For the voiceless stop-affricate judgement [c, ʧ] in jinek ‘snake’, RAs judged the sound as the palatal stop and there was disagreement between non-native transcribers. There was general agreement between RAs and transcribers on the fricative-affricate judgements [ʧ, s] and [ʃ, ʤ] that these sounds were affricates.

Word-initial [h] judgements involved six tokens: non-native transcribers perceived [h] as present in three tokens and absent in the other three tokens. With one endpoint as [hV] and one as the following vowel in the word with no [h], RAs all responded strongly towards the [h] endpoint. The final word-initial judgement was [θ, s] in a token of ‘three’, an English code-switch. While non-native transcribers perceived

[s], RAs judged the sound as [θ].

Overall word-initially, the RA responses on the visual analogue scales provided insight on the transcriptions that were ambiguous and/or where non-native transcribers disagreed. Where these sounds involved stop-fricative distinctions, the native speaking

RAs generally judged them as stops. For two of the three voicing pairings RAs tended to give continuous responses. Judgements involving palatal stops and affricates were CHAPTER 6 192

generally perceived as palatal stops by RAs, but as affricates by non-native transcribers.

Initial [s] was always judged as being present by the RAs while non-native transcribers sometimes found it ambiguous. There were two judgements on word-initial sounds in

English code-switches, [z, s] circle and [θ, s] three, in which RAs disagreed with the non-native transcribers. Both non-native transcribers agreed with each other on these sounds.

6.5.2 Word-Medial

Figure 6.3 displays RA and non-native transcriber responses for word-medial sounds arranged by judgement type. The first five pairings along the horizontal axis are stop-fricative distinctions, the sixth is voicing followed by an affricate-stop pairing, and the last two involve place differences in fricatives. Again the green dots indicate the sound perceived by non-native transcribers when they were in agreement with each other, and the purple dots where there was ambiguity or transcribers disagreed with each other.

The first judgement shows there was a non-native transcriber disagreement as to whether the sound was a palatal stop [ɟ] or palato-alveolar fricative [ʒ]. RAs judged the sound as a palatal stop. The next pairing shows agreement that the sound was a stop between RAs and both non-native transcribers on the [ɟ, ð] judgement. For the [d, ð] pairing, non-native transcribers perceived one sound as fricative and the other as stop, with transcribers disagreeing on the stop in the first consonant in ajasaid ‘other side’.

RAs judged these sounds as the fricative [ð] and between the fricative and the mid-point of the scale. For the [ʃ, ɟ] pairing the average RA response was around the mid-point, with some responses at either end, agreeing with non-native transcribers who also judged these sounds in both categories. The [v, b] responses show that all four of the sounds were perceived by non-native transcribers as the stop [b], but non-native CHAPTER 6 193

transcribers disagreed on two of them, gibit ‘give’ and libim ‘leave’. RAs judged them between the stop and the fricative with the mean response at the mid-point.

The voicing pairing [z, s] shows that non-native transcribers agreed that one sound was voiced and the other was voiceless. RAs judged both sounds on the continuous scale between the endpoints. For the affricate-palatal stop pairing, non- native transcribers perceived three as [ɟ] and three as [ʤ], with some disagreement.

RAs judged most of these as closer to the palatal stop and one as the affricate, indicated by the long whisker extending out to [ʤ]. The voiceless place judgement [ʃ, s] shows that non-native transcribers perceived the sound as [ʃ] and had some disagreement. RAs judged it between both sounds but closer to [s] than [ʃ]. The results for [z, ɟ] indicate that non-native transcribers perceived an ambiguous palatal stop for the second consonant in bijinbat ‘fish’, and RAs perceived a palatal stop.

Overall word-medially RAs generally agreed with the judgements of the non- native transcribers. The main differences between RAs and non-native transcribers were in the stop-fricative judgements [d, ð] and [b, v]. These sounds were transcribed by non- native transcribers as stops, though for half of (1 out of 2 [d, ð] and 2 out 4 [b, v]) there were disagreements among non-native transcribers between the stop and fricative. On the continuous scale RAs judged some of these sounds toward the fricative. There was also a difference between the non-native transcribers’ judgements and RAs’ judgements for the place pairing [ʃ, s]. Non-native transcribers perceived the sound as [ʃ], but RAs responded at the [s] endpoint and on the continuum between the two points.

6.5.3 Word-Final

The results for word-final sounds are shown in Figure 6.4. The first four pairings involve stop-fricative judgements and the last two voicing. Sounds in the first judgements, [b, v] in lub ‘love’, were perceived as an ambiguous stop by non-native CHAPTER 6 194

listeners and RAs generally perceived a stop. The sounds in the [c, ʃ] judgements for the final consonants in binij 'that's it!,

The next pairings involved voicing judgements. For sounds in [f, v] judgements non-native transcribers perceived [v], and on the scale RAs made judgements between the voiced and voiceless endpoints with a mean around the midpoint. For the [z, s] pairing RAs judged the sounds between the two endpoints but generally closer to the voiceless fricative [s]. The whisker extending out to [z] corresponds with the sound that non-native transcribers also judged as [z].

Overall for the word-final sounds there was generally agreement between non- native transcribers and RAs. The main discrepancy was for [f, v], which involved an

English code-switch five and a borrowing giraffe. While the final sound in these words was perceived by non-native transcribers as the voiced fricative [v], there was inter- transcriber disagreement for one token, five, which was an English code-switch, and

RAs responded along the full scale.

CHAPTER 6 198

6.6 Discussion

In this study native speakers of Gurindji Kriol used visual analogue scales to make judgements of Gurindji Kriol sounds that had been transcribed phonetically as fricatives or were ambiguous to transcribers. The phonetic transcriptions of Gurindji

Kriol were made by native Australian English speakers, one of them with direct experience with Gurindji Kriol. To investigate the relationships between perceptions of native speakers and nonnative transcribers on Gurindji Kriol segments, native speaker judgements were then compared with the transcriber judgements in initial, medial and final word positions. The transcribers are referred to here as ‘non-native transcribers’, as they were not native speakers of Gurindji Kriol.

The aim was twofold: 1) to investigate how visual analogue scales can be used in the field to elicit native speaker perceptions of sounds in a phonologically or phonetically underdocumented language, and 2) to further our understanding of fricative variation in Gurindji Kriol by analysing native speaker judgements across a range of types of judgements and word positions. We found that the methodology used here was successful at eliciting judgements of sounds from native speakers who had no technical

IPA training, with some limitations that could be addressed in future studies. From the native speaker responses we gained further insight into transcription ambiguities. Word- initially, ambiguities involving stop-fricative distinctions tended to be judged as stops by native speakers, and ambiguities between palatal stops and palato-alveolar affricates were in general judged as palatal stops by native speakers. Word-medial stop-fricative ambiguities were often judged toward fricatives by native speakers when transcribers had judged them as stops. In all word positions voicing judgements were often judged along the continuous scale by native speakers. Results also indicated agreement CHAPTER 6 199

between native speakers and non-native transcribers for many judgements, particularly those in medial and final word positions.

Visual analogue scales have been used previously to measure adult perceptions of covert contrast, accuracy, and ambiguous sounds in child speech (Julien & Munson,

2012; Munson et al., 2010; Schellinger, 2008; Uberg-Carlson et al., 2009), judgements of disordered speech (Kempster et al., 2009), and to investigate perceptual sensitivity to within-category phonetic detail in synthetic speech (Kong & Edwards, 2011;

Skorniakova & Ito, 2011). The previous research suggested visual analogue scales would be beneficial in speech-language pathology contexts for measuring children’s productions and in analysing speech of multilingual children (Edwards & Munson,

2012; Munson, Schellinger, Urberg-Carlson, 2012). The current study indicates visual analogue scales can additionally be used in an adult L2 context for the purpose of checking phonetic transcription of non-native transcribers (i.e. not native speakers of the language being transcribed) with judgements of native speakers who have not been trained in the technical aspects of transcription. The visual analogue scales enabled us to elicit and analyse the implicit linguistic knowledge of native speakers, which provided information about discrepancies and ambiguities in the phonetic transcriptions made by non-native transcribers.

The current method may be particularly useful in contexts such as phonetically underdocumented languages, where it is often impractical to train native speakers to the technical level required for IPA transcription. Further, in a language like Gurindji Kriol the sound system is highly variable and the sound categories of native speakers are unknown. Using continuous scales for sound judgements therefore provided additional information for interpreting the categorical IPA transcriber judgements, particularly ambiguous transcriptions and inter-transcriber discrepancies. CHAPTER 6 200

The two major advantages of the current design were its ease of use with the native speaker RAs and that the visual analogue scales allowed for both continuous and categorical judgements. That is, RAs could respond along the scale for a continuous judgement or at the endpoints for categorical judgements. There were also some limitations, however. In particular the endpoints on the scales and the time demands required by this design should be carefully considered for future research. Here we will discuss the study results and specific information that native speaker scale responses provided about phonetic transcription and fricative variation in Gurindji Kriol. We then consider the limitations of the current design, particularly the time demands in relation to the extra information that was gained from carrying out this study.

6.6.1 Effects of Language Background

RAs were native speakers of Gurindji Kriol and were also bilingual in English.

Phonetic transcribers were native Australian English speakers who were trained in phonetic transcription and had varying levels of experience with Gurindji Kriol – one transcriber was not familiar with Gurindji Kriol prior to this study, the other (the first author) had some experience listening to the language gained throughout the research.

Language background may have affected the results differentially for RAs and transcribers, and different levels of familiarity with Gurindji Kriol could explain some of the inter-transcriber discrepancies.

Language background may have directly affected the results if perceptual judgements of Gurindji Kriol sounds were influenced by the listener’s L1 phonological categories (Coussé et al., 2004; Hardison, 2003; Strange & Jenkins, 1978). Effects would be seen in native speaker and transcriber judgements of sounds that occur in one language and not in the other. The sounds in the current study that occur in Gurindji

Kriol and are not present in Australian English are the voiced and voiceless palatal stops, CHAPTER 6 201

[ɟ ,c]. These sounds occur in traditional Gurindji words, and in Gurindji Kriol are also pronounced in some Kriol-derived words in variation with fricatives, e.g. the initial sound in jinek ‘snake’ is sometimes pronounced with a palatal stop [ɟ] and sometimes a voiceless alveolar fricative [s]. When palatal stops are pronounced with frication they sound very similar to the palato-alveolar affricates [ʧ, ʤ], and there were several inter- transcriber discrepancies between palatal stops and affricates.

The results show that in initial and medial positions there were differences between RAs and transcribers in judgements between palatal stops and affricates (there were no palatal stop-affricate judgements for word-final sounds). In initial position there were six sounds transcribed as the affricates [ʧ] or [ʤ], and the mean RA response was a categorical palatal stop [c] or [ɟ]. There was one mid-point RA response for a [ɟ, ʤ] judgement suggesting the Gurindji Kriol sound was perceived as in- between the two endpoints, and this was in a token of jeya ‘there’ that transcribers had disagreed on. Word-medially there were three [ʤ, ɟ] sounds transcribed as the palatal stop and three as the affricate. Again the mean RA response was closer to the palatal stop endpoint, although more of the scale was used than for the word-initial judgements of this type. Overall these findings suggest it may be difficult to distinguish between palatal stops and palato-alveolar affricates and that listeners tend to judge them in terms of their native language phonology. Future research could involve acoustic analyses of controlled recordings to investigate the acoustic variability of these sounds.

Some of the native speaker judgements were unexpected because they differed from non-native transcriber judgements, with the RA responses tending toward the phonology of English, in which they are bilingual. For example initial and medial interdental fricatives [θ, ð] are present in English but do not occur in Gurindji and are extremely rare in Gurindji Kriol. RAs judged an initial sound as a voiceless dental CHAPTER 6 202

fricative [θ] rather than the alveolar stop [t], and two medial sounds toward the voiced dental fricative [ð] rather than the alveolar stop [d]. In these cases language background may have had indirect effects, as RAs and transcribers may have been differentially influenced by English spelling interference and lexical expectation effects. The phonological awareness training and visual analogue scale procedure were designed so that RAs would not rely on either English or Gurindji Kriol spelling to complete the task. However, the words involved in these judgements (e.g. smuth-wan ‘smooth’) are phonologically similar to the corresponding words in English, in which the RAs are bilingual. Their knowledge of English spelling may have interfered with perceptual judgements of the more difficult sounds such as dental fricatives.

The word-initial [h] judgements also appear to have been affected by language background. Non-native transcribers had judged half the [h] words as being pronounced with an initial [h] and the other half as having no pronounced [h], while native speakers judged all the words as being pronounced with an initial [h]. Word-initial [h] appears to occur variably in Gurindji Kriol and may be a form of hypercorrection when it occurs in vowel-initial Kriol-derived words, e.g. [hæpʊl] apple. The current results may be a reflection of the hypercorrection of initial [h] in production. It is also possible that [h] is a covert contrast in Gurindji Kriol, that is, a phonetic contrast that is only apparent at a subphonemic level. If there were a covert contrast with [h] in Gurindji Kriol this would mean it may be more readily perceived by native speakers than non-native speakers.

While it is difficult to analyse [h] reliably in acoustic analyses, further controlled perceptual studies would be useful to examine the variable occurrence of [h] in Gurindji

Kriol.

Non-native speaker transcribers and native speaker RAs may have both been influenced by lexical expectation effects. For example for the initial sound in the CHAPTER 6 203

English code-switch word circle transcribers agreed on the voiceless alveolar fricative

[s], as that is the expected pronunciation in English. RAs however judged this sound toward the voiced alveolar fricative endpoint [z]. On the other hand there were words like giraffe, another English code-switch, in which transcribers judged the final consonant as the voiced fricative [v] and RAs judged the voiceless fricative [f], which is the expected sound from the spelling and usual English pronunciation. To elucidate these effects further research and acoustic analyses of controlled recordings would be beneficial; however these results indicate a starting point for the types of perceptual differences that may occur.

A final effect of language background may explain some of the inter-transcriber disagreements. While both transcribers were trained in phonetic transcription, one had no prior experience with Gurindji Kriol and the other had some familiarity with the language and had learned some of it directly from native speakers. As Oller and Eilers

(1975) found, language background effects are not necessarily negative and familiarity with the language being transcribed can improve the accuracy of transcription. Knowing the ‘target’ meaning of words and utterances can help direct transcriber attention toward the relevant phonetic detail (Oller & Eilers, 1975). Half the tokens in the current study were chosen for the visual analogue scale judgements because there was transcriber discrepancy or the sounds were ambiguous, and so the proportions of disagreement in the current data are higher than in the full dataset. At least some of these disagreements are likely due to familiarity with Gurindji Kriol, which changes even throughout the transcription process as transcribers listen to more of the speech data.

In sum, the differing language backgrounds of both native speakers and transcribers may explain some of the judgement differences in the current data.

Disagreements between the two transcribers may be due to their different levels of CHAPTER 6 204

exposure to Gurindji Kriol conversational speech. Effects of L1 phonological categories are shown in the different judgements of native and non-native speakers between palatal stops and palato-alveolar affricates. Interference from English spelling and lexical expectation effects may account for some differences between RA and transcriber judgements where RAs unexpectedly judged sounds as ‘toward’ endpoints that were rarer sounds in Gurindji Kriol.

6.6.2 Word Position Effects

Results showed differences by word position in agreement between RAs and transcribers, and in inter-RA agreement. In particular word-initial responses appeared to differ overall from responses to medial and final sounds. In general, RA judgements of initial sounds showed more inter-RA agreement but also more discrepancies between

RAs and transcribers than medial and final word positions. In the phonological awareness training, RAs found the initial sounds easier to isolate and identify than the medial and final sounds, which is consistent with previous research as in early literacy development, initial sounds are easier to segment and reflect on (Gillon, 2004).

Additionally, the consonants in the endpoints were in onset positions (CV), which may have made for an easier comparison with the word-initial judgements as these were also mostly in onset CV contexts.

For word-medial sounds, the mean RA responses were often along the scale (i.e. not categorical judgements at the endpoints) for stop-fricative judgements such as [d, ð],

[ʃ, ɟ] and [v, b]. This may be due to lenition occurring in medial position, particularly as most of these sounds were in intervocalic environments. This context is susceptible to lenition, as greater articulatory effort is arguably required to produce a constriction relative to other contexts (Kirchner, 2004). Future research using acoustic analyses and more controlled perceptual testing would be useful to determine whether these sounds, CHAPTER 6 205

or some of them, are produced along a continuum, that is stops may be produced with varying amounts of frication until they are indistinguishable from fricatives. Further studies using acoustic analyses would have to involve controlled recordings of speech production, which may affect stylistic influences on pronunciation by eliciting more formal speech. The current data was from naturalistic conversational recordings, thus transcription and other perceptual judgements are more reliable than acoustic analyses due to background noise, overlapping speech, low recording levels and clipping.

Overall, the current study using the continuous visual analogue scales with native speakers of Gurindji Kriol helps to explain some of the ambiguities and inter- transcriber discrepancies in the phonetic transcription that are likely due to differences in language background. Some sounds, such as fricative voicing and word-medial stop- fricative continuancy, may vary along a continuum while others such as the presence of word-initial [h] may indicate possible hypercorrection or covert contrasts in Gurindji

Kriol. Further research involving acoustic analyses of controlled recordings that also control for phonetic environments and perceptual experiments are needed to investigate the relative contributions of the possible effects suggested here.

It is also important to note that the RA judgements, and thus our interpretations of them, are restricted by the endpoints provided in the visual analogue scales. A major advantage of the scales is that they allowed for continuous judgements; however, this was a trade-off in that there were only two categorical endpoints and these were decided by a transcriber. RAs did have the option of responding with ‘other’ if they thought the

Gurindji Kriol sound was something different to the endpoints provided, and sometimes they did use that response. The endpoints may still have biased RAs toward making an assumption that the Gurindji Kriol sound ‘belonged’ to one of the two endpoints. The endpoints may have been problematic in other respects as they required certain CHAPTER 6 206

decisions to be made in advance by the researchers. These are discussed in further detail in the following section.

6.6.3 Limitations

There are two main limitations to the current design and, by way of recommendation for future research, it is important to make decisions about these early on. One relates to the endpoints used in the scales – what to use as endpoints, how listeners may represent them in relation to the sounds being rated, and how they may affect the results. The second important consideration is the limited number of RAs in the current study and that not all of them undertook the phonological awareness training.

We then consider whether the quality of information gained from the visual analogue scale task is advantageous enough to add another time-consuming process to phonetic transcription.

The endpoints used in the current study were IPA segments in CV contexts from the UCLA phonetics website and were spoken by an American male speaker trained in phonetics. In the scale tasks the native speakers heard the consonants of interest in CV endpoints, and the Gurindji Kriol sound in the context of the whole utterance. This allowed for comparison with the transcriber IPA judgements, as transcribers also judged the sounds in the context of the whole word and utterance. Native speakers and transcribers were therefore all exposed to linguistic context and sentence cues. The rationale for using the UCLA IPA segments as endpoints was that in Gurindji Kriol fricatives are highly variable (Buchan et al., 2012) and the native speaker target forms are unknown. To use whole words or any kind of Gurindji Kriol categories as endpoints would raise a new set of issues around how speakers represent those categories and may not capture the full variability of the segments being judged. CHAPTER 6 207

The UCLA segments were used because they are clear typical examples of IPA categories produced by a trained phonetician. The consonants of interest were in CV contexts in the UCLA segments, which may have limited the study as the endpoints did not indicate the actual range of the Gurindji Kriol sounds that were judged. Rather, the endpoints represented approximations to which RAs could respond. The phonological awareness training was designed to address some of the issues with the endpoints by focusing on sound segmentation and isolating sounds in different word positions. One aim of the training was to develop RAs’ phonological awareness skills so that they could compare the consonants in the endpoints to Gurindji Kriol sounds in initial, medial and final word positions in the tokens. Knowing more about the variation in sounds and how they are typically pronounced in the language being studied would allow for a closer approximation of the sounds used in the endpoints of a visual analogue scale study.

A further limitation of the current study is that a relatively small number of RAs participated, and not all of them completed the phonological awareness training that was conducted in a previous visit. The native speakers who underwent the training assisted those who did not do it, but these differences in the phonological awareness training may have contributed to some of the inter-RA disagreements in judgements. Future research using the current method would benefit from recruiting more RAs and retaining the same people throughout the study, that is for both the training and the administration of the visual analogue scales.

6.6.4 Conclusion

Visual analogue scales are useful tools for eliciting linguistic knowledge about phonology from native speakers who do not have technical training in phonetic transcription. Comparing native speaker judgements on fricatives in Gurindji Kriol with CHAPTER 6 208

the categorical IPA judgements made by non-native speaker phonetic transcribers provided insight into inter-transcriber disagreements and alerted us to sounds that may vary over a continuum rather than being categorical, such as voicing, and possible covert contrasts.

Methodologically, using visual analogue scales with native speakers of Gurindji

Kriol in the field was generally successful. Native speakers have implicit linguistic knowledge of their language, and the visual analogue scales allowed us to elicit their knowledge of the sound system to compare with non-native transcriber IPA judgements.

Results suggested that ambiguities and discrepancies in phonetic transcriptions may not always be due to random error but may be more systematic, with factors such as language background and phonological variation in the language being transcribed explaining some transcriber disagreements.

CHAPTER SEVEN

Conclusion

CHAPTER 7 210

Phonetic variation in the speech to which children are exposed plays an important role in children’s language development. Children’s processing of sound structures in the input forms the basis of much of their linguistic knowledge. According to exemplar-based models of phonological development, acquisition of the native language phonology is an ongoing process and the input continually influences the organisation of phonological representations in memory (Hawkins, 2003;

Pierrehumbert, 2003). Across languages the input contains phonetic variability associated with linguistic and socioindexical information. There is a large literature on phonological variation in speech input to infants (e.g. Cristià, 2010, 2011; Davis &

Lindblom, 2001; Fernald et al., 1989; Grieser & Kuhl, 1988; Katz, Cohn & Moore,

1996; Kirchhoff & Schimmel, 2005; Kuhl et al., 1997; Snow, 1977; Stern, Spieker,

Barnett & MacKain, 1983). The phonology of the input continues to change after the first year of life as child-directed speech develops into adult-directed speech. There are, however, relatively few longitudinal studies of how and when this happens.

The purpose of this thesis was twofold. The main aim was to investigate how phonetic variation in fricatives in mothers’ speech to children changes as children age in two language varieties in northern Australia – Gurindji Kriol and Australian English. A subsidiary aim was to investigate methodological issues with IPA transcription of a phonologically underdescribed language, which involved eliciting perceptual judgements from native speakers of Gurindji Kriol to compare with phonetic transcriptions. Gurindji Kriol is a mixed language spoken at Kalkaringi, an Aboriginal community in northern Australia. The phonology is highly variable, particularly in fricatives in Kriol-derived words. The Australian English corpus collected and analysed for this thesis was from Katherine, a regional town in the same geographic region as

Kalkaringi. In both language varieties we analysed maternal speech longitudinally as CHAPTER 7 211

children aged from approximately 1;6 to 2;0 to 2;6. This is considered an important time in children’s language acquisition as their knowledge of phonological processes in the native language continues to develop and their receptive and productive vocabularies expand (e.g. Nelson, 1973; Swingley, 2007).

This thesis consisted of three empirical research studies to address three main research questions: 1) Study 1, Chapter 4 – What is the nature and extent of variation in fricative production in Gurindji Kriol maternal speech? 2) Study 2, Chapter 5 – How does phonetic reduction in fricatives change as children age in northern Australian

English maternal speech? 3) Study 3, Chapter 6 – How can we elicit native speaker perceptions of fricative variation in Gurindji Kriol using Visual Analogue Scales

(VAS), and what information does this add to non-native transcribers’ phonetic transcriptions? In this chapter I discuss the main findings and conclusions across all three studies, the implications and limitations of this research, and directions for future research.

7.1 Core Findings

The overall results of the studies in this thesis show that children are exposed to considerable phonetic variation at a segmental level in fricatives in both language varieties, and that variation changes in maternal speech over child ages 1;6, 2;0 and 2;6.

Fricative variation is complex and different types were found in each language variety, likely associated with different kinds of linguistic and socioindexical information.

In Gurindji Kriol maternal speech we found that although fricatives in Kriol- derived words were highly variable, at least some of the variability was systematic.

Manner of articulation variation between stops and fricatives was more likely to occur in word-initial position (rather than medial or final), at labio-dental and alveolar places of articulation, and in lexical items that were open- rather than closed-class. In all CHAPTER 7 212

tokens that could potentially be pronounced with a fricative (defined as words whose corresponding English cognates contain fricatives), in word-initial position fricatives were significantly more likely in mothers’ speech when children were approximately

2;6 than when they were 1;6. In tokens of words that were found to contain stop- fricative variation, word-initial fricatives (with the exception of /h/, which did not occur frequently enough to be analysed) were again more likely at 2;6 than 1;6. In these tokens fricatives were also significantly more likely in medial position at child age 2;0 and 2;6 than at 1;6. These results took into account both phonological environment and interspeaker variation. Mothers used proportionately more fricatives as children got older partly due to lexical choice, as mothers also proportionately increased their use of words that could be pronounced with fricatives as children aged. Findings can be interpreted in terms of fine-tuning, the concept that mothers make adjustments to their speech differentially as children age in response to children’s own receptive and productive language development. The proportionate increase in fricatives over time may serve to gradually introduce children to the variable phonetic forms that are inherent in adult speech in the community.

In northern Australian English a different type of variability in fricatives in maternal speech was investigated. We analysed word-initial /h/ and word-final /v/ deletion as examples of phonological reduction due to connected speech processes that are frequent in casual speech. It was found that deletion occurred in a stable set of lexical items that were mostly function words. In mothers’ speech deletion proportionately increased between child ages 1;6 and 2;0, and decreased between 2;0 and 2;6. The effect was significant for /h/ deletion and approached significance for /v/, which had fewer occurrences in the corpus. Further analysis showed that the effect of child age on deletion was not likely to be caused by changes in speech rate. The non- CHAPTER 7 213

linear change over time was unexpected, as the few previous studies that have examined sociophonetic variation in speech input to children after infancy suggested a gradual linear change from child-directed to adult-directed speech style (Foulkes, Docherty &

Watt, 2005; Smith, Durham & Fortune, 2009). Possible explanations for the Australian

English maternal speech findings also relate to fine-tuning, as mothers may have modelled speech to promote children’s language acquisition.

There are several mechanisms that could lead to a non-linear change in deletion, in particular changes in the distributions of speech acts and types of social interactions, which are associated with different speech styles along a continuum of casual to careful speech. For example, Smith, Durham and Fortune (2007) found that in the informal speech contexts play and routine, child-directed speech was more likely to contain a non-standard phonetic variant than in the more formal contexts of teaching and discipline. Speech acts at different points on a careful-casual or formal-informal continuum, such as questions, recasts and repetitions, may become more frequent in mothers’ speech as children start consistently producing more words. Relative frequencies of different types of interactions between mothers and children, e.g. play, explicit teaching, behaviour management, may also shift over time with children’s social and cognitive development. Further research is needed to examine whether there are stylistic changes over time in maternal speech and how it relates to changes in phonetic variability. As with Gurindji Kriol, in northern Australian English one function of changing phonetic variability in input to children over time may be to introduce them to the variation that is characteristic of the adult speech community.

This thesis also considered phonetic transcription methodology. In Study 3

(Chapter 6) we elicited native speaker judgements of Gurindji Kriol segments using

VAS and compared these to IPA transcriptions made by non-native transcribers. This CHAPTER 7 214

methodology provided further insight into variation in Gurindji Kriol, and elucidated possible reasons behind ambiguities and inter-transcriber discrepancies in the transcriptions. Language background had an effect as native speakers and non-native transcribers tended to judge ambiguous sounds, such as palatal stops/palato-alveolar affricates, in terms of their native language phonology. Differences between native speaker and non-native transcriber may have been due to several effects on perception of Gurindji Kriol phones, including lexical expectation, orthographic influence and perceptual hypercorrection. Findings also showed agreement between native and non- native speakers on many segments, particularly those in medial and final word positions.

7.2 Implications

The findings of this thesis show that fricatives in speech input to children are variable and change over time in Gurindji Kriol and northern Australian English. More generally, the findings indicate that theoretical models of phonological acquisition need to account for a variable and dynamic input. As discussed in the theoretical framework in section 2.4, exemplar-based theories are well positioned to do this as they propose mechanisms for how children process and store variable phonetic detail and associated information in long-term memory (Beckman, Munson & Edwards, 2007; Hawkins,

2003; Pierrehumbert, 2003). The current studies were not designed to test the various theoretical models, rather exemplar-based models form the theoretical rationale for describing variation in the input. Exemplar models account for phonetic variation in the input including phonetic variation related to fine-tuning, which was explored in Studies

1 and 2. According to exemplar theories perceptual events contribute to the formation of categories in memory, and phonetic variants in the input influence the distributions of exemplars from which categories form. Fine-tuning in the form of differential CHAPTER 7 215

phonological modifications in the input would therefore also contribute to fine-tuning of exemplar distributions in children’s long-term memories.

This research has practical implications for language professionals in northern

Australia, particularly for teachers at the school in Kalkaringi and neighbouring communities where similar language varieties are spoken. Although Gurindji Kriol is the home language of children in Kalkaringi, English is the language of the school, and children are taught in English when they start going to school. Prior to the larger project that this research is a part of there were no detailed systematic descriptions of the phonology of Gurindji Kriol and thus there was little information for teachers to use as a basis for teaching English language and literacy. Information about sounds in Gurindji

Kriol and the similarities and differences to Australian English could help guide teachers’ explicit teaching of sounds and words in English. In particular the current findings suggest that Gurindji children may benefit from explicit teaching of pronunciation in English as an L2, particularly in manner of articulation distinctions in

English between fricatives and stops in different word positions. For example the results of Study 1 suggest that children are exposed to fricatives in word-initial position, with the majority being labio-dental and alveolar in open-class words. English also contains word-initial labio-dental dental and alveolar fricatives, but in contrast to Gurindji Kriol

English also has a relatively high frequency of interdental word-initial fricatives, largely due to their presence in English closed-class words (the, this, that, then, there, etc.). The corresponding Kriol-derived words in Gurindji Kriol were nearly all pronounced with word-initial stops, and children may need special instruction to help them perceive and produce these sounds in English.

Children are also exposed to word-medial fricatives in Gurindji Kriol. Some medial fricatives in Kriol-derived words alternate frequently with stops, which means CHAPTER 7 216

words may sometimes sound quite similar to the corresponding English word and sometimes different, depending on whether it is pronounced with a fricative or stop (e.g. the medial consonant in neba ‘never’). If teachers hear children producing fricatives they may assume that the children have a command of English phonology, without necessarily realising that they are producing Gurindji Kriol phonology where stops and fricatives are variable in some phonological and lexical contexts.

The rarest position for fricatives in Kriol-derived words was word-final. Again,

Gurindji children may require explicit instruction in word-final fricatives in English.

Many word-final fricatives in English are morphological /s/ denoting plurals, possession, and present tense. Therefore in addition to phonological instruction children may also need further teaching of English morphology as part of learning word-final fricatives. The results from the Visual Analogue Scales with adults in Study 3 suggested that it may also be difficult to distinguish fricative voicing in word-final position.

Further research is needed to examine voicing in Gurindji Kriol fricatives; however, the present research does suggest variability in voicing. This is important when it comes to teaching English, which has fricative voicing contrasts.

Overall, the findings of these studies have practical implications for teachers at the school in Kalkaringi and neighbouring areas. They can also be used in the same way for teaching traditional Gurindji, for example starting with the sounds that children have already learned from Gurindji Kriol input and then later teaching the words that sound relatively different. The different distributions of fricatives in each word position have pedagogical implications, as it is important for teachers to know that learners will likely be more familiar with some fricatives in particular positions. For example, relative to other fricatives, word-initial [s] was common in Kriol-derived words in maternal speech. Being familiar with the initial [s] sound from their home language may make it CHAPTER 7 217

easier to learn [s] in medial and final positions in English words if teachers can make students aware that they are the same sounds.

In the feedback trip to Katherine and Kalkaringi conducted at the end of the project, we presented these findings to teachers at Kalkaringi school. This involved providing information and sound charts to teachers about Gurindji Kriol phonology described in the current studies and studies in the larger project. Information included brief descriptions of Gurindji Kriol fricatives and stop-fricative variation from the studies in this thesis, as well as stop consonant voicing variation (Jones & Meakins,

2012a), the vowel system (Jones, Meakins & Buchan, 2011; Jones, Meakins &

Muawiyath, 2012), and baby-talk phonology (Jones & Meakins, 2012b). The focus of the information provided to teachers was on similarities and differences between sounds in Gurindji Kriol and Australian English, with the aim of developing teachers’ knowledge about Gurindji mothers’ speech to children, sounds that children may have difficulty with, and sounds that they are likely to be familiar with.

7.3 Limitations

As the limitations of each study have been identified in the discussion sections in Chapters 4, 5 and 6, the wider limitations of the thesis as a whole will now be discussed. First, both the Katherine English corpus and the subsample of the Gurindji

Kriol corpus analysed for these studies consisted of a small number of speakers (four

Katherine English speaking mothers, three Gurindji Kriol speaking mothers). Small speaker numbers is a common issue in many child speech studies involving corpus analyses, due to the large demands on time and resources that it takes to create a corpus.

A relatively small number of speakers was built into the current design as it allowed for in-depth analysis of naturalistic data, that is, analysis of occurrences of segments and processes in naturalistic conversational speech; but it does limit the generalisability of CHAPTER 7 218

results and further research is needed to determine how widespread the current findings are.

A second methodological issue to consider when interpreting these findings is that the data were not specifically child-directed but included all speech spoken by mothers in the presence of the target children. This was again by design, based on previous conceptualisations of input to children in the literature (see Chapter 2 pp. 19-

21). There may, however, be differences between speech directed specifically to children and other speech in their presence that would not be revealed by the analyses in this thesis.

Another broader limitation inherent in the design of this study is that the data were from phonetic transcription, which is primarily a perceptual activity and may be subject to perceptual biases. Potential problems with phonetic transcription data are discussed extensively in the background section of Chapter 6. The disadvantages of phonetic transcription as a data analysis tool formed part of the rationale for Study 3

(Chapter 6), where we addressed reliability issues by comparing phonetically transcribed segments with native speaker judgments. As well as checking the phonetic transcriptions with native speaker judgements in Study 3, to address potential biases reliability checks were also undertaken for the transcription data analysed in Studies 1 and 2. Future research could involve acoustic analyses of Gurindji Kriol speech recorded in a controlled and quiet environment, which may reveal finer-grained variation than transcription based analyses will allow. Acoustic analyses of fricatives was not feasible for the data in these studies, as they are from naturalistic recordings that contain varying amounts of background noise and were not controlled for context.

Also, the Gurindji Kriol conversational data were originally recorded for the purpose of morphological analyses rather than detailed acoustic processing and some audio CHAPTER 7 219

recordings of Gurindji Kriol were made on mini-discs, which compress audio data.

Phonetic transcription in conjunction with reliability analyses were therefore more appropriate methods for analysing fricatives in the current data than acoustic analyses.

7.4 Directions for Future Research

The empirical research in this thesis was largely exploratory, as to our knowledge it is the first quantitative description of fricatives in an Australian contact language, and the Katherine English corpus is the first phonetically transcribed longitudinal corpus of mother-child interactions in Australian English for toddlers. The findings of this research therefore open up several broad avenues for future research.

The present studies examined fricatives in a defined set of Kriol-derived words.

Investigating fricatives more broadly and using acoustic analyses and perceptual testing could be done to investigate whether or not there are any fricative contrasts in Gurindji

Kriol, as well as to examine the relative contributions of specific processes (e.g. lenition, socioindexical and stylistic factors) to fricative variation. These findings can also be used as a basis for researching fricatives in Kriol and other Australian contact language varieties.

In both Gurindji Kriol and northern Australian English it would be interesting to analyse children’s receptive and productive language longitudinally over the preschool years. Such research would provide information about the developmental trajectories of children’s acquisition of phonetic variation in the input, and the specific effects that change over time in maternal speech has on children’s language acquisition. We have longitudinal data of the Australian English speaking children’s productions as the target children in the Katherine English corpus did wear microphones for most of the sessions.

Future research on this data may provide further insight on the relationships between input and children’s language production. CHAPTER 7 220

Another question for future research on both language varieties is how much variation is related to speech styles and the types of interactions? We suggested changes in these factors as a possible contributor to the effect of child age on phonological reduction in maternal Katherine English (Chapter 5). Further study is needed to investigate possible effects of situational factors on phonological variation in mothers’ speech and their influence on children’s language.

Also raised in the Katherine English study was a question about sociolinguistic markers in regional Australian English, or linguistic variables that carry information about social groups and/or speech styles. Research to identify these variables would require a sociolinguistic investigation of Katherine English and other varieties of

Australian English spoken in different regions and social groups. Studies of change over time in identified sociophonetic markers in varieties of Australian English maternal speech would allow for comparisons with previous studies in other English varieties (in particular Foulkes, Docherty and Watt’s (2005) study of Tyneside English child- directed speech), and further insight into children’s acquisition of phonetic variation related to socioindexical factors.

The methodological study in Chapter 6 also highlights areas for further research.

Findings indicated that Visual Analogue Scales could be used successfully in the field to elicit perceptual judgements on phonetic segments from adult speakers with no technical training in transcription. More generally it suggests that visual analogue scales can be used in adult L2 contexts to compare the effects of different language backgrounds on perceptual judgements of sounds in different word positions. Future studies could investigate the application of this to other contexts. For example in adult

L2 language learning visual analogue scales could be used to compare learner and native speaker judgements for the purpose of providing specific feedback to L2 learners CHAPTER 7 221

about their perceptual knowledge of native-like sounds in specific phonological contexts.

The pedagogical implications of the research also open avenues for further research. The larger study included analyses of phonetic variation at acoustic and segmental levels of vowels and consonants, which applied to a teaching context is a lot of information to get through in a classroom. It would be beneficial for teachers to know which sounds to prioritise and the contrasts that are important for intelligibility and comprehension. The relative importance of phonological features in understanding an

L2 learner’s speech is referred to as functional load. There is no single standard measure of functional load, but it typically involves a combination of factors such as frequencies of phonemes, distributions of phonemes within syllable structures, the number of minimal pairs and their lexical frequency, and the frequency of particular phones within minimal pair words (Brown, 1988; King, 1967; Munro & Derwing, 2006; Surendran &

Niyogi, 2006). Phones that have high frequencies and distinguish a number of minimal pairs have high functional load (such as many vowels in English), while low frequency phones that distinguish relatively few pairs (such as [ʃ, ʒ] in English) have low functional load. Speech errors involving high functional load sounds affect comprehensibility and intelligibility of L2 speakers to a greater extent than errors involving low functional load sounds (Munro & Derwing, 2006). Empirical information about functional load would therefore be useful to teachers, who could use it to prioritise teaching high functional load sound contrasts.

For functional load information to be useful to teachers in Kalkaringi and surrounding areas, further research needs to be done on children’s productions. This thesis is an analysis of the sounds children are likely to be familiar with through exposure to speech in their home language. Future studies of children’s speech would CHAPTER 7 222

provide information about the sounds they have difficulty producing, which could then be used by teachers in conjunction with functional loads to determine the relative importance of the difficult sounds (Brown, 1988). Further questions raised in the

Katherine English study (Chapter 5) and the Visual Analogues Scales study (Chapter 6) could also be studied from a functional load approach. For example, are phonological reduction and other casual speech processes more likely to occur in low than high functional load segments? What are the relationships between functional load, phonetic variation and perception of Gurindji Kriol sounds?

7.5 Conclusion

In conclusion, this thesis has provided analyses of fricative variation in speech that children are exposed to in two language varieties in northern Australia, Gurindji

Kriol and northern Australian English. Further, variation changes in mothers’ speech as children age between approximately 1;6, 2;0 and 2;6 in some processes and phonological environments. This research highlights the value of longitudinal research into the language input to children after infancy, and of systematic descriptions of variation in contact language varieties. Results indicate the need for models of language acquisition to account for a dynamic phonology in the input. REFERENCES 223

References

Akhtar, N. (2005). The robustness of learning through overhearing. Developmental

Science, 8(2), 199-209.

Anderson, J. L., Morgan, J. L., & White, K. S. (2003). A statistical basis for speech sound

discrimination. Language and Speech, 46(2-3), 155-182.

Aslin, R. N., Saffran, J. R., & Newport, E. L. (1998). Computation of conditional

probability statistics by 8-month-old infants. Psychological Science, 9(4), 321-324.

Au, T. K., Knightly, L. M., Jun, S. A., & Oh, S. (2002). Overhearing a language during

childhood. American Psychological Society, 13(3), 238-243.

Ball, P., Giles, H., & Hewstone, M. (1985). Interpersonal accommodation and situational

construals: An integrative formalisation. In H. Giles & R. N. S. Clair (Eds.), Recent

advances in language, communication and social psychology (pp. 263-286).

London: Erlbaum.

Bard, E. G., & Anderson, A. (1983). The unintelligibility of speech to children. Journal of

Child Language, 10(2), 265-292.

Bauer, L. (2008). Lenition revisited. Journal of Linguistics, 44(3), 605-624.

Becker, C. A. (1980). Semantic context effects in visual word recognition: An analysis of

semantic strategies. Memory and Cognition, 8(6), 493-512.

Beckman, M. E., Munson, B., & Edwards, J. (2007). Vocabulary growth and the

developmental expansion of types of phonological knowledge. In J. S. Cole & J. I.

Hualde (Eds.), Papers in laboratory phonology (Vol. 9, pp. 241-264). New York:

Mouton de Gruyter.

Behrmann, M., & Bub, D. (1992). Surface dyslexia and dysgraphia: Dual routes, single REFERENCES 224

lexicon. Cognitive Neuropsychology, 9(3), 209-251.

Bell, A. (1984). Language style as audience design. Language in Society, 13(2), 145-204.

Bent, T., Bradlow, A., & Smith, B. L. (2007). Segmental errors in different word positions

and their effects on intelligibility of non-native speech: All's well that begins well.

In M. Munro & O.-S. Bohn (Eds.), Language experience in second language

speech learning (pp. 331-347). Amsterdam: John Benjamins.

Bernstein-Ratner, N. (1984a). Patterns of vowel modification in mother-child speech.

Journal of Phonetics, 11(3), 557-578.

Bernstein-Ratner, N. (1984b). Phonological rule usage in mother-child speech. Journal of

Phonetics, 12(3), 245-254.

Browman, C. P., & Goldstein, L. M. (1987). Tiers in articulatory phonology, with some

implications for casual speech. In J. Kingston & M. E. Beckman (Eds.), Papers in

laboratory phonology I: Between the grammar and physics of speech (pp. 341-

376). Cambridge: Cambridge University Press.

Brown, A. (1988). Functional load and the teaching of pronunciation. TESOL Quarterly,

22(4), 593-606.

Bryant, G. A., & Clark Barrett, H. (2007). Recognizing intentions in infant-directed

speech: Evidence for universals. Psychological Science, 18(8), 746-751.

Buchan, H. (2011). Phon: Free software for phonological transcription and analysis.

Language Documentation and Conservation, 5, 81-87.

Buchan, H., & Jones, C. (2013). Phonological reduction in maternal speech in northern

Australian English: Change over time. Journal of Child Language, doi:

10.1017/S0305000913000123 REFERENCES 225

Buchan, H., Jones, C., & Meakins, F. (2012). Fricative variation in maternal Gurindji

Kriol. Manuscript in preparation.

Butcher, A. (2006). Australian Aboriginal Languages: Consonant-salient phonologies and

the 'place-of-articulation imperative'. In J. M. Harrington & M. Tabain (Eds.),

Speech production: models, phonetic processes and techniques (pp. 187-210). New

York: Psychology Press.

Chevrot, J. P., Dugua, C., & Fayol, M. (2009). Liaison acquisition, word segmentation and

construction in French: A usage-based account. Journal of Child Language, 36(3),

557-596.

Chevrot, J. P., Nardy, A., & Barbu, S. (2011). Developmental dynamics of SES-related

differences in children's production of obligatory and variable phonological

alternations. Language Sciences, 33(1), 180-191.

Chong, A. J. (2011). Lenition in Gaalpu: An optimality theoretic analysis. Australian

Journal of Linguistics 31(4), 473-490.

Church, B. A., & Schacter, D. L. (1994). Perceptual specificity of auditory priming:

Implicit memory for voice intonation and fundamental frequency. Journal of

Experimental Psychology: Learning, Memory, and Cognition, 20(3), 521-533.

Coady, J. A., & Aslin, R. N. (2004). Young children's sensitivity to probabilistic

phonotactics in the developing lexicon. Journal of Experimental Child Psychology,

89(3), 183-213.

Cole, R., & Jakimik, J. (1980). A model of speech perception. In R. Cole (Ed.), Perception

and production of speech (pp. 133-163). Hillsdale, NJ: Lawrence Erlbaum

Associates. REFERENCES 226

Coltheart, M., Curtis, B., Atkins, P., & Haller, M. (1993). Models of reading aloud: Dual-

route and parallel-distributed-processing approaches. Psychological Review,

100(4), 589-608.

Cooper, R. P., & Aslin, R. N. (1990). Preference for infant-directed speech in the first

month after birth. Child Development, 61(5), 1584-1595.

Coussé, E., Gillis, S., Kloots, H., & Swerts, M. (2004). The influence of the labeller’s

regional background on phonetic transcriptions: Implications for the evaluation of

spoken language resources. In M. Lino, M. Xavier, F. Ferreira, R. Costa, & R. Silva

(Eds.), Proceedings of the 4th International Conference of Language Resources

and Evaluation, Paris: ELRA.

Cox, F. (2006). Australian English pronunciation into the 21st century. Prospect: The

Journal of the Adult Migrant Education Program, 21(1), 3-21.

Cox, F. (2012). Australian English: Pronunciation and transcription. Melbourne:

Cambridge University Press.

Cox, F., & Palethorpe, S. (2010). Australian Voices, Macquarie University. Retrieved from

http://clas.mq.edu.au/australian-voices/australian-voices.

Cox, F. M., & Palethorpe, S. (2004). The border effect: Vowel differences across the

NSW/Victorian border. In C. Moskovsky (Ed.). Proceedings of Australian

Linguistics Society Conference (pp. 1-14). Newcastle, Australia: Australian

Linguistics Society.

Cox, F. M., & Palethorpe, S. (2007). Illustrations of the IPA: Australian English. Journal

of the International Phonetic Association, 37(3), 341-350.

Cox, F. M., & Palethorpe, S. (2010). Broadness variation in Australian English. In M. REFERENCES 227

Tabain, J. Fletcher, D. Grayden, J. Hajek, & A. Butcher (Eds.). Proceedings of the

12th Australasian International Conference of Speech Science and Technology,

Melbourne, Australia.

Cristià, A. (2010). Phonetic enhancement of sibilants in infant-directed speech. Journal of

Acoustical Society of America, 128(1), 424-434.

Cristià, A. (2011). Fine-grained variation in caregivers' /s/ predicts their infants' /s/

category. Journal of Acoustical Society of America, 129(5), 3271-3280.

Cruttenden, A. (1994). Phonetic and prosodic aspects of baby talk. In C. Gallaway & B. J.

Richardson (Eds.), Input and interaction in language acquisition (pp. 135-152).

Cambridge: Cambridge University Press.

Cucchiarini, C. (1996). Assessing transcription agreement: Methodological aspects.

Clinical Linguistics and Phonetics, 10(2), 131-155.

Cutler, A. (1993). Phonological cues to open- and closed-class words in the processing of

spoken sentences. Journal of Psycholinguistic Research, 22(2), 109-131.

Daland, R. (2013). Variation in the input: A case study of manner class frequencies.

Journal of Child Language, doi:101017/S0305000912000372.

Davis, B. L., & Lindblom, B. (2001). Phonetic variability in Baby Talk and development

of vowel categories. In F. Lacerda, C. v. Hofsten, & M. Heiman (Eds.), Emerging

cognitive abilities in early infancy (pp. 135-171). Mahwah, New Jersey: Lawrence

Erlbaum. de Jong, K. J., Beckman, M. E., & Edwards, J. (1993). The interplay between prosodic

structure and coarticulation. Language and Speech, 36(2), 197-212.

Dinnsen, D. A. (1992). Variation in developing and fully developed phonetic inventories. REFERENCES 228

In C. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.), Phonological development:

Models, research, implications (pp. 191-210). Timonium, MD: York Press.

Dixon, R. M. W. (2002). Australian languages: Their nature and development.

Cambridge: Cambridge University Press.

Dominey, P. F., & Dodane, C. (2004). Indeterminacy in language acquisition: The role of

child directed speech and joint attention. Journal of Neurolinguistics, 17(2), 121-

145.

Eckert, P. (2001). Style and social meaning. In P. Eckert & J. Rickford (Eds.), Style and

sociolinguistic variation (pp. 119-126). New York: Cambridge University Press.

Edwards, J., & Beckman, M. E. (2008). Methodological questions in studying consonant

acquisition. Clinical Linguistics and Phonetics, 22(12), 939-958.

Edwards, J., Beckman, M. E., & Munson, B. (2004). The interaction between vocabulary

size and phonotactic probability effects of children's production accuracy and

fluency in nonword repetition. Journal of Speech, Language and Hearing

Research, 47(2), 421-436.

Edwards, J., & Munson, B. (2012). Transcribing the speech of multilingual children with

speech sound disorder. In S. MacLeod & B. Goldstein (Eds.), Multilingual aspects

of speech sound disorders in children. Clevedon, UK: Multilingual Matters.

Escure, G. (1977). Hierarchies and phonological weakening. Lingua, 43(1), 55-64.

Ferguson, C. A. (1977). Baby talk as a simplified register. In C. E. Snow & C. A. Ferguson

(Eds.), Talking to children: language input and acquisition (pp. 209-235).

Cambridge: Cambridge University Press.

Fernald, A. (1985). Four-month-old infants prefer to listen to motherese. Infant Behaviour REFERENCES 229

and Development, 8(2), 181-195.

Fernald, A. (1989). Intonation and communicative intent in mothers' speech to infants: Is

the melody the message? Child Development, 60(6), 1497-1510.

Fernald, A., & Morikawa, H. (1993). Common themes and cultural variations in Japanese

and American mothers' speech to infants. Child Development, 64(3), 637-656.

Fernald, A., Taeschner, T., Dunn, J., Papousek, M., de Boysson-Bardies, B., & Fukui, I.

(1989). A cross-language study of prosodic modifications in mothers' and fathers'

speech to preverbal infants. Journal of Child Language, 16(3), 477-501.

Finegan, F. E., & Biber, D. (2001). Register variation and social dialect variation: The

Register Axiom. In P. Eckert & J. R. Rickford (Eds.), Style and sociolinguistic

variation. Cambridge: Cambridge University Press.

Forster, K. I. (1976). Accessing the mental lexicon. In F. J. Wales & E. Walker (Eds.), New

approaches to language mechanisms (pp. 257-287). Amsterdam: North-Holland.

Foulkes, P., & Docherty, G. (2006). The social life of phonetics and phonology. Journal of

Phonetics, 34(4), 409-438.

Foulkes, P., Docherty, G., & Watt, D. (2005). Phonological variation in child-directed

speech. Language, 81(1), 177-206.

Giles, H., Coupland, N., & Coupland, J. (1991). Accommodation theory: Communication,

context, and consequence. In H. Giles, J. Coupland, & N. Coupland (Eds.),

Contexts of accommodation (pp. 1-68). New York, NY: Cambridge University

Press.

Gillon, G. T. (2004). Phonological awareness: From research to practice. New York:

Guilford Press. REFERENCES 230

Goddijn, S., & Binnenpoorte, D. (2003). Assessing manually corrected broad phonetic

transcriptions in the Spoken Dutch Corpus. Proceedings of the 15th International

Congress of Phonetic Sciences (ICPhS), Barcelona, Spain.

Goldinger, S. D. (1996). Words and voices: Episodic traces in spoken word identification

and recognition memory. Journal of Experimental Psychology: Learning, Memory

and Cognition, 22(5), 1166-1183.

Goldinger, S. D. (1998). Echoes of echoes? An episodic theory of lexical access.

Psychological Review, 105(2), 251-279.

Goldinger, S. D., & Azuma, T. (2004). Episodic memory reflected in printed word naming.

Psychonomic Bulletin & Review, 11(4), 716-722.

Grieser, D. L., & Kuhl, P. K. (1988). Maternal speech to infants in a tonal language:

Support for universal prosodic features in motherese. Developmental Psychology,

24(1), 14-20.

Hardison, D. M. (2003). Acquisition of second-language speech: Effects of visual cues,

context, and talker variability. Applied Psycholinguistics, 24(4), 495-522.

Harrington, J., Cox, F. M., & Evans, Z. (1997). An acoustic phonetic study of broad,

general, and cultivated Australian English vowels. Australian Journal of

Linguistics, 17(2), 155-184.

Hawkins, S. (2003). Roles and representations of systematic fine phonetic detail in speech

understanding. Journal of Phonetics, 31(3), 373-405.

Hazan, V., & Barrett, S. (2000). The development of phonemic categorization in children

aged 6-12. Journal of Phonetics, 28(4), 277-396.

Hock, H. H. (1986). Principles of historical linguistics. Berlin: Mouton de Gruyter. REFERENCES 231

Honeybone, P. (2008). Lenition, weakening and consonantal strength: Tracing concepts

through the history of phonology. In J. Brandão de Carvalho, T. Scheer, & P.

Ségéral (Eds.), Lenition and fortition (pp. 9-93). Berlin: Mouton de Gruyter.

Horvath, B. M. (1985). Variation in Australian English: The sociolects of Sydney.

Cambridge: Cambridge University Press.

Hudson, J. (1985). Grammatical and semantic aspects of Fitzroy Valley Kriol. Darwin:

Summer Institute of Linguistics.

Ingram, J. C. L. (1989). Connected speech processes in Australian English. Australian

Journal of Linguistics, 9, 21-49.

Jones, C., & Meakins, F. (2012a). Variability in stop consonant voicing in Gurindji Kriol,

a mixed language of northern Australia. Manuscript submitted for publication.

Jones, C., & Meakins, F. (2012b). The form and function of janyarrp: Gurindji Kriol child-

directed speech. Manuscript submitted for publication.

Jones, C., Meakins, F., & Buchan, H. (2011). Comparing vowels in Gurindji Kriol and

Katherine English: Citation speech data. Australian Journal of Linguistics, 31(3),

305-326.

Jones, C., Meakins, F., & Muawiyath, S. (to appear). Learning vowel categories from

maternal speech in Gurindji Kriol. Language Learning, 62.

Julien, H., & Munson, B. (to appear). Modifying speech to children based on their

perceived phonetic accuracy. Journal of Speech, Language and Hearing Research.

Jusczyk, P. W., Cutler, A., & Redanz, N. J. (1993). Infants' preference for the predominant

stress patterns of English words. Child Development, 64(3), 675-687.

Jusczyk, P. W., Friederici, A. D., Wessels, J. M. I., Svenkerud, V. Y., & Jusczyk, A. M. REFERENCES 232

(1993). Infants' sensitivity to the sound patterns of native language words. Journal

of Memory and Language, 32(3), 402-420.

Jusczyk, P. W., Luce, P. A., & Charles-Luce, J. (1994). Infants' sensitivity to phonotactic

patterns in the native language. Journal of Memory and Language, 33(5), 630-645.

Kaplan, A. (2011). Perceptual pressures on lenition. Language and Speech, 54(3), 285-305.

Katz, G. S., Cohn, J. F., & Moore, C. A. (1996). A combination of vocal f0 dynamic and

summary features discriminates between three pragmatic categories of infant-

directed speech. Child Development, 67(1), 205-217.

Kearns, K. P., & Simmons, N. N. (1988). Interobserver reliability and perceptual ratings:

More than meets the ear. Journal of Speech and Hearing Research, 31(4), 131-136.

Kempster, G. B., Gerratt, B. R., Abbott, K. V., Barkmeier-Kraemer, J., & Hillman, R. E.

(2009). Consensus auditory-perceptual evaluation of voice: Development of a

standardized clinical protocol. American Journal of Speech-Language Pathology,

18(2), 124-132.

Kent, R. (1992). The biology of phonological development. In C. Ferguson, L. Menn, & C.

Stoel-Gammon (Eds.), Phonological development: Models, research, implications

(pp. 65-90). Timonium, MD: York Press.

Kerswill, P., & Williams, A. (2000). Creating a New Town koine: Children and language

change in Milton Keynes. Language in Society, 29(1), 65-115.

Kingston, J. (2008). Lenition. Proceedings of 3rd Conference on Laboratory Approaches

to Spanish Phonology, Somerville: Cascadilla Press.

Kirchhoff, K., & Schimmel, S. (2005). Statistical properties of infant-directed versus adult-

directed speech: Insights from speech recognition. Journal of the Acoustical Society REFERENCES 233

of America, 117(4), 2238-2246.

Kirchner, R. (2004). Consonant lenition. In B. Hayes, R. Kirchner, & D. Steriade (Eds.),

Phonetically-Based Phonology (pp. 313-345). West Nyack, NY: Cambridge

University Press.

Kitamura, C., Thanavishuth, C., Burnham, D., & Luksaneeyanawin, S. (2002).

Universality and specificity in infant-directed speech: Pitch modifications as a

function of infant age and sex in a tonal and non-tonal language. Infant Behaviour

and Development, 24(4), 372-392.

Ko, E. S. (2012). Nonlinear development of speaking rate in child-directed speech. Lingua,

122(8), 841-857.

Kong, E. J., & Edwards, J. (2011). Individual differences in speech perception: Evidence

from visual analogue scaling and eye-tracking. Proceedings of the International

Congress of Phonetic Science XVII, Hong Kong.

Koreman, J. (2006). Perceived speech rate: The effects of articulation rate and speaking

style in spontaneous speech. Journal of the Acoustical Society of America, 119(1),

582-596.

Kuhl, P. K., Andruski, J. E., Chistovich, I. A., Chistovich, L. A., Kozhevnikova, E. V.,

Ryskina, V. L. (1997). Cross-language analysis of phonetic units in language

addressed to infants. Science, 277(5326), 684-686.

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.

Labov, W. (1984). Field methods in the project on linguistic change and variation. In J. REFERENCES 234

Baugh & J. Sherzer (Eds.), Language in use: Readings in sociolinguistics (pp. 28-

53). Englewood Cliffs, NJ: Prentice-Hall.

Labov, W. (1989). The child as a linguistic historian. Language Variation and Change 1,

85-97.

Labov, W. (2001). Principles of linguistic change. Volume 2: Social factors. Oxford:

Blackwell.

Ladefoged, P., & Maddieson, I. (1996). The sounds of the world's languages. Oxford &

Malden, MA: Blackwell.

Lee, S., Davis, B., & MacNeilage, P. F. (2008). Segmental properties of input to infants: A

study of Korean. Journal of Child Language, 35(3), 591-617.

Lee, S., & Davis, B. L. (2010). Segmental distribution patterns of English infant- and

adult-directed speech. Journal of Child Language, 37(4), 767-791.

Lindblom, B. (1990). Explaining phonetic variation: A sketch of the H&H theory. In W. J.

Hardcastle & A. Marchal (Eds.), Speech production and speech modelling (pp.

403-439). Dordrecht, Netherlands: Kluwer Academic Publishers.

Liu, H. M., Tsao, F. M., & Kuhl, P. K. (2009). Age-related changes in acoustic

modifications of Mandarin maternal speech to preverbal infants and five-year-old

children: A longitudinal study. Journal of Child Language, 36(4), 909-922.

Loakes, D., Hajek, J., & Fletcher, J. (2010). The /el/-/æ/ sound change in Australian

English: A preliminary perception experiment. In Y. Treis & R. de Busser (Eds.),

Selected papers from the 2009 conference of the Australian Linguistic Society.

Louko, L. J., & Edwards, M. L. (2001). Issues in collecting and transcribing speech

samples. Topics in Language Disorders, 21(4), 1-11. REFERENCES 235

MacWhinney, B. (2000). The CHILDES project: Tools for analyzing talk (3rd ed.).

Mahwah, NJ: Lawrence Erlbaum Associates Inc.

Martinez-Sussmann, C., Akhtar, N., Diesendruck, G., & Markson, L. (2011). Orienting to

third-party conversations. Journal of Child Language, 38(2), 273-296.

Maye, J., Weiss, D. J., & Aslin, R. N. (2008). Statistical phonetic learning in infants:

Facilitation and feature generalization. Developmental Science, 11(1), 122-134.

McConvell, P. (2008). Mixed languages as outcomes of code-switching: Recent examples

from Australia and their implications. Journal of Language Contact, THEMA II,

187-212.

McConvell, P., & Meakins, F. (2005). Gurindji Kriol: A mixed language emerges from

code-switching. Australian Journal of Linguistics, 25(1), 9-30.

Meakins, F. (2007). Case marking in contact: The development and function of case

morphology in Gurindji Kriol, an Australian mixed language (Doctoral

dissertation). University of Melbourne, Australia.

Meakins, F. (2008). Unravelling languages: Multilingualism and language contact in

Kalkaringi. In J. Simpson & G. Wigglesworth (Eds.), Children’s language and

multilingualism: Indigenous language use at home and school (pp. 247-264).

London: Continuum International Press.

Meakins, F. (2009). The case of the shifty ergative marker: A pragmatic shift in the

ergative marker in one Australian mixed language. In J. Barddal & S. Chelliah

(Eds.), The role of semantics and pragmatics in the development of case (pp. 59-

91). Amsterdam: John Benjamins.

Meakins, F. (2010). The development of asymmetrical serial verb constructions in an REFERENCES 236

Australian mixed language. Linguistic Typology, 14(1), 1-38.

Meakins, F. (2011). Case-marking in contact: The development and function of case

morphology in Gurindji Kriol. Amsterdam: John Benjamins.

Meakins, F. (in press). Gurindji Kriol. In S. Michaelis, P. Maurer, M. Haspelmath, & M.

Huber (Eds.), Atlas of pidgin and creole structures (Vol III, pp. 131-139). Oxford:

Oxford University Press.

Mitchell, A. G., & Delbridge, A. (1965). The pronunciation of English in Australia.

Sydney: Angus and Robertson.

Monaghan, P., Chater, N., & Christiansen, M. H. (2005). The differential role of

phonological and distributional cues in grammatical categorisation. Cognition

96(2), 143-182.

Morphy, F. (1983). Djapu, a Yolngu dialect. In R. M. W. Dixon & B. J. Blake (Eds.),

Handbook of Australian Languages (Vol 3). Canberra: ANU Press.

Morton, J. (1979). Facilitation in word recognition: Experiments causing change in the

logogen model. In P. A. Kolers, M. E. Wrolstad, & H. Bouma (Eds.), Processing

visible language 1 (pp. 259-268). New York: Plenum.

Munro, M. J., & Derwing, T. M. (2006). The functional load principle in ESL

pronunciation instruction: An exploratory study. System, 34(4), 520-531.

Munson, B., Edwards, J., & Beckman, M. E. (2005). Relationships between nonword

repetition accuracy and other measures of linguistic development in children with

phonological disorders. Journal of Speech, Language and Hearing Research, 47(1),

61-78.

Munson, B., Edwards, J., & Beckman, M. E. (2011). Phonological representations in REFERENCES 237

language acquisition: Climbing the ladder of abstraction. In A. C. Cohn, C.

Fougeron, & M. K. Huffman (Eds.), Handbook of laboratory phonology. Oxford:

Oxford University Press.

Munson, B., Edwards, J., Schellinger, S. K., Beckman, M. E., & Meyer, M. K. (2010).

Deconstructing phonetic transcription: Covert contrast, perceptual bias, and an

extraterrestrial view of Vox Humana. Clinical Linguistics and Phonetics, 24(4-5),

245-260.

Munson, B., Johnson, J., & Edwards, J. (2012). The role of experience in the perception of

phonetic detail in children's speech: A comparison of speech-language pathologists

with clinically untrained listeners. American Journal of Speech-Language

Pathology, 21(2), 124-139.

Munson, B., Kurtz, B. A., & Windsor, J. (2005). The influence of vocabulary size,

phonotactic probability, and wordlikeness of nonword repetitions of children with

and without specific language impairment. Journal of Speech, Language and

Hearing Research, 48(5), 1033-1047.

Munson, B., Schellinger, S. K., & Urberg-Carlson, K. (2012). Measuring speech-sound

learning using visual analog scaling. Perspectives of Language Learning and

Education, 19(1), 19-30.

Munson, B., Swenson, C. L., & Manthei, S. C. (2005). Lexical and phonological

organization in children: Evidence from repetition tasks. Journal of Speech,

Language and Hearing Research, 48(1), 108-124.

Nelson, K. (1973). Structure and strategy in learning to talk. Monographs of the Society for

Research in Child Development, 38(1 and 2, Serial No. 149). REFERENCES 238

Nygaard, L. C., & Pisoni, D. B. (1998). Talker-specific learning in speech perception.

Perception and Psychophysics, 60(3), 355-376.

Oller, D. K., & Eilers, R. E. (1975). Phonetic expectation and transcription validity.

Phonetica, 31(3-4), 288-304.

Palmeri, T. J., Goldinger, S. D., & Pisoni, D. B. (1993). Episodic encoding of voice

attributes and recognition memory for spoken words. Journal of Experimental

Psychology: Learning, Memory, and Cognition, 19(2), 309-328.

Picheny, M. A., Durlach, N. I., & Braider, L. D. (1986). Speaking clearly for the hard of

hearing II: Acoustic characteristics of clear and conversational speech. Journal of

Speech and Hearing Research, 29(4), 434-446.

Pierrehumbert, J. B. (2003). Phonetic diversity, statistical learning, and acquisition of

phonology. Language and Speech, 46(2-3), 115-154.

Pisoni, D. B. (1997). Some thoughts on "normalisation" in speech perception. In K.

Johnson & J. W. Mullennix (Eds.), Talker variability in speech processing (pp. 9-

32). San Diego: Academic Press.

Pollock, K. E., & Hinton, M. L. (2001). Phonetic transcription of African American

Vernacular English. Communication Disorders Quarterly, 23(1), 47-53.

Pruitt, J. S., Jenkins, J. J., & Strange, W. (2006). Training the perception of Hindi dental

and retroflex stops by native speakers of American English and Japanese. Journal

of the Acoustical Society of America, 119(3), 1684-1696.

Ralston, J. V., & Johnson, K. (1990). Context effects as a function of perceptual set.

Journal of the Acoustical Society of America, 87(S1), S46-S47.

Ramsdell, H. L., Oller, K. D., & Ethington, C. A. (2007). Predicting phonetic transcription REFERENCES 239

agreement: Insights from research in infant vocalizations. Clinical Linguistics and

Phonetics, 21(10), 793-831.

Roberts, J. (1997). Acquisition of variable rules: A study of (-t,d) deletion in preschool

children. Journal of Child Language 24(2), 351-372.

Rose, Y., Hedlund, G. J., Byrne, R., Wareham, T., & MacWhinney, B. (2007). Phon 1.2: A

computational basis for phonological database elaboration and model testing. In P.

Buttery, A. Villavicenci, & A. Korhonen (Eds.), Proceedings of the workshop on

cognitive aspects of computational language acquisition, 45th Annual Meeting of

the Association for Computational Linguistics (pp. 17-24). Stroudsburg, PA: ACL.

Rost, G. C., & McMurray, B. (2010). Finding the signal by adding noise: The role of

noncontrastive phonetic variability in early word learning. Infancy, 15(6), 608-635.

Sandefur, J. (1991). A sketch of the structure of Kriol. In S. Romaine (Ed.), Language in

Australia (pp. 204-212). Cambridge: Cambridge University Press.

Sandefur, J. R. (1979). An Australian creole in the Northern Territory: A description of

Ngukurr-Bamyili dialects (Part 1) (Vol. 3). Darwin: Summer Institute of

Linguistics.

Schellinger, S. (2008). The role of intermediate productions and listener expectations on

the perception of children's speech. (Unpublished master’s thesis). University of

Wisconsin-Madison.

Seidenberg, M. S., & McClelland, J. L. (1990). A distributed, developmental model of

word recognition and naming. Psychological Review, 96(4), 523-568.

Shi, R., Werker, J. F., & Cutler, A. (2006). Recognition and representation of function

words in English-learning infants. Infancy, 10(2), 187-198. REFERENCES 240

Shockey, L., & Bond, Z. S. (1980). Phonological processes in speech addressed to

children. Phonetica, 37(4), 267-274.

Shriberg, L. D., & Lof, G. L. (1991). Reliability studies in broad and narrow phonetic

transcription. Clinical Linguistics and Phonetics, 5(3), 225-279.

Simpson, J., & Wigglesworth, G. (2008). Children's language and multilingualism:

Indigenous language use at home and school. London: Continuum International

Press.

Singh, L., Morgan, J. L., & Best, C. T. (2002). Infants' listening preferences: Baby talk or

happy talk? Infancy, 3(3), 365-394.

Skorniakova, O., & Ito, K. (2011). Sub-phonemic variation during lexical identification:

Evidence from visual analogue scale (VAS) goodness ratings. Proceedings of the

International Congress of Phonetic Science XVII, Hong Kong.

Slawinski, E. B., & Fitzgerald, L. K. (1998). Perceptual development of the categorization

of the /r-w/ contrast in normal children. Journal of Phonetics, 26(1), 27-43.

Smith, J., Durham, M., & Fortune, L. (2007). "Mam, my trousers is fa'in doon!":

Community, caregiver, and child in the acquisition of variation in a Scottish dialect.

Language Variation and Change, 19(1), 63-99.

Smith, J., Durham, M., & Fortune, L. (2009). Universal and dialect-specific pathways of

acquisition: Caregivers, children and t/d deletion. Language Variation and Change,

21(1), 69-95.

Snow, C. E. (1977). Mothers' speech research: From input to interaction. In C. E. Snow &

C. Ferguson (Eds.), Talking to children: Language input and acquisition (pp. 31-

51). Cambridge: Cambridge University Press. REFERENCES 241

Snow, C. E., & Ferguson, C. A. (Eds.). (1977). Talking to children: Language input and

acquisition. Cambridge: Cambridge University Press.

Soderstrom, M. (2007). Beyond babytalk: Re-evaluating the nature and content of speech

input to preverbal infants. Developmental Review, 27(4), 501-532.

Solé, M. (2010). Effects of syllable position on sound change: An aerodynamic study of

final fricative weakening. Journal of Phonetics, 38(2), 289-305.

Stern, D. N., Spieker, S., Barnett, R. K., & MacKain, K. (1983). The prosody of maternal

speech: Infant age and context related changes. Journal of Child Language, 10(1),

1-15.

Stoel-Gammon, C. (2011). Relationships between lexical and phonological development in

young children. Journal of Child Language, 38(1), 1-34.

Strange, W. H., & Jenkins, J. J. (1978). Role of linguistic experience in perception of

speech. In R. D. Walk & H. L. Pick (Eds.), Perception and experience. New York:

Plenum Press.

Surendran, D., & Niyogi, P. (2006). Quantifying the functional load of phonemic

oppositions, distinctive features, and suprasegmentals. In O. N. Thomsen (Ed.),

Competing models of language change: Evolution and beyond. Amsterdam: John

Benjamins.

Swingley, D. (2007). Lexical exposure and word-form encoding in 1.5-year-olds.

Developmental Psychology, 43(2), 454-464.

Swingley, D. (2009). Contributions of infant word learning to language development.

Philosophical Transactions of the Royal Society B, 364(1536), 3617-3632.

Tagliamonte, S. A., & Temple, R. (2005). New perspectives on an ol' variable: (t,d) in REFERENCES 242

British English. Language Variation and Change, 17(3), 281-302.

Tenpenny, P. L. (1995). Abstractionist versus episodic theories of repetition priming and

word identification. Psychonomic Bulletin and Review, 2(3), 339-363.

Tollfree, L. F. (1999). South-east London English: discrete versus continuous modelling of

consonantal reduction. In P. Foulkes & G. J. Docherty (Eds.), Urban Voices (pp.

163-184). London: Arnold.

Tollfree, L. (2001). Variation and change in Australian English consonants. In D. Blair &

P. Collins (Eds.), Varieties of English around the world: English in Australia (pp.

45-67). Amsterdam: John Benjamins.

Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.),

Organization of memory (pp. 382-402). London: Academic Press.

Urberg-Carlson, K., Munson, B., & Kaiser, E. (2009). Gradient measures of children's

speech production: Visual analog scale and equal appearing interval scale measures

of fricative goodness. Journal of the Acoustical Society of America, 125(4), 2529-

2529. van de Weijer, J. (2002). How much does an infant hear in a day? In J. Costa & M. J.

Freitas (Chair), Associação Portuguesa de Linguistíca. Symposium conducted at

the meeting of the GALA2001 Conference on Language Acquisition, Lisboa.

Vihman, M. M., Nakai, S., DePaolis, R. A., & Hallé, P. (2004). The role of accentual

pattern in early lexical representation. Journal of Memory and Language, 50(3),

336-353.

Warren, R. M., & Warren, R. P. (1970). Auditory illusions and confusions. Scientific

American, 223(12), 30-36. REFERENCES 243

Wassink, A. B., Wright, R. A., & Franklin, A. D. (2007). Intraspeaker variability in vowel

production: An investigation of motherese, hyperspeech, and Lombard speech in

Jamaican speakers. Journal of Phonetics, 35(3), 363-379.

Wedel, A. B. (2006). Exemplar models, evolution and language change. The Linguistic

Review, 23(3), 247-274.

Wells, J. C. (1982). Accents of English. Cambridge: Cambridge University Press.

Werker, J. F., Fennell, C. T., Corcoran, K. M., & Stager, C. L. (2002). Infants' ability to

learn phonetically similar words: Effects of age and vocabulary size. Infancy, 3(1),

1-30.

Werker, J. F., & McLeod, P. J. (1989). Infant preference for both male and female infant-

directed talk: A developmental study of attentional affective responsiveness.

Canadian Journal of Psychology, 43(3), 230-246.

Werker, J. F., & Tees, R. C. (1999). Influences on infant speech processing: Toward a new

synthesis. Annual Review of Psychology, 50(1), 509-535.

Wheeler, M. A., Stuss, D. T., & Tulving, E. (1997). Toward a theory of episodic memory:

The frontal lobes and autonoetic consciousness. Psychological Bulletin, 121(3),

331-354.

Wilkinson, M. (1991). Djambarrpuyngu: A Yolngu variety of Northern Australia. PhD

thesis, University of Sydney.

Williams, A., & Kerswill, P. (1999). Dialect levelling: Continuity vs. change in Milton

Keynes, Reading and Hull. In P. Foulkes & G. J. Docherty (Eds.), Urban voices

(pp. 141-162). London: Arnold.

Yallop, C. (2003). A. G. Mitchell and the development of Australian pronunciation. REFERENCES 244

Australian Journal of Linguistics, 23(2), 129-141.

Youssef, V. (1991). Variation as a feature of language acquisition in the Trinidad context.

Language in Society 22(1), 257-274.

APPENDIX A

Participant Information

and Consent Forms APPENDIX A 246

15 July 2009

INFORMATION SHEET FOR DAGURAGU PARTICIPANTS

Research Project: Sounds in Speech to Children

Dear

You are invited to participate in a study being conducted by Dr Caroline Jones and Heather Buchan (University of Wollongong). Between 2004 and 2007 you participated in the ACLA research project with Felicity Meakins who made video and audio recordings of you talking to your children and other people around you. We are inviting you to help us write down the sounds in these recordings for a new project. This project is to find out about the sounds you use talking to children in Daguragu and Katherine.

INVESTIGATORS Dr Caroline Jones Heather Buchan Faculty of Education Faculty of Education [email protected] [email protected]

WHAT WE WOULD LIKE YOU TO DO If you choose to participate we will ask you to sit down at a computer with us and go through the recordings a bit at a time. We will show you how we wrote it down, and ask you to say bits slowly and some words clearly so we can get it right. We will also ask you to tell us about when you are talking to your child. We are happy to work in short sessions when you are free.

If you agree, we would like to record these sessions on video and audio. We would also ask you about how we should use and store recordings of you talking about the videos.

If you choose to take part you will be paid at $30.70 an hour before tax. We do not think there are risks for you. You can choose if you want to do it or not. You can stop whenever you like. You don’t have to tell us why and nothing bad will happen. Just let us know, and we can delete the recordings of you talking about the videos.

This project is funded by a grant from the Australian Research Council. There is no direct benefit to you. The project can be used to help teachers and community members with teaching traditional languages and English.

WHO WILL SEE MY INFORMATION?

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 247

No-one will be able to identify you or your child from the results of the study. Only the researchers will have access to this information. If you agree, we might use some photos or recordings when we present the information (for example at a conference).

All information will be stored securely at the University of Wollongong for at least 5 years and only the researchers will have access to it. You can choose to store your information at NCI and/or AIATSIS, or not. -NCI (National Computational Infrastructure) used to be called APAC (Australian Partnership for Advanced Computing). It stores the information you gave to Felicity Meakins for the ACLA (Aboriginal Child Language Acquisition) project. -AIATSIS is a national collection of audio and visual records of Australian Aboriginal and Torres Strait Islander cultures and histories.

If you would like to check that you are OK with the information or recordings from the study, you need to contact Dr Caroline Jones on (02) 4221 4905.

When you have read this information sheet, Dr Caroline Jones and Heather Buchan are free to talk about any questions. We will be happy to talk to you at any time in the project if you think of something later, or if any problems come up.

ETHICS REVIEW AND COMPLAINTS This study has been looked at by the Human Research Ethics Committee (Social Science, Humanities and Behavioural Science) of the University of Wollongong. If you have any problems with how the project has been done, you can contact the UoW Ethics Officer on (02) 4221 4457.

Thank you for your interest in this study.

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 248

CONSENT FORM FOR DAGURAGU PARTICIPANTS

Research Project: Sounds in Speech to Children Researchers’ Names: Dr Caroline Jones and Heather Buchan

1. I have read the participation information sheet and have been able to ask the researchers any questions.

2. I have discussed with the researchers how my data is going to be used and stored and I am happy with how the study has been explained to me.

3. I understand that my participation in this research is voluntary (I can choose) and I may withdraw at any time from the study without having to give a reason and without affecting my treatment in the community or at school in any way.

4. I understand that this study does not have any known risk for me, and I have read the information sheet and asked any questions about risks.

5. I understand that I will watch and listen to audio and visual recordings of me speaking to my children. These recordings are some of the ones that were made by Felicity Meakins between 2004 and 2007 for the ACLA project. I understand that I will help to write down my speech in the recordings.

6. I agree to be audio and video recorded while I help the researchers write it down. The researchers will delete anything I want them to.

7. I agree that I will get copies of all the recordings.

8. If I have any concerns or complaints about way the research is or has been conducted I can contact the Ethics Officer, Human Research Ethics Committee, Office of Research, University of Wollongong on 4221 4457.

Signed Date

...... /...... /...... Name (please print)

......

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 249

On this page you can tell us how the recordings can be used, and who can access the recordings. Please circle what you agree to:

Storing the videos and making sure you can get copies:

I would / would not like a video file of the recordings to be kept at Diwurruwurru- jaru (Katherine regional Aboriginal language centre). Tick one:
anyone can access
only me and my family
anyone with my permission
only me

I do / do not allow for the data I give to this study to be stored at NCI with the data I gave to the ACLA project.

Tick what you agree to: Future researchers can access:
transcripts
audio
video

future researchers have to contact me first, explain the research, and send me a copy when it’s done
future researchers have to contact me first and get my permission, explain the research, and send me a copy when it’s done If I am not available, the researcher must contact: ______

In five years time, I would / would not like to have recordings stored at AIATSIS.

other researchers can access
other people have to get my permission to access

Use in public (e.g. in books, conferences):

I do / do not give my permission for the researchers of this project to print parts of transcripts in books / journals and other printed reports.
names must be changed

I do / do not give my permission for the researchers to show short clips of video at conferences and meetings.
no names in the audio
faces in video blurred

I do / do not give my permission for the researchers to print photos for academic publications and reports / show photos at conferences and meetings / put photos on website.
any
check with me first

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 250

Signed

...... /...... /...... Name (please print) date

......

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 251

15 July 2009

INFORMATION SHEET FOR AUSTRALIAN ENGLISH PARTICIPANTS

Research Project: Sounds in Speech to Children

You are invited to participate in a study being conducted by Dr Caroline Jones and Heather Buchan (Education, University of Wollongong). The purpose of this project is to look at sound patterns in adults’ everyday speech to children in regional northern Australia.

INVESTIGATORS Dr Caroline Jones Heather Buchan Faculty of Education Faculty of Education [email protected] [email protected]

WHAT WE WOULD LIKE YOU TO DO If you choose to participate we will ask if we can record you in normal conversation to your child and other people around you during a normal day. To do this we will use a video camera, microphones and audio recorder.

We will ask to come back to make recordings every 6 months for a year and a half. We would like to make 3 to 4 hours of recordings each time. We will transcribe and do acoustic analyses of speech in the recordings and analyse the sound patterns.

We would also like to talk with you about how you want recordings of you and your child to be used and stored, and record these discussions.

If you choose to take part you will be paid $25 an hour to reimburse you for your time. We do not think there are risks for you or your child. You are free to decide if you want to be involved in this project or not and you can stop participating at any time. You do not need to give a reason for changing your mind, and there won’t be any negative effects for you if you choose to stop participating even if the study has started. All you have to do is let the researcher know, and any information we already have about you will be destroyed.

This project is funded by a grant from the Australian Research Council. Your participation will not directly benefit you. Participating will help us describe the sound patterns of speech to children in regional Australian English, and this information can be used to help teachers in northern Australia with teaching language to children.

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 252

WHO WILL SEE YOUR INFORMATION No-one will be able to identify you or your child from the results of the study. Only the researchers will have access to this information. If you agree, we might use some photos or recordings when we present the information. But we will only use them if you allow us, and you do not have to agree to this to participate.

All information will be stored securely at the University of Wollongong for at least 5 years and only the researchers will have access to it. In the attached consent form we ask you to tell us how you would like your data to be used and stored. If you would like to check that you are OK with the information or recordings from the study, you need to contact Dr Caroline Jones on (02) 4221 4905.

When you have read this information sheet, Dr Caroline Jones and Heather Buchan are free to talk about any questions you might have. They will be happy to talk to you at any time in the project if you think of something later, or if any problems come up.

ETHICS REVIEW AND COMPLAINTS This study has been reviewed by the Human Research Ethics Committee (Social Science, Humanities and Behavioural Science) of the University of Wollongong. If you have any concerns or complaints regarding the way this research has been conducted, you can contact the UoW Ethics Officer on (02) 4221 4457. Alternatively you can approach the Project Manager at Diwurruwurru-jaru Aboriginal Corporation or your local language centre representative.

Thank you for your interest in this study.

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 253

CONSENT FORM FOR AUSTRALIAN ENGLISH PARTICIPANTS

Research Project: Sounds in Speech to Children Researchers’ Names: Dr Caroline Jones and Heather Buchan

9. I have read the participation information sheet and have been able to ask the researchers any questions I may have had.

10. I have discussed with the researchers how my data is going to be used and stored and I am happy with how the study has been explained to me.

11. I understand that my participation in this research is voluntary and I may withdraw at any time from the study without having to give a reason and without affecting my treatment in the community in any way.

12. I understand that this study does not have any known risk for me, and I have read the information sheet and asked any questions I may have about the risks.

13. I understand that I will be recorded on video and audio in everyday communication with my child and other people around me who I would normally communicate with.

14. I understand that the researchers will contact me and ask to make recordings every 6 months for a year and a half, and each time they will make 3 to 4 hours of recordings.

15. I understand that I will be asked regularly to look at the recordings to make sure they are okay. I may also be asked to clarify to the researchers what I am saying in some of these recordings.

16. I understand that all of my work with the researchers will be video and audio recorded.

17. If I have any concerns or complaints regarding the way the research is or has been conducted I can contact the Ethics Officer, Human Research Ethics Committee, Office of Research, University of Wollongong on (02) 4221 4457.

Please sign on the next page if you agree to these terms.

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 254

To choose how and where we will keep the recordings, please circle whether you do or do not give permission for these things:

Storing the videos and making sure you can get copies:

I would / would not like a DVD copy of the recordings of my family.

TalkBank is a collection of videos, audio and transcriptions on the internet (http://talkbank.org/). It is accessed by researchers in child development.

I do / do not give my permission for my audio recordings to be stored on the TalkBank database, where it will be available to other researchers once this project has finished.

If you agree to storage on TalkBank please select your preferences:

What material would you like stored on TalkBank? (Please tick all that apply)
audio
transcript
video Do you want to be anonymous? (Please tick all that apply)
no, you can use my and my child’s real name
yes, by changing last names only
yes, by changing first names and last names
yes, by removing names from audio recordings
yes, by blurring faces in video recordings

Printing parts of the transcripts and showing videos and photos at conferences and language meetings:

I do / do not give my permission for the researchers of this project to print parts of transcripts in books / journals and other printed reports.
only with pseudonyms

I do / do not give my permission for the researchers to show short extracts of video at conferences and meetings.
only if the audio contains no name
only if the video blurs faces

I do / do not give my permission for the researchers to print photos for academic publications and reports / show photos at conferences and meetings / display photos on a project website
any
ask me about specific photos before using them

------

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 255

Signed

...... /...... /...... Name (please print) date

......

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 256

15 July 2009

PARTICIPANT INFORMATION SHEET FOR AUSTRALIAN ENGLISH CHILDREN

Research Project: Sounds in Speech to Children

Your child is invited to participate in a study being conducted by Dr Caroline Jones and Heather Buchan (Education, University of Wollongong). The purpose of this project is to look at sound patterns in adults’ everyday speech to children in regional northern Australia.

INVESTIGATORS Dr Caroline Jones Heather Buchan Faculty of Education Faculty of Education [email protected] [email protected]

WHAT WE WOULD LIKE YOU TO DO If you choose to participate we will ask if we can record you in normal conversation with your child and other people around you during a typical day. To do this we will use a video camera, microphones and audio recorder.

At the start of each recording session we will explain to your child in a simple way what we are doing and why, and we will ask your child if he or she would like to take part. You are welcome to help us with this.

We will ask to come back to make recordings every 6 months for a year and a half. We would like to make 3 to 4 hours of recordings each time. We will transcribe and do acoustic analyses of speech in the recordings and analyse the sound patterns.

We do not think there are risks for your child being involved in this study. You are free to decide if you want your child to be involved in this project or not, and you can take them out of the study at any time. You do not need to give a reason for changing your mind, and there won’t be any negative effects for you if you choose to withdraw your child even if the study has started. All you have to do is let the researcher know, and any information we already have about your child will be destroyed.

This project is funded by a grant from the Australian Research Council. Participation will not directly benefit your child. Participating will help us describe the sound patterns of speech to children in regional Australian English, and this information can be used to help teachers in northern Australia with teaching language to children.

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 257

WHO WILL SEE YOUR CHILD’S INFORMATION No-one will be able to identify you or your child from the results of the study. Only the researchers will have access to this information. If you agree on behalf of your child, we might use some photos or recordings when we present the information. But we will only use them if you allow us, and you do not have to agree to this for your child to participate.

All information will be stored securely at the University of Wollongong for at least 5 years and only the researchers will have access to it. In the attached consent form we ask you to tell us how you would like your child’s data to be used and stored.

If you would like to check that you are OK with the information or recordings from the study, you need to contact Dr Caroline Jones on (02) 4221 4905.

When you have read this information sheet, Dr Caroline Jones and Heather Buchan are free to talk about any questions you might have. We will be happy to talk to you at any time in the project if you think of something later, or if any problems come up.

ETHICS REVIEW AND COMPLAINTS This study has been reviewed by the Human Research Ethics Committee (Social Science, Humanities and Behavioural Science) of the University of Wollongong. If you have any concerns or complaints regarding the way this research has been conducted, you can contact the UoW Ethics Officer on (02) 4221 4457.

Thank you for your interest in this study.

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 258

CONSENT FORM FOR AUSTRALIAN ENGLISH CHILDREN PARTICIPANTS

TO BE READ AND SIGNED BY THE CHILD’S PARENT OR LEGAL GUARDIAN

Research Project: Sounds in Speech to Children Researchers’ Names: Dr Caroline Jones and Heather Buchan

1. I have read the participation information sheet and have been able to ask the researchers any questions I may have had.

2. I have discussed with the researchers how my child’s data is going to be used and stored and I am happy with how the study has been explained to me.

3. I understand that my child’s participation in this research is voluntary and I may withdraw my child at any time from the study without having to give a reason. I understand that this will not affect my child’s treatment in the community in any way.

4. I understand that this study does not have any known risk for my child, and I have read the information sheet and asked any questions I may have about the risks.

5. I understand that my child will be recorded on video and audio in everyday communication with me and other people around me who we would normally communicate with.

6. I understand that the researchers will contact me and ask to make recordings every 6 months for a year and a half, and each time they will make 3 to 4 hours of recordings.

7. I understand that I will be asked regularly to look at the recordings to make sure they are okay. I may also be asked to clarify to the researchers what my child or I are saying in some of these recordings.

8. I understand that the video and audio recorder will be recording all the time the researchers are with my child and myself.

9. If I have any concerns or complaints regarding the way the research is or has been conducted I can contact the Ethics Officer, Human Research Ethics Committee, Office of Research, University of Wollongong on (02) 4221 4457.

Please sign on the next page if you agree to these terms.

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 259

To choose how and where we will keep the recordings, please circle whether you do or do not give permission for these things as your child’s legal guardian:

Storing the videos and making sure you can get copies:

I would / would not like a DVD copy of the recordings of my family.

TalkBank is a collection of videos, audio and transcriptions on the internet (http://talkbank.org/). It is accessed by researchers in child development.

I do / do not give my permission for audio recordings to be stored on the TalkBank database, where it will be available to other researchers once this project has finished.

If you agree to storage on TalkBank please select your preferences:

What material would you like stored on TalkBank? (Please tick all that apply)
audio
transcript
video Do you want your child to be anonymous? (Please tick all that apply)
no, you can use my and my child’s real name
yes, by changing last names only
yes, by changing first names and last names
yes, by removing names from audio recordings
yes, by blurring faces in video recordings

Printing parts of the transcripts and showing videos and photos at conferences and language meetings:

I do / do not give my permission for the researchers of this project to print parts of transcripts in books / journals and other printed reports.
only with pseudonyms

I do / do not give my permission for the researchers to show short extracts of video at conferences and meetings.
only if the audio contains no name
only if the video blurs faces

I do / do not give my permission for the researchers to print photos for academic publications and reports / show photos at conferences and meetings / display photos on a project website
any
ask me about specific photos before using them

------

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au APPENDIX A 260

I give permission for my child ______(please insert your child’s name) to participate in this research.

Parent/ Guardian Signature ______

Date______Name (please print) ______

Faculty of Education University of Wollongong NSW 2522 Australia Telephone: (61 2) 4221 4905 Facsimile: (61 2) 4221 4657 [email protected] www.uow.edu.au

APPENDIX B

List of Gurindji Kriol Words Containing

Potential Fricatives APPENDIX B 262

List of words containing ‘potential fricatives’ that were variable, always a fricative, and always a stop. Words in bold are English code-switches. Underlined words are recent borrowings.

Variable All fricatives All stops abta PREP accident N abim V baind V atsaid DIR ajan DET biksim baldan V V bif N (medial) bej (initial) ADV disko N dat DEM bej (final) ADV dragonflai N den CONG bidio N fait V dens V fence bijin (initial) V N dey PRO (initial) fence bijin (medial) V N dij (initial) DEM (final) bijinlain N fij (final) N dijan (initial) DEM (medial) bijinlain N filim V dijei (initial) DEM (initial) ebriting biksim (initial) V flai N PRO (medial 1) ebriting binij (initial) DIS fon N PRO (medial 2) binij (final) DIS giraffe N eniting PRO bo PREP is V hab V jamting jamting bolouim V N N (initial) (medial) brom PREP klosim V jarran DEM bust V laswan ADJ jeya DEM but N live ADJ jumok N clothes N pressure N knife N sausage dij (final) DEM N lab V (initial) sausage dijan (medial) DEM N labta AUX (medial) dijei (medial) DEM separet ADJ lipt V drib V shade N mousey N elephant N shirt N naja DEM fij (initial) N sign N neba AUX APPENDIX B 263

Variable All fricatives All stops flour N singlet N neks ADV flower N sirkul N oji N futbal N skipping V tilip N gib V skirt N

gibit V so CONJ

hawuj N sori V

inside DIR square N

jarrei DEM star N

ji V stat-im V

jidan V stik N

jinek N sting V

jingat V stirimraun V

jouim V stori N

juka N swim V

jutim V telafon N

juwing N three ADV

libim V

lijin V

machine N

mub V

najan N

nes N

nojing EXCLAM

pleis N

pujim V

raifol N

said DIR

stop V

ting N

waip N

wajim V





APPENDIX B 264

Variable All fricatives All stops

washing N

yestadei N