close

Вход

Забыли?

вход по аккаунту

?

Production of Anticipatory Vowel Nasalization and Word-Final Nasal Consonants in Two Dialects of Spanish

код для вставкиСкачать
Production of anticipatory vowel
nasalization and word-final nasal
consonants in two dialects of Spanish
Silvina Bongiovanni
Submitted to the faculty of the University Graduate School in partial fulfillment of
the requirements for the degree Doctor of Philosophy in the Department of
Linguistics and the Department of Spanish and Portuguese, Indiana University
June 2018
ProQuest Number: 10824742
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
ProQuest 10824742
Published by ProQuest LLC (2018 ). Copyright of the Dissertation is held by the Author.
All rights reserved.
This work is protected against unauthorized copying under Title 17, United States Code
Microform Edition © ProQuest LLC.
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
Accepted by the Graduate Faculty, Indiana University, in partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
Doctoral Commitee
Committee Co-Chair, Kenneth de Jong, PhD
Committee Co-Chair, Erik Willis, PhD
Julie Auger, PhD
Manuel Díaz-Campos, PhD
May 3rd, 2018
ii
Copyright © 2018
Silvina Bongiovanni
iii
Acknowledgments
I had always thought of the effort and responsibility of this doctoral project as
all mine in its entirety. Yet, in writing these pages it became all too clear how this
dissertation, and more broadly my graduate career, would not have been possible
without the help, support, and friendship of so many people.
First and foremost, my deepest gratitude goes to Julie Auger, Kenneth de Jong,
Manuel Díaz-Campos and Erik Willis, the members of my dissertation committee.
Erik and Ken, my dissertation co-chairs, have encouraged me time and again to follow
curiosity and to chase new research questions. Without the foundation of their own
work and their tutelage, I would have been lost in this line of research. I also wish
to thank Manuel and Julie for making me a stronger thinker and better writer with
their support, advice and scrutiny, at the different stages of graduate school. As I am
wrapping up this project, I realize how much the teaching and guidance of these four
scholars have shaped who I am as a researcher and educator. They challenged me by
expecting more from me than I thought I could produce myself and they taught me
how to teach by example. It has been my privilege to study and train under them.
I am extremely grateful to Kimberly Geeslin for her support and her mentorship
throughout graduate school. Though not a member of my dissertation committee, Kim
has answered frantic emails and offered words of wisdom and reassurance too many
times to count. Her patience, integrity, generosity, and willingness to advocate on my
behalf have meant more to me than I can fully express here. Patrícia Amaral and Kelly
iv
Berkson have also been important mentors. Their invaluable advice, encouragement,
and support as I navigated the final years of graduate school helped me become
more confident in my work. They have proven to be good friends and understanding
colleagues; I feel extremely lucky that I was able to get to know them and work
alongside them.
Processing and analysis of the data could not have been undertaken without
the input of Daniel Dakota. Danny wrote the Python script to manage the huge
corpus of analysis (+20,000 files!). He also answered hundreds of questions about
programming and typesetting, always with patience and in great detail. Daniel ignited
everything I know about LaTeX, R and Praat scripting, and for this I will be always
grateful. I would also like to thank Michael Dow, the IU ‘pioneer’ in nasometry. In
addition to being the inspiration behind the methodological approach in this project,
Michael shared with me the Praat script that served as the basis for programming the
extraction of energy measurements and without which I would not have known where
to start with the recordings.
I also would like to acknowledge the financial and logistic support from various
entities and individuals. The Dissertation Completion Fellowship from the College
of Arts and Sciences allowed me to focus on my research full time. Additionally,
the Grant-In-Aid from the Graduate School and the Householder Research Award
from the Department of Linguistics funded the travel expenses to collect the data in
the Dominican Republic. In Santo Domingo, José Alejandro Rodríguez Núñez and
Ana Margarita Haché de Yunén of Pontificia Universidad Católica Madre y Maestra
facilitated participant recruitment and data collection. And finally, the purchase of
the nasometer was possible thanks to budgetary magic by Robert Botne and Ken de
Jong, in the Department of Linguistics.
There is no way graduate school would have been half as rewarding without the
v
support of wonderful friends. Nora Gardner, Cristobal Garza, Kaitlin Guidarelli, Erin
Lavin, Avizia Long, Rafael Monteiro, Alysa Schroff and Sara Zahler helped me stay
grounded and move forward at different points, both good and not so good. During
the lonely parts of the process, Nicolás Bamballi, Kelly Casper-Cushman, Christie
Cole, Danielle Daidone, Travis Evans-Sago, Carly Henderson, Beth Herring, Daniel
Jung, Matt Kanwit, Ryan Lidster, Samson Lotven, Vanessa Elias, Sean McKinnon,
Ian Michalski, Andrea Mojedano, Olivia Salzano, Megan Solon, and Melissa Whatley
provided good laughs and companionship. And I am always thankful to the pandilla
that started it all –Carolina Ayala, Ana Brown, Lucas Fiszman, Ana Moreno, and
Florencia Sartori– as well as to Jimena Montaña and my friends in Buenos Aires,
without whom home would not be home anymore.
Finally, nothing I do would have been possible without the love and support of
my family. My parents and my brothers have always supported my crazy endeavors,
from taking over our kitchen with candle-making to pursuing graduate school abroad.
I know that my academic ambition was not easy for them as it meant that was I very
far away, but they enthusiastically encouraged me to go after my goals nonetheless.
Everything I do is with the hope of making them proud.
vi
Silvina Bongiovanni
Production of anticipatory vowel nasalization and word-final nasal
consonants in two dialects of Spanish
People in different regions have certain ways of talking that set them apart from
others. The study of dialectal variation allows examining what happens to linguistic
systems as they diverge and allows for an investigation of the details of variability in
parallel systems. As such, regional variation provides a window into the principles
that underlie language differences, language innovation and language variation and
change. This dissertation focuses on dialectal differences in Spanish with regard to
word-final nasal consonants and anticipatory vowel nasalization.
Research on dialectal differences in Spanish repeatedly makes the observation that
the dialects with a preference for velarized variants of /n/ (such as Caribbean dialects)
have extensive anticipatory vowel nasalization, due to weakening of the word-final nasal
consonant. Despite these long-held claims, comparisons across Spanish dialects are
lacking. This dissertation fills this void by comparing production of word-final nasal
consonants and anticipatory vowel nasalization in a Caribbean and in a non-Caribbean
dialect of Spanish. In combining the analysis of both variables, this doctoral project
permits the discussion of not only dialectal differences in the production of each
acoustic feature, but co-variation between them as well.
In order to investigate these dialectal differences, I conducted a phonetic study
of Santo Domingo (Dominican Republic, 31 speakers) and Buenos Aires (Argentina,
28 speakers) Spanish using a nasometer, a split-channel microphone system which
records nasal and oral signals separately but simultaneously. Measurements of nasal
vii
and oral energy, as well duration, were extracted to characterize the time-course of
nasality and weakening of the nasal consonant.
Dialect groups were found to differ in terms of the time-course of nasality, with
Santo Domingo Spanish presenting earlier onset of velum lowering. Dialects were not
different in terms of weakening of the nasal consonant. The analysis of the relationship
between the time-course of nasality and nasal consonant weakening revealed that
as the nasal consonant weakened, lowering of the velum took place earlier, and this
trend obtained for both dialects. These findings, thus, confirm previous claims that
Caribbean dialects of Spanish presented more extensive anticipatory vowel nasalization,
but challenge previous descriptions of the process as a compensatory mechanism.
Committee Co-Chair, Kenneth de Jong, PhD
Committee Co-Chair, Erik Willis, PhD
Julie Auger, PhD
Manuel Díaz-Campos, PhD
viii
Table of Contents
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xv
1 Introduction . . . . . . . . . . . . . . . . . . . . . . .
1.1 Terms and definitions . . . . . . . . . . . . . . . . . .
1.2 Phonological interpretation of vowel nasality . . . . .
1.3 Overview of target dialects . . . . . . . . . . . . . . .
1.3.1 Buenos Aires Spanish . . . . . . . . . . . . . .
1.3.2 Santo Domingo Spanish . . . . . . . . . . . .
1.4 Goals . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5 Contributions . . . . . . . . . . . . . . . . . . . . . .
1.6 Structure of the thesis . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
3
4
6
6
9
12
13
15
2 Phonetics of nasals and nasalization . . . . . . . . . .
2.1 Articulation of nasal segments . . . . . . . . . . . . . .
2.2 Acoustics of nasal segments . . . . . . . . . . . . . . .
2.2.1 Nasal consonants . . . . . . . . . . . . . . . . .
2.2.2 Nasal(ized) vowels . . . . . . . . . . . . . . . .
2.3 Coarticulatory vowel nasalization . . . . . . . . . . . .
2.4 Diachronic evolution of vowel nasalization . . . . . . .
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
17
17
22
23
24
29
33
37
3 Word-final nasals and nasalization in Spanish . . . .
3.1 Historical development of the Spanish nasal system . .
3.2 Synchronic variation . . . . . . . . . . . . . . . . . . .
3.2.1 Variationist studies . . . . . . . . . . . . . . . .
3.2.2 Instrumental studies . . . . . . . . . . . . . . .
3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 39
. . 41
. 44
. 46
. 50
. . 61
4 The research protocol . .
4.1 Research sites . . . . . .
4.2 Participants . . . . . . .
4.2.1 Recruitment . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ix
. . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
.
.
.
.
62
67
70
70
4.2.2 Demographic information . . . . .
4.3 Instrument . . . . . . . . . . . . . . . . . .
4.3.1 Stimuli design . . . . . . . . . . . .
4.3.2 Elicitation task . . . . . . . . . . .
4.3.3 Background questionnaire . . . . .
4.4 Experimental Procedure . . . . . . . . . .
4.4.1 Equipment . . . . . . . . . . . . . .
4.4.2 Procedure . . . . . . . . . . . . . .
4.5 Data processing . . . . . . . . . . . . . . .
4.5.1 Data management . . . . . . . . .
4.5.2 Data annotation . . . . . . . . . .
4.6 Instrumental analyses . . . . . . . . . . . .
4.6.1 Weakening of the word-final nasal .
4.6.2 Time-course of nasality . . . . . . .
4.6.3 Relationship between time-course of
the nasal consonant . . . . . . . . .
4.7 Statistical analyses . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 71
. . . . . . . . . . . . . . . 72
. . . . . . . . . . . . . . . 72
. . . . . . . . . . . . . . . 74
. . . . . . . . . . . . . . . 77
. . . . . . . . . . . . . . . 78
. . . . . . . . . . . . . . . 78
. . . . . . . . . . . . . . . 80
. . . . . . . . . . . . . . . . 81
. . . . . . . . . . . . . . . . 81
. . . . . . . . . . . . . . . 82
. . . . . . . . . . . . . . . 93
. . . . . . . . . . . . . . . 94
. . . . . . . . . . . . . . . 95
nasality and weakening of
. . . . . . . . . . . . . . . 97
. . . . . . . . . . . . . . . 97
5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1 Control and target items . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Weakening of the nasal consonant . . . . . . . . . . . . . . . . . . . .
5.2.1 Duration of the nasal consonant . . . . . . . . . . . . . . . . .
5.2.2 Constriction in the nasal consonant . . . . . . . . . . . . . . .
5.3 Time-course of nasality . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4 Relationship between the time-course of nasality and weakening of the
nasal consonant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5 Summary of the chapter . . . . . . . . . . . . . . . . . . . . . . . . .
98
98
100
100
110
119
136
138
6 Discussion and conclusion . . . . . . . . . . . . . . . . . . . . . . .
6.1 Regional variation in weakening of the nasal consonant . . . . . . .
6.2 Regional variation in time-course of nasality . . . . . . . . . . . . .
6.3 The influence of linguistic variables . . . . . . . . . . . . . . . . . .
6.4 The relationship of time-course and duration of the nasal consonant
6.5 Instrumental analyses and sociolinguistics . . . . . . . . . . . . . .
6.6 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.7 Evaluation and recommendations for future research . . . . . . . . .
6.8 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
140
140
149
154
160
163
165
167
. 171
Appendices . . . . . . . . . . . . . . . . . . . . . . . . . .
A
Background questionnaire . . . . . . . . . . . . . . .
B
Participant demographic information . . . . . . . . .
C
Stimuli . . . . . . . . . . . . . . . . . . . . . . . . . .
D
Data Management . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
182
183
190
193
202
x
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Curriculum Vitae
xi
List of Figures
1.1
1.2
2.1
2.2
2.3
2.4
2.5
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
4.10
4.11
4.12
4.13
4.14
Map of dialectal areas in Argentina (Vidal de Battini, 1964) . . . . .
Map of dialectal areas in the Dominican Republic based on the production of coda liquids (Jiménez Sabater, 1975) . . . . . . . . . . . . . .
Midsagittal view of the the vocal tract . . . . . . . . . .
Anterior, posterior and lateral view of tongue muscles . .
Two-tube model of the vowel tract . . . . . . . . . . . .
Vocal tract transfer functions for oral, moderate nasalized
nasalized /i/ . . . . . . . . . . . . . . . . . . . . . . . . .
Vocal tract transfer functions for oral, moderate nasalized
nasalized /A/ . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
and heavily
. . . . . . .
and heavily
. . . . . . .
Map of Latin America highlighting the location of Santo Domingo and
Buenos Aires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zoomed map highlighting the location of Santo Domingo . . . . . . .
Zoomed map highlighting the location of Buenos Aires . . . . . . . .
PorwerPoint sample slides with carrier phrases. . . . . . . . . . . . .
PowerPoint slides depicting context of the carrier phrase task . . . . .
Nasometer being held by a participant . . . . . . . . . . . . . . . . .
Waveform and spectrogram of the word ‘bend’ . . . . . . . . . . . . .
Sample spectrogram of target word sartén / saRtén / ‘skillet’ in the
phrase la sartén caliente ‘the hot skillet’ . . . . . . . . . . . . . . . .
Sample spectrogram of target word tiraron / tiRáRon / ‘they threw’ in
the phrase cuando tiraron tabaco ‘when they threw tobacco’ . . . . .
Sample spectrogram of target word limón / limón / ‘lemon’ in the
phrase un ácido limón ‘an acid lemon’ . . . . . . . . . . . . . . . . .
Sample spectrogram of target word veintidos / beintidós / ‘twenty two’
in the phrase los veintidos aliados ‘the twenty two allies’ . . . . . . .
Sample spectrogram of target word veintidos /beintidós/ ‘twenty two’
in the phrase el número veintidos ‘the number twenty two’ . . . . . .
Sample spectrogram of target word veintidos /beintidós/ ‘twenty two’
in the phrase el número veintidos ‘the number twenty two’ . . . . . .
Sample spectrogram of target word esperan / espéRan / ‘(they) wait’ in
the phrase esperan callados ‘they wait silently’ . . . . . . . . . . . . .
xii
7
10
18
20
23
26
27
68
69
69
76
77
79
80
84
85
85
86
87
88
89
4.15 Sample spectrogram of target word capitán / kapitán / ‘captain’ in the
phrase el capitán astuto ‘the clever captain’ . . . . . . . . . . . . . . . 90
4.16 Sample spectrogram of target word canón /kánon/ ‘cannon’ in the
phrase el canón alegre ‘the joyful cannon’ . . . . . . . . . . . . . . . . . 91
4.17 Sample spectrogram of target word tiraron / tiRáRon / ‘they threw’ in
the phrase digo que tiraron ‘I say they threw’ . . . . . . . . . . . . . 92
4.18 Sample of minimum and maximum readings for 30 random tokens . . 96
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
5.10
5.11
5.12
5.13
5.14
5.15
5.16
5.17
5.18
5.19
5.20
Time-course of nasal energy for CVC, CVN and NVN of 10% of the
tokens in the corpus, per dialect . . . . . . . . . . . . . . . . . . . . . 99
Box, jitter plot and density plot for duration of the nasal consonant (in
ms), per dialect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Box, jitter plot and density plot for duration uration of the nasal
consonant (in ms) for syllable, stress and vowel type, per dialect . . . 105
Box, jitter plot and density plot for duration of the nasal consonant (in
ms) for syllable, stress and vowel type, per dialect . . . . . . . . . . . 107
Box, jitter plot and density plot for duration of the nasal consonant (in
ms) for pre-coronal and pre-dorsal environments, per dialect . . . . . 109
Box, jitter plot and density plot for consonant-to-vowel oral energy
ratio, per dialect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Box, jitter plot and density plot for consonant-to-vowel oral energy
ratio for syllable, stress and vowel type . . . . . . . . . . . . . . . . . 114
Line graph for interaction of environment with dialect group for consonantto-vowel oral energy ratio . . . . . . . . . . . . . . . . . . . . . . . . 115
Box, jitter plot and density plot for consonant-to-vowel oral energy
ratio for pre-pausal and pre-vocalic environments, per dialect . . . . . 116
Box, jitter plot and density plot for consonant-to-vowel oral energy
ratio for pre-coronal and pre-dorsal environments, per dialect . . . . . 118
Histograms for time-course of nasality per dialect . . . . . . . . . . . . 121
Time-course of nasal energy for CVN, per dialect . . . . . . . . . . . 123
Time-course of nasal energy for NVN, per dialect . . . . . . . . . . . 123
Histograms for time-course of nasality per syllable type and dialect . 125
Histograms for time-course of nasality per stress condition and dialect 127
Histograms for time-course of nasality per vowel condition and dialect 129
Histograms for time-course of nasality for the pre-pausal and pre-vocalic
conditions, BAS data . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Histograms for time-course of nasality for the pre-coronal and pre-dorsal
conditions, BAS data . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Scatterplot for duration of the nasal consonant as a function of onset
of the gesture, per dialect . . . . . . . . . . . . . . . . . . . . . . . . 137
Scatterplot for consonant-to-vowel oral enegy ratio as a function of
onset of the gesture, per dialect . . . . . . . . . . . . . . . . . . . . . 138
xiii
6.1
6.2
6.3
6.4
6.5
6.6
6.7
D.1
D.2
D.3
Duration (in ms) and consonant-to-vowel oral energy ratio of the nasal
consonant, per dialect . . . . . . . . . . . . . . . . . . . . . . . . . .
Duration of the nasal consonant (in ms) as a function of consonant-tovowel oral energy ratio, per dialect. . . . . . . . . . . . . . . . . . . .
Duration of word-final /s/ (in ms), per dialect. . . . . . . . . . . . . .
Time-course of nasality per dialect. . . . . . . . . . . . . . . . . . . .
Duration of the vowel (in ms), per syllable type. . . . . . . . . . . .
Oscilogram of a sine wave illustrating amplitude terms . . . . . . . .
Box, jitter plot and density plot for consonant-to-vowel oral energy
ratio (raw and transformed on the RAU scale), per dialect . . . . . .
Screen capture of the user interface with options to save the text file .
ASCII Filter Options . . . . . . . . . . . . . . . . . . . . . . . . . . .
Screen capture of the user interface with options to import the text file
into Open Office spreadsheet . . . . . . . . . . . . . . . . . . . . . . .
xiv
. 141
143
146
. 151
155
180
. 181
203
203
204
List of Tables
2.1
2.2
Frequency values for murmurs reported in the literature . . . . . . . .
Summary of acoustic measurements of vowel nasality . . . . . . . . .
25
28
3.1
3.2
3.3
40
42
3.6
Preconsonantal Spanish nasal allophones . . . . . . . . . . . . . . . .
Summary of the historical development of Spanish nasal consonants .
Variants of syllable-coda (word-internal and word-final) nasal consonants in the sociolinguistic literature . . . . . . . . . . . . . . . . . .
Frequency of syllable-coda nasal variants per domain in the sociolinguistic literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Frequency of deletion of word-final nasals per phonological environment
in the sociolinguistic literature . . . . . . . . . . . . . . . . . . . . . .
Summary of findings in Martínez (2015) . . . . . . . . . . . . . . . .
4.1
4.2
4.3
Example words for phonological contexts under study . . . . . . . . .
Example carrier phrases . . . . . . . . . . . . . . . . . . . . . . . . .
Excluded tokens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
75
93
3.4
3.5
5.1
47
48
49
60
Descriptive statistics on the duration of the nasal consonant (in ms),
per dialect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.2 ANOVA results for main effects and interaction terms for the analysis
of duration of the nasal consonant as dependent variable . . . . . . . 103
5.3 Descriptive statistics on the duration of the nasal consonant (in ms),
per syllable type and dialect . . . . . . . . . . . . . . . . . . . . . . . 103
5.4 Descriptive statistics on the duration of the nasal consonant (in ms),
per stress condition and dialect . . . . . . . . . . . . . . . . . . . . . 103
5.5 Descriptive statistics on the duration of the nasal consonant (in ms),
per stress condition and dialect . . . . . . . . . . . . . . . . . . . . . 104
5.6 Descriptive statistics on the duration of the nasal consonant (in ms)
for pre-pausal and pre-vocalic environments, per dialect . . . . . . . . 106
5.7 Descriptive statistics on the duration of the nasal consonant (in ms)
for pre-coronal and pre-dorsal environments, per dialect . . . . . . . . 108
5.8 Descriptive statistics of consonant-to-vowel oral energy ratio, per dialect110
5.9 ANOVA results for main effects and significant interaction for the
analysis of oral energy ratio as dependent variable . . . . . . . . . . . 112
xv
5.10 Descriptive statistics of consonant-to-vowel oral energy ratio for syllable
type, per dialect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.11 Descriptive statistics of consonant-to-vowel oral energy ratio for stress
condition, per dialect . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.12 Descriptive statistics of consonant-to-vowel oral energy ratio for vowel
type, per dialect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.13 Descriptive statistics of consonant-to-vowel oral energy ratio for prepausal and pre-vocalic, per dialect . . . . . . . . . . . . . . . . . . . .
5.14 Descriptive statistics of consonant-to-vowel oral energy ratio for predorsal and pre-coronal, per dialect . . . . . . . . . . . . . . . . . . . .
5.15 Descriptive statistics on time-course of nasality per dialect group . . .
5.16 Descriptive statistics on time-course of nasality per syllable type, BAS
data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.17 Descriptive statistics on time-course of nasality per syllable type, SDS
data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.18 Descriptive statistics on time-course of nasality per stress condition,
BAS data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.19 Descriptive statistics on time-course of nasality per stress condition,
SDS data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.20 Descriptive statistics on time-course of nasality per vowel type . . . .
5.21 Descriptive statistics on time-course of nasality for pre-pausal and
pre-vocalic environments, BAS data. . . . . . . . . . . . . . . . . . .
5.22 Descriptive statistics on time-course of nasality for pre-coronal and
pre-dorsal environments, BAS data. . . . . . . . . . . . . . . . . . .
5.23 Descriptive statistics on time-course of nasality for pre-pausal and
pre-vocalic environments, SDS data. . . . . . . . . . . . . . . . . . .
5.24 Descriptive statistics on time-course of nasality for pre-coronal and
pre-dorsal environments, SDS data . . . . . . . . . . . . . . . . . . .
113
113
113
116
117
122
125
126
127
128
130
134
134
134
135
A.1 Detailed demographic information of participants in the sample . . . 190
C.1 Stimuli phrases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
xvi
Chapter 1
INTRODUCTION
People in different regions have ways of talking that set them apart from others.
The study of dialectal variation allows to examine what can happen to linguistic
systems as they diverge and allows for an investigation of the details of variability itself
in parallel systems. As such, regional variation provides a window into the principles
underlying language innovation and language variation and change. To this end, this
dissertation examines cross-dialectal variation and focuses on differences in Spanish in
anticipatory vowel nasalization and word-final nasal consonants.
This dissertation finds its place at the crux of Hispanic linguistics, phonetics
and phonology, and sociolinguistics. Within Hispanic Linguistics, word-final nasal
consonants have occupied a prominent place both in dialectological and phonological
studies. Dialectological compendiums of Spanish, such as Canfield (1981), Lipski
(1994) and Resnick (1975), have proposed the realization of word-final nasal consonants
as one of the features that distinguish Spanish dialects: they either produce an alveolar
variant ([n]) or a velar variant ([N]). And in Spanish phonology, nasal place assimilation
and word-final velarization have been the object of many theoretical studies (Baković,
2000; Colina, 2009; Harris, 1984a, 1984b; Hualde, 1989a, 1989b; Piñeros, 2006, 2014;
Ramsammy, 2013).
Outside of the realm of Hispanic linguistics, nasals had a different fate, yet
1
just as interesting. The complex interplay between the phonetic and phonological
factors involving nasals and nasalization has attracted the attention of much phonetics
research. This work has examined the phonetic details of nasal(ized) segments, their
patterns of sound change, and the interplay between the production and the perception
of these segments. In particular, it is work examining cross-linguistic differences in
vowel nasalization that has served as the inspiration for this dissertation (e.g. Beddor
& Krakow, 1999; Cohn, 1993; Solé, 1992). Vowel nasalization has operated widely
in Romance languages, and for this reason, much of this work has focused in this
particular language family (e.g. Delvaux, Demolin, Harmegnies, & Soquet, 2008; Dow,
2014; Hajek, 1997; Sampson, 1999). However, despite the long tradition in Spanish
phonology and dialectology, the phonetics literature has little paid attention to Spanish
nasality.
A notable exception to the previous statement is Solé (1992, 1994, 1995), who
shows that vowel nasalization in Spanish is best described as a physiological result of
co-articulation with an adjacent nasal consonant, and it is not intended as a feature
of the vowel. However, the literature in Hispanic sociolinguistics contradicts this
‘laboratory description’ by reporting that some dialects of Spanish, especially in
the Caribbean region, may have phonologized vowel nasalization due to lenition of
word-final nasals (Campos-Astorkiza, 2012; Colantoni, 2011b; Hualde, 2014; Lipski,
2011; Vaquero, 1996). Despite these diverse descriptions, comparisons of patterns of
anticipatory vowel nasalization and nasal consonants across Spanish dialects are still
lacking (Colantoni, 2011b, p. 21). This dissertation addresses this gap by investigating
the production of anticipatory vowel nasalization and word-final consonants in a
Caribbean and a non-Caribbean dialect of Spanish. As such, this dissertation, marries
descriptions of nasality in the Hispanic linguistics tradition with phonetic research on
nasality.
2
The approach undertaken in this study is experimental and fits into the growing
body of work that applies a laboratory approach to the study of variation and change
in sound systems. While this approach may seem to undermine the sociolinguistic
enterprise, the use of experimental techniques provides more accurate characterizations
of phonological phenomena. In doing so, this line of work furthers our understanding
of the details of how universal and language-specific patterns shape specific sound
changes. The next three sections offer a brief presentation of terms and definitions, a
discussion of the phonological interpretation of vowel nasality and a description of
the dialects under study, prior to outlining the precise goals and contributions of this
dissertation.
1.1
Terms and definitions
Before the major issues that inform this doctoral project are laid out, some
explanation of terms frequently used in this thesis is due. The reader may have already
noticed that three terms appear with frequency: nasality, nasal and nasalization. I use
‘nasality’ to refer to segments produced with a lowered velum, be that a consonant or
a vowel. Thus, ‘nasality’ is taken to be an encompassing term. ‘Nasal’ is used to refer
to segments that are phonologically nasal, or [+nasal] in a featural approach, such
as nasal consonants, /m/ and /n/, and nasal vowels, /õ/ and /ũ/. When ‘nasal’ (and
its plural ’nasals’) appears by itself, it indicates nasal consonants. Vowels that are
[+nasal], on the other hand, will be referred to as ‘nasal vowels’. Finally, ‘nasalization’
refers to the process of producing a vowel while the velum is lowered, and ‘nasalized’
the adjective that qualifies the result of this process. This dissertation focuses on
a specific type of vowel nasalization, anticipatory (or regressive) vowel nasalization.
With anticipatory nasalization, we focus on the pre-nasal vowel, where the vowel
3
and the nasal consonant are tautosyllabic. Another type of nasalization relates to
carry-over nasalization. In this case, the vowel and the nasal are also in the same
syllable, but the nasal consonant precedes the vowel (i.e. NV).
Three other terms are crucial to this dissertation: distinctive nasality, allophonic
nasality, and phonetic nasality. The next section addresses these differences.
1.2
Phonological interpretation of vowel nasality
In phonological terms, nasality in vowels can have different phonological statuses.
Sampson (1999, p. 14-19) distinguishes five different types of vowel nasality: (a) generic,
(b) universal phonetic, (c) allophonic, (d) distinctive, and (e) prosodic (‘phonological’ in
Sampson, 1999). Generic nasality refers to vowel nasality that results from pathological
reasons (i.e., rhinolalia, cleft palate) and bears no phonological significance. Generic
nasality also describes oral segments may (and often do) present some degree of nasality
even though the segment is not nasalized underlyingly (Bell-Berti, 1976; Henderson,
1984).
Types (b) through (e), on the other hand, serve a function in the linguistic system.
Types (b) and (c) constitute non-distinctive conditioned nasality (that is [-nasal]),
given that the presence of nasality is conditioned by the phonetic environment in
which the vowel appears (i.e., adjacent to a nasal consonant). On one hand, universal
phonetic nasality refers to nasality that results from co-articulation with an adjacent
nasal consonant. As a consequence of the gestural overlap, part of the vowel adjacent
to the nasal segment presents some degree of nasality. This gestural retiming is
considered a low-level physiological adjustment, and for this reason, phonetic universal
nasality has also been labeled ‘low-level’ nasality. As mentioned in the previous section,
Spanish nasality has been described as a phonetic universal (Solé, 1992, 1995, see
4
section 3.2.2.1 in chapter 3).
In allophonic nasality, on the other hand, vowels present enhanced levels of nasality
when adjacent to the nasal consonant, and as a result, it has phonologized. In other
words, speakers have adopted nasality as an intended property of the vowel. In this
case, the gestural retiming constitutes a higher-level adjustment or a ‘preprogrammed’
activity (Solé, 1992, 1995), and for this reason, allophonic nasality has also been
termed high-level nasality.
In the case of distinctive nasality, type (d), vowel nasality cannot be ascribed
to the (synchronic) phonetic environment (e.g. an adjacent nasal consonant). Nasal
vowels appear in contexts in which oral vowels may occur, and the nasality cannot be
predicted by the phonetic environment. For example, French has pairs such as feinte
‘trick’ - fête ‘party’, where the nasality of [Ẽ] in fête cannot predicted on the phonetic
environment, as there is no adjacent nasal consonant. Thus, in this case, nasality is
distinctive as oral and nasal vowel are contrastive.
Finally, prosodic nasality (‘phonological’ in Sampson, 1999) refers to cases in which
nasality is not associated with a single vowel segment, but it occurs distinctively over
sequences of segments. In languages with prosodic nasality, such as some languages in
the Brazilian Amazonia, vowel nasality serves to distinguish individual morphemes at
the lexical level or a morphological marker that grammatically modifies lexical items.
This dissertation focuses on non-distinctive conditioned nasality: allophonic and
phonetic nasality. The fact that in some languages the presence of vowel nasality
has been phonologized and intended as part of the vocalic segment (English, for
example), while in other languages (such as Spanish), is a residue of co-articulation
indicates that languages differ in the patterns of contextual nasalization (Beddor
& Gottfried, 1995; Cohn, 1993; Sampson, 1999; Solé, 1992, 1995) (see chapters 2
and 3 for details as to how these two types of nasality surface in production and in
5
perception). Together with the aforementioned claims that in some dialects of Spanish
(for example, Caribbean dialects) pre-nasal vowels present extensive nasalization (see
details in section 3.2.1 in chapter 3), these facts suggest that the phonological status
of vowel nasality is different across dialects of Spanish. The present doctoral research
investigates these dialectal differences.
1.3
Overview of target dialects
This section presents a brief overview of the linguistic characteristics of the dialects
under study: Buenos Aires and Santo Domingo Spanish. More details regarding the
research sites are included in chapter 4, when the experimental protocol is presented.
1.3.1
Buenos Aires Spanish
The most comprehensive description of dialectal areas in Argentina, and to-date
probably the only one, belongs to Berta Vidal de Battini (Vidal de Battini, 1964).
She divides Argentina into five dialect areas, illustrated in Figure 1.1: (a) Rioplatense,
(b) Central, (c) Northwest, (d) Guaraní and (e) Cuyo. In the interest of time and
space, I will focus on the dialectal area that this doctoral project explores: Buenos
Aires Spanish, to which we turn next.
6
Figure 1.1: Map of dialectal areas in Argentina (Vidal de Battini, 1964)
Buenos Aires Spanish would fall in the realm of Vidal de Battini’s Rioplatense
area, which also includes the province of Buenos Aires, the province of Santa Fe, areas
of the province of Entre Ríos and La Pampa, as well as parts of the region of Patagonia.
Buenos Aires Spanish, and more generously Rioplatense Spanish, is often taken as
the prototype of what Argentines sound like, and therefore, taken to be ‘Argentine’
Spanish. But before we continue, due diligence compels me to set the record straight:
7
Argentine Spanish “does not end at the General Paz” (Colantoni, 2011a, p. 209.)1
Having made this important disclaimer, we can now focus on Buenos Aires Spanish
(henceforth, BAS).
Also called Rioplatense or Porteño Spanish, BAS refers to the dialect of Spanish
spoken in the region around the River Plate, i.e., the Buenos Aires metropolitan
area and most of Uruguay. Some of the linguistic features that have caught the
attention of research include “voseo”, i.e. the generalized use of subject pronoun
vos ‘you2nd.singular.informal’, instead of tú (e.g. Carricaburo, 2013; Fontanella de
Weinberg, 1968, 1979; M. Johnson, 2016; Siracusa, 1972), and a distinctive lexical
stock by the name of “lunfardo”, which incorporates lexical items from languages in
the Iberian and Italian peninsulas, Portuguese, Guaraní, and Quechua, among other
languages (e.g. Conde, 2011; Gobello, 1989, 1996; Lorenzino, 2016).
With regard to the sound system, three aspects of BAS have garnered much
attention. Like many other dialects of Spanish, BAS weakens /s/ in coda position.
Broadly speaking, BAS falls in between s-deleting dialects (such as Andalucia and
Caribbean Spanish, and s-maintaining dialects like Mexican Spanish) (Hualde, 2014, p.
159). In BAS, /s/ is maintained in intervocalic position and before a pause, and it is
aspirated or deleted before a consonant, both word-internally as word-finally (Bybee,
2000; Hooper, 1981; Terrell, 1978).
BAS is also well-known for its palatal system (or lack thereof) (Harris & Kaisse,
1999). In BAS, as in most Spanish dialects, the palatal lateral, /L/, and the palatal
obstruent, /J/, have merged. In Hispanic linguistics, this process is called yeísmo.
BAS underwent two additional processes. First, the phoneme /J/ became a voiced
post-alveolar sibilant, /Z/ (zheísmo; Fontanella de Weinberg, 1973). The second
process took place during the twentieth century and started with the working class:
the devoicing of /Z/ to [S], also known as sheísmo (Fontanella de Weinberg, 1978; Wolf
8
& Jiménez, 1979). Recent work has examined the sociolinguistic details of the latter
process (Chang, 2008; Rohena-Madrazo, 2013, 2015).
One final aspect of BAS phonetics and phonology that has garnered attention
from linguistic research relates to its intonation, which is well-known by its pronounced
falling contour of statements (Colantoni & Gurlekian, 2004; Fontanella de Weinberg,
1966, 1980, inter alia). Kaisse (2001) describes this “long fall” as a high tone on the
most prominent syllable of a phrase combined with a fall to a low tone within that
same syllable, or an H* + L pitch accent. Of note with this contour is that the fall
occurs on the salient syllable, even when followed by unstressed syllables, and the
salient syllable may be exaggerated in duration, up to five times the length of the
surrounding syllables.
1.3.2
Santo Domingo Spanish
Much like with BAS and Argentine Spanish, descriptions of Santo Domingo Spanish
(henceforth, SDS) are conflated with the the broader term ‘Dominican Spanish’. Thus,
in this section, we offer an overview of Dominican Spanish, with special reference to
SDS where available. But before we do so, we will take a brief tangent to situate SDS
within the context of the Dominican Spanish.
Several sources (e.g. Lipski, 1994) cite Henríquez Ureña (1940) and Jiménez Sabater
(1975) as the most complete descriptions on Dominican Spanish. These descriptions
divide the Dominican Republic into three dialectal areas, illustrated in Figure 1.2: (a)
Santo Domingo, and the area surrounding the Distrito Nacional, (b) the Cibao region
and (c) the eastern tip of the island.
Some of the linguistic features that linguists have paid special attention to include
frequent use of overt subjects (Alfaraz, 2015; Jiménez Sabater, 1975; Martínez-Sanz,
2011; Otheguy, Zentella, & Livert, 2007) and the lack of subject-verb inversion in
9
declarative, interrogative and infinitival constructions (Bergen, 1976; Heap, 1990;
Jiménez Sabater, 1975, inter alia).
Within the sound system, a distinguishing feature among Dominican dialects is
the realization of syllable-coda liquids. As an example of dialectal areas, the map in
figure 1.2 indicates the distribution of variation in coda-liquids. With regard to the
rhotic phoneme, speakers of SDS realize coda /R/ as [l], such that the name Arturo is
pronounced as A[l]turo, unlike in other Dominican regions where liquids vocalize or
are retained (Alba, 2000, p. 20, Lipski, 1994, p. 239.) Núñez Cedeño (1980) found
that lambdacism is socially stratified, with the lower and middle class producing the
[l] variant more often, and in spontaneus speech, across all socioeconomic levels.
Figure 1.2: Map of dialectal areas in the Dominican Republic based on the production of
coda liquids (Jiménez Sabater, 1975)
10
Another distinguishing feature of SDS relates to the weakening or lenition of
syllable-final /s/. Unlike BAS, which also exhibits weakening of /s/, the preferred
realization in SDS is deletion. However, studies comparing across speaker groups
and tasks (e.g. Alba, 2016; Núñez Cedeño, 1980; Terrell, 1982, 1986) have found
that deletion and aspiration are socially stratified as well, with younger and lowersocioeconomic level and informal tasks favoring /s/ reduction.
Another aspect of the syllable-final /s/ in SDS that has been mentioned in
the literature relates to instances of hypercorrection, that is, cases where speakers
hypercorrect deletion and insert /s/ in places where /s/ is not part of the underlying
representation. One of the findings in Terrell (1982) and Terrell (1986) is that /s/insertion is more common among speakers with a lower level of education and in
conversational tasks (as opposed to tasks that involve reading). These results have led
researchers to suggest that for illiterate speakers lexical items no longer contain any
trace of coda-/s/. In other words, that they have lost /s/. However, recent research
with data from the Cibao region has shown that the phenomenon is not as pervasive
as had been reported in the earlier literature (Bullock, Toribio, & Amengual, 2014)
Crucial to the study undertaken in this dissertation, another segment that is
subject to variation in SDS is /n/, which can be realized as a velarized variant ([N])
or a deleted variant in word-final position. Because chapter 3 presents a detailed
overview of this literature, for now I will only note that SDS, in the same vein as
other Caribbean dialects, shows a tendency to produce word-final nasal consonants as
velar and sometimes even deleted.
Having established some basic notions in the study of nasality as well we introduced
the dialects under study, the next sections lays out the goals and contributions of this
doctoral project.
11
1.4
Goals
While much research has examined phonological variation in Spanish, from labo-
ratory and sociolinguistic perspectives, most studies focus on one variety and a narrow
array of segmental phenomena. In sociolinguistics research, coda-/s/ has by far been
the star of the show. In experimental phonology, rhotics and the lenition of voiced
stops have stolen the limelight. Extending this research on sound variation in change
in Spanish to less studied domains and across dialects of Spanish, extends empirical
coverage of the phonological phenomena examined in Hispanic sociophonetics and
diversifies methodological approaches and parameters measured in the study of sound
change. One such domain is nasality. To this end, the present study was designed
with the following goals in mind:
1. Contribute to the body on Spanish laboratory on sound variation by providing
an instrumental examination of the production of anticipatory vowel nasalization
and word-final nasal consonants.
2. Provide a detailed acoustic characterization of the aforementioned phonological
variables by:
(a) comparing the realization of anticipatory vowel nasalization and word-final
nasal consonants in two dialects of Spanish, and
(b) identifying the linguistic factors that influence and best describe patterns
of variation.
To accomplish these goals, this dissertation follows an experimental approach
to examine the production of anticipatory vowel nasalization and word-final nasal
consonants by native speakers of two dialects of Spanish. Data was collected within
12
the dialectal regions, in Santo Domingo (Dominican Republic) and Buenos Aires
(Argentina). The findings of this study highlight linguistic and social factors that
influence variation in the production of nasality for native speakers from different
Spanish-speaking regions. In the next section, the contributions are detailed.
1.5
Contributions
The study reported here bridges together research in very different subfields
and approaches in the study of language variation, including phonetics, phonology,
sociolinguistics and dialectology, and historical linguistics. As such, it makes contributions in several areas in linguistics. Perhaps the most immediate contribution of
this dissertation is to Spanish dialectology and Spanish phonetics and phonology. As
emphasized previously, development of nasalized allophones in Caribbean dialects has
been reported in the literature but never instrumentally examined. This dissertation
provides experimental confirmation to the patterns attested in the early sociolinguistics
literature.
By analyzing both the production of vowel nasality and nasal consonants, this
dissertation also contributes to Spanish phonetics and phonology. Much has been
said about how nasalized allophones fit in a continuum of lenition. This has been
the approach of most phonological treatments on the topic (for an overview on
these, the reader is directed to Goodin-Mayeda, 2016). Experimental work, on the
other hand, has investigated one or the other, and the preference for which one is
due to the research questions under investigation, to methodological limitations, or
both. To provide an example, Solé (1992) investigates the production of nasality
by means of nasography, but the analysis does not include the production of nasal
consonants as the goal of this research was to contrast the phonetic underpinnings of
13
allophonic (English) and phonetic (Spanish) nasalization. Colantoni and Kochetov
(2012) and Ramsammy (2011, 2013), on the other hand, employ EPG to analyze
place of articulation and degree of constriction of word-final nasal consonants, but the
methodology prevented the researchers from addressing vowel nasality. The research
reported in this dissertation combines the analysis of anticipatory vowel nasalization
and word-final nasals, and in doing so, permits the discussion of not only dialectal
differences in the production of each acoustic feature, but also, co-variation between
them.
Extending research to incorporate laboratory methods to the study of dialectal
variation is also meaningful to the field of sociolinguistics more broadly. Laboratory
approaches contribute to the study of linguistic variation in a social context. In fact, one
of the reasons why research in the domain of nasality has lagged behind may be related
to the fact that studying larger populations, outside a laboratory setting, and within
the dialectal area, is difficult. In order to characterize the production of anticipatory
vowel nasalization and word-final nasal consonants this dissertation employs nasometry.
Unlike other instruments used to examine the production of nasality, whether in nasal
consonant or nasal(ized) vowels (for example, electromiography, EPG, or nasography
in the articulatory literature), nasometry lends itself very nicely to fieldwork. Thus,
this dissertation provides an innovative method for data collection. And while this
doctoral project presently examines regional differences, it lays the groundwork to
examine differences across speaker groups, such as educational level, socioeconomic
class, gender and sexual identity.
Additionally, as will be explained in greater detail in chapter 4, nasometry lies
between acoustics and articulation. Traditional acoustic analyses are not reliable
methods to capture different degrees of nasal constriction and anticipatory vowel
nasalization. Given that nasometry provides separate yet simultaneous signals for
14
the oral and the nasal tract, the approach undertaken here convincingly captures
the gradient nature of the variants under analysis, while at the same time allowing
non-invasive and field-work friendly data collection procedures. As such, another
contribution of this dissertation relates to acoustical analysis of production data in
nasality.
Another important characteristic of this study is the examination of multiple
linguistic variables (vowel height, stress and environment). In addition to demonstrating how nasality differs between dialects, these variables were selected in light of the
literature on coarticulatory vowel nasalization and on universals of vowel nasalization
(see sections 2.3 and 2.4 in chapter 2). The elicitation of multiple linguistic variables
help us uncover fine-grained variability in both Spanish dialects, while confirming
some well-established cross-dialectal differences. The insights we gain allow us to
design in the future more controlled and systematic investigations of the variation
within and across dialects of Spanish, as well as sociolinguistic groups.
Finally, this dissertation constitutes a first step in a research agenda that seeks
to examine the relationship between production and perception in phonetic and
phonological variation. By incorporating perceptual protocols, this larger research
program will contribute to the growing body of work examining the connection
between the production and the perception system. But before perceptual work can
be undertaken, the production signal has to be understood. More details with regard
to future research are presented in chapter 6.
1.6
Structure of the thesis
This dissertation is organized as follows. In chapter 2, I present an overview of
the phonetics of nasals and nasalization. This chapter begins with a articulatory and
15
acoustic description of nasal(ized) segments, and discusses co-articulatory vowel nasalization. It ends with an overview of the literature on evolution of vowel nasalization.
Chapter 3 presents a review of work examining word-final nasals and anticipatory
vowel nasalization in Spanish. This chapter draws specifically from the sociolinguistics
literature on the realization of nasal consonants and recent experimental work that
offer evidence for the reports of phonologization of nasalization. An introduction to the
general methodological approach of this dissertation leading to the experimental study
is presented in chapter 4. Important demographic information about the 58 speakers
that were recruited for this dissertation is given as well as detailed information on
how the data was processed, measured and analyzed. The results of the study are
presented in chapter 5. Chapter 6 discusses the findings and provides further insight
on their linguistic and theoretical implications. Finally, chapter 7 summarizes the
main findings of this dissertation, and also offers possibilities for future research.
16
Chapter 2
PHONETICS OF NASALS AND NASALIZATION (AND SOME
HISTORICAL LINGUISTICS)
Nasal consonants and vowels have had a prominent place in phonetics research.
The exact way in which nasality is phonetically realized not only depends on whether
nasality is phonolological or not, but also on other features within the segment,
such as vowel height, and on the combination with other segments in speech (i.e.
coarticulation). One major key to understanding the behavior of nasal(ized) segments
is understanding how they are produced. To this end, this chapter opens with an
overview of the articulatory characteristics of nasal segments (section 2.1) followed
by a survey of the acoustic consequences of nasality (section 2.2). In section 2.3, we
discuss co-articulatory vowel nasalization. We close the chapter with a review of the
literature on diachronic development of vowel nasalization, in section 2.4.
2.1
Articulation of nasal segments
Nasal segments are articulated by a combination of the movement of the tongue
and lips in the oral cavity and lowering of the velum. This section focuses on the second
movement. We first discuss the physiology of the nasal cavity and velopharingeal port,
and we then review the velophayngeal port mechanism.
Figure 2.1 presents an illustration of the vocal tract. To produce nasal sounds,
17
the air has to flow upward into the nasopharynx and into the nasal passages, through
the velopharyngeal port, the opening that separates the mouth from the nose. Thus,
during the production of nasal segments (consonants or vowels), two cavities are
involved: the oral and the nasal cavity. What allows or prevents the connection
between them is the action of the velum. The velum, the soft muscular rear part of the
roof of the mouth, can raise to close off the nasal cavity from the rest of the vocal tract
(e.g. when we are swallowing food) or it can lower and allow air to go through the
nose (e.g. when we are breathing.) These two settings –lowered and raised– are used
in the production of speech sounds and determine the binary distinction between oral
and nasal sounds. Before discussing the mechanisms that implement the oral/nasal
distinction, I will review the physiology of the parts involved: the nasal cavity and
the muscles of the velum.2
Figure 2.1: Midsagittal view of the the vocal tract.
When the velum lowers, the velophraryngeal port is opened, and the nasal
tract connects with the oral cavity. Multiple muscles are involved in closing the
velopahryngeal port. However, there is some disagreement, especially in the earlier
18
literature, as to whether and to which extent, particular muscles are involved in raising
and lowering the velum. The levator veli palatini is widely considered as the primary
muscle in raising the velum (Bell-Berti, 1976, 1993; Fritzell, 1969; Lubker, Fritzell,
& Lindqvist, 1970). It connects the front surface of the velum with the skull. When
it contracts, it pulls the velum to close the velopharyngeal port, whereas when it
relaxes, the velum is lowered and leaves the velopharyngeal port opened. Two other
sets of muscles are involved in raising the velum. The pharyngeal constrictor, whose
main function is to propel food down toward the oesophagus, raises the velum when
contracted by drawing the lateral and posterior pharyngeal walls inward. The tensor
palatini muscle, on the other hand, spreads and tenses the velum as it is raised.
Lowering of the velum has been linked to the action of the palatoglossus muscle.
The palatoglossus muscle connects the velum to the tongue. It lowers the velum,
or raises the back of tongue, or both. However, there is contradictory evidence
regarding the specific role of this muscle. Some research suggests that opening of the
velopharyngeal port is the result of the relaxation of the muscles involved in raising
the velum –such as the levator veli palatani– and that there is limited activity of
the palatoglossus (Bell-Berti, 1976, p.239; Bell-Berti & Hirose, 1973, p. 204).3 In
fact, activity of the palatoglossus was not found to be as strong as when swallowing
and breathing (Lubker et al., 1970, pp. 18-19). Other research shows activity of this
muscle during opening of the velum, which suggests that lowering of the velum is not
just the result of muscle relaxation and gravity.4
A well-known phonetic factor influencing nasalization is the co-variation between
tongue height and velum height via the palatoglossus. With (oral) low vowels, the
hyoglossus muscle, which is active during the production of low vowels, pulls the
tongue body down; the tongue body, in turn, pulls down on the palatoglossus; and
the palatoglossus pulls down on the velum. As a result, low vowels such as [A] exhibit
19
some incidental nasalization, as opposed to high vowels like [i] and [u], which would
provide physiological evidence concerning why the observed pattern of low vowels
exhibiting greater extent of nasality (Gick, Wilson, & Derrick, 2012, p. 136; Zsiga,
2013, pp. 86-87).
Figure 2.2: Anterior, posterior and lateral view of tongue muscles.
Research has addressed differences in nasalization across vowel height (Clumeck,
1976; Kuehn & Moon, 1998; Rochet & Rochet, 1991, among many others). When
phonetic correlates, such as duration of nasalization, degree of velic opening and
velopharyngeal closure force are examined, low vowels are the most nasal. For
example, Clumeck (1976) examines the correlation between vowel height and degree
20
of nasalization in six languages (American English, Swedish, French, Amoy Chinese,
Hindi and Brazilian Portuguese) by measuring velic movement with a nasograph.5 For
five of the six languages, low vowels were the most nasal in terms of size of velic opening
and duration of vowel nasalization, possibly because of an already low velum position
and inherently longer duration of the vowel. Kuehn and Moon (1998), additionally,
examined velopharyngeal closure force and levator palatini muscle activity by means
of a force sensing bulb and electromiography,6 and arrived at similar conclusions
–speakers exhibited greater velopharyngeal closure force with high vowels than with
low vowels, though differences in activity of levator palatini muscles across vowel types
did not reach statistical significance. Kuehn and Moon concluded that while speakers
activated the levator palatini muscles for all oral vowels, high vowels exhibited tighter
closure. In acoustic studies this trend was also borne out: Diakoumakou (2004), who
examined anticipatory and carryover nasalization in Modern Greek, found that low
vowels were temporarily more nasalized than high vowels.
However, variation across vowel types appears to be more nuanced, as exceptions
to these trends have been found. In Clumeck (1976) some speakers and some languages
(e.g. Hindi) did not exhibit the co-vartiation between velum and tongue height. At the
same time, some differences either did not reach statistical significance or presented
great inter-speaker variation (e.g. French). Additionally, when other phonetic cues
are examined, such as nasal airflow and sound pressure levels, the opposite trend is
found –high vowels are the most nasal (Al-Bamerni, 1983; Rochet & Rochet, 1991).
At first blush, the articulatory mechanisms that facilitate greater velum lowering
during the production of low vowels provide physiological confirmation to patterns of
phonologization of nasality (see section 2.4 for a review of the diachronic literature).
However, this literature also suggests that the inverse relationship between vowel height
and nasalization does not hold across the board and that, in addition to variation as a
21
function of linguistic environment, patterns of co-articulatory vowel nasalization may
also be language-specific. These are addressed in section 2.3.
2.2
Acoustics of nasal segments
We now move on to discuss the acoustics of nasal segments. The addition of
a side-chamber complicates the pattern of resonances and complicates the acoustic
analysis of nasal consonants and nasal(ized) vowels. Much work has been done with
the aim of identifying the spectral characteristics of nasal segments. A detailed account
of the acoustics of nasality would fall beyond the scope of this dissertation. However,
in order to contextualize the present doctoral project, this section presents a summary
of these findings. See Huffman and Krakow (1993, and references therein) for more
detailed treatments.
The two cavities involved in the production of nasal segments have to be modeled
by means of a two-tube model, illustrated in Figure 2.3. One tube represents the nasal
cavity, and is closed-open: it begins at the closed glottis and ends in the nasal cavity,
which is opened at the nostrils. The other tube represents the oral cavity, and can
be viewed as a side chamber. With nasal consonants the side chamber is closed at
the place of articulation of the consonant (e.g. the alveolar ridge for /n/), whereas
for nasal(ized) vowels is open. The acoustics of nasal consonants are presented in the
next section, and in the following, the acoustics of nasal(ized) vowels.
22
Figure 2.3: Two-tube model of the vowel tract
2.2.1
Nasal consonants
During the production of nasal consonants, the oral airflow is blocked by means of
a complete constriction in the oral cavity. The sound produced by the glottal action
of phonation is propagated through the velar port and the nasal passages, and escapes
through the nose. This sound is called the nasal murmur.
Nasal murmurs are known for a number of spectral characteristics. The nasal
cavity is not flexible, and does not change in size or shape. The nasal passages remain
constant for different nasal consonants, and for this reason, the murmur spectrum
does not differ greatly between nasal consonants. Nasal murmurs have a first formant,
also referred as the nasal formant, that is very low in frequency, due to the longer
length of total cavity, and low in amplitude.
The interaction, or coupling, between the nasal and oral cavities results in the
presence of anti-resonances in the spectrum. The source component (i.e. the vocal
folds) causes resonances both in the nasal and oral cavities. The resonance frequencies
in the nasal cavity that are close to the resonances in the oral cavity are canceled (or
absorbed) from the overall spectrum. Consequently, the acoustic output has frequency
23
regions in which the amplitudes of the source component are severely attenuated.
These anti-formants, or zeros, show up in the spectrum as valleys.
In theory, spectral characteristics function as acoustic cues for place of articulation
(Fant, 1960; Fujimura, 1962). The frequency of the nasal formant and oral zeros are
inversely related to the volume of the cavity, as determined by the position of the
tongue in the oral cavity. A more forward articulation (e.g. /n/) will produce a larger
cavity volume. By contrast, a more retracted articulation will produce a smaller cavity
volume, and larger frequency values for the nasal formant and oral zeros. However,
as the previous statements reveal, spectral characteristics depend on the volume of
the sinus and the dimensions of this opening, and research has shown assymetries in
the nasal cavity and large individual variation regarding the morphology of the nasal
cavity (Dang, Honda, & Suzuki, 1994; K. Johnson, 2012, p. 193). It is no surprise,
then, that work seeking to identify place-specific characteristics of the nasal murmur
have yielded inconsistent results (e.g. Fujimura, 1962; House, 1957; Recasens, 1983; for
Spanish: Albalá, 1992; Borzone de Manrique, 1980; García & Rodríguez, 1997). For
illustration purposes, Table 2.1 below presents murmur values found in the literature.7
2.2.2
Nasal(ized) vowels
The vocal tract filtering that characterizes the production of nasal segments also
has implications for the acoustics of vowels. One consequence of lowering the velum is
that, compared to an oral vowel, nasal(ized) vowels show greater F1 bandwidth. With
nasal(ized) vowels, there is greater surface area that absorbs more sound and, as a
result, F1 bandwidth is greater in a nasal(ized) vowel than in an oral vowel.
Much like nasal consonants, the resonances in the nasal cavity become antiformants in the spectrum of the nasal(ized) vowels. Typically nasal(ized) vowels
present three spectral prominences below 1000Hz: an oral formant (F1), a nasal
24
Table 2.1: Frequency values for murmurs reported in the literature. “N1”-“N4” represents
first through fourth nasal formants, while “NZ” the first anti-formant.
Study
Recasens (1983)
Language
Catalan
Borzone de
Manrique (1980)
Spanish
Albalá (1992)
Spanish
García and
Rodríguez (1997)
Spanish
Consonant
[m]
[n]
[ñ]
[N]
[m]
[n]
[ñ]
[m]
[n]
[ñ]
[m]
[n]
[ñ]
N1
200
250
250
350
300
370
380
480
480
480
298
318
296
N2
1120
850
1025
1200
900
950
1320
1052
1052
1004
1369
1612
1616
N3
N4 NZ
1360 2100
1550 2025 1780
2100 3125 2650
2030 2540 3700
2000
2200
2320
2240
2400
2338
2468
2580
formant and an anti-formant. These prominences interact with each other in a
complex way. The frequency of F1 varies as a function of the position of the vowel
height: high vowels exhibit a low F1 whereas low vowels a high F1. With the antiformant, the frequency depends on the degree of coupling between the nasal and
oral cavities. That is, how much nasalization is present in the signal. With slight
nasalization, the anti-formant is higher than the nasal formant, so the former may
cancel the latter. With heavy nasalization, the frequency of the anti-formant will
be higher and will cancel the F1. At the same time, the addition of nasality in the
signal causes the F1 to present a wide bandwidth. As a result, the nasal formant
becomes more prominent, and thus, leads to misperception of the acoustic signal.
Figures 2.4 and 2.5 below, taken from Beddor, Krakow, and Goldstein (1986), present
transfer functions for /i/ and /A/ that illustrate the interaction between spectral
prominences with oral and nasal(ized) high and low vowels (for details, the reader
is directed to this seminal paper.) With high vowels, when nasality is added to the
signal, additional spectral energy appears at a higher frequency than the original oral
25
F1 (Figure 2.4). As a result, a nasalized high vowel is perceived as lower, compared
to its oral counterpart. For low vowels, on the other hand, the first spectral peak is
lower than in an oral vowel, which can potentially cause low vowels to be perceived
as higher than their oral counterparts (Figure 2.5). Thus, this complex interplay
between components of the spectrum may cause the hearer to perceive a change in
vowel height, and judge as similar different nasal vowel types.8
Figure 2.4: Vocal tract transfer functions for oral, moderate nasalized and heavily nasalized
/i/ taken from Beddor et al. (1986).
26
Figure 2.5: Vocal tract transfer functions for oral, moderate nasalized and heavily nasalized
/A/ taken from Beddor et al. (1986).
We turn next to the acoustic measurements of nasality that have been proposed in
the literature. Table 2.2 summarizes a battery of measurements to examine the effect
of nasality on the formant/spectral structure. These measurements characterize the
slope and width of the spectral prominences, discussed in the above paragraph. They
are taken at different time points throughout the nasal(ized) vowel and compared
to their oral counterparts (see Diakoumakou, 2004; Scarborough, Styler, & Zellou,
2011; Zellou & Tamminga, 2014 for examples of research using these measures.)
More recently, Pruthi and Espy-Wilson (2007) propose another set of nine acoustic
parameters that describe sprectral flattening, but in the interest of space and scope,
only the ‘traditional’ measurements are presented here.
Another parametrized way to quantifying nasality comes from the clinical literature
with nasalance, which provides an estimate of the relationship between oral and
nasal energy in a vowel. Excessive nasal resonance in speech is one of the typical
characteristics of individuals with cleft palate. Clinicians need to asses the degree to
27
Table 2.2: Summary of acoustic measurements of vowel nasality
Amplitude of the first formant (A1) minus amplitude of the first
harmonic (H1). Given that nasalization increases the bandwidth of F1,
A1 - H1 so the difference will be greater for oral vowels than for
nasal(ized) vowels (Huffman, 1990). This method has been replaced
by other methods (Thomas, 2010, p. 243)
Amplitude of F1 minus (A1) the amplitude of first nasal formant (P0
A1 - P0 –often lower than F1; M. Y. Chen, 1995, 1997). The larger the
difference, the less nasality in the signal.
Amplitude of F1 (A1) minus the amplitude of the second nasal formant
(P1 –between F1 and F2; M. Y. Chen, 1995, 1997). The larger the
difference, the less nasality in the signal. This measurement is easier to
A1 - P1
use than A1 - P0 because with high vowels the nasal formant (P0) can
be difficult to distinguish from the oral F1
(which is low) (Thomas, 2010, p. 243).
The first peak minus the second peak. With [ĩ], the spectral spread
between the two peaks is captured as the distance between F1 (N1)
and the nasal formant (N2). With [ũ], on the other hand, the first peak
N1 - N2
(N1) represents F1 and the second peak (N2), F2. The higher the
difference, the greater the degree of nasal coupling
(Maeda, 1993, p. 160).
which their speech is nasalized in order to decide on an intervention scheme. Nasalance
is obtained by recording oral and nasal signals separately with a specialized device,
the nasometer, and computing the proportion of nasal to total energy. Nasometry is
the approach taken in the current study, though we will not be employing nasalance
as a measure of nasaliy. More details will be presented in chapter 4, when the
instrumentation and analysis are discussed.
Taken together, the aforementioned acoustic parameters provide spectral indicators
of nasality that allow quantifying the degree or the magnitude of nasal coupling.
However, a segment can be more nasalized in terms of the temporal domain; that is,
in terms of the time-course of nasality. Detecting the onset/offset of nasality through
spectrographic analysis has been challenging (Fant, 1960). Thus, acoustic analysis
presents a challenge for describing the temporal domain of velum lowering (and, thus,
28
nasality) and how it overlaps with the articulation of adjacent sounds (though see
Diakoumakou, 2004). The different timing patterns, and how they inform the research
reported here, are discussed in the next section.
2.3
Coarticulatory vowel nasalization
Speech sounds are part of a sequence of articulations, and there is overlap, or
coarticulation, with nearby sounds. Co-articulatory vowel nasalization, the focus of
this study, is spatially and temporally more extensive in some environments than in
others. It is in environments where nasalization is more extensive that phonologization
of nasality may begin. Some of these environments have been mentioned earlier in this
chapter: low vowels present greater extent of nasalization than non-low vowels. In this
section, we discuss how timing of velum lowering may be affected by non-segmental
influences, such as position within the syllable or stress, as well as cross-linguistic
differences.
Not all nasal consonants are created equal: syllable-final nasal consonants present
lower positions of the velum than syllable-initial nasal consonants (Henderson, 1984;
Krakow, 1989). Work by Krakow (1989, 1993, 1999) additionally shows that timing of
gestural movement, oral and velic, varies across syllable positions. For word-final /m/,
there is close temporal proximity between the end of velum lowering and the end of
lip raising. That is, the velum lowers as the lips initiate the constriction. With initial
nasals, the velum goes down when the lips release the constriction. These findings
indicate that with word-final nasal consonants velum lowering is greater and takes
place earlier than with word-initial nasal consonants. As a result, nasalization is more
likely to affect vowels preceding a word-final nasal than those following word-initial
nasal consonants. These findings provide articulatory facts for the diachronic changes
29
that have been observed, and which will be discussed in the next section, section 2.4.
Co-articulatory vowel nasalization is also susceptible to the effects of stress.
Vaissiére (1988, p. 128) finds that velum lowering is greater in stressed than in
unstressed environments. However, the two speakers in the sample differed in terms
of velum movement velocity. For one speaker, the maximum slope of the downward
movement of the velocity of velum lowering was greater with stressed syllables than
with unstressed. The other speaker showed no such contrast among environments.
Krakow (1993, p. 103) also found that stress effects can be different for different
speakers. Recall that low vowels have been shown to exhibit greater degrees of nasality
(in terms of magnitude and time-course) that non-low vowels. In Krakow (1993),
stress enhanced differences in velic position between low and high vowels (i.e. velum
lower for /A/ than for /i/), for one speaker. The other participant showed a lower
velum with stressed syllables than with unstressed syllables, regardless of vowel height.
It appears, then, that stress and vowel height have a combinatory effect, with low
stressed vowels exhibiting more nasalization than high stressed, high unstressed or low
unstressed vowels. In the same vein, Krakow (1999) finds that the velum lowers earlier
when the nasal consonant is in the stressed syllable (however, note that this difference
was not statistically significant for coda nasals, but it was significant for onset nasals.)
In addition to variation due to environment, co-articulatory vowel nasalization
varies across languages. These cross-linguistic differences provide evidence as to the
phonetic underpinnings of the phonological interpretation of vowel nasality. There are
different phonetic requirements for contrastive and non-contrastive effects of vowel
nasalization, as shown by Clumeck (1976); Cohn (1993); Rochet and Rochet (1991).
Cohn (1993), for example, examines the phonetic realization of the feature [nasal] in
French, American English and Sudanese pre-nasal vowels. Crucially, vowel nasality is
prosodic in Sudanese, distinctive in French and allophonic in American English. Given
30
the scope of this doctoral project, I will focus on the English and French data in Cohn
(1993), which show a very clear difference between pre-nasal vowel phonemes that
are [+/-nasal]. Cross-linguistic differences surfaced not only as differences in the
timing of velum lowering, but also in terms of the rapid or gradient increase of vowel
nasalization over the course of the segment. Unlike phonemic nasal vowels, pre-nasal
vowels in French exhibited no significant airflow for most of their duration. Pre-nasal
oral vowels (e.g., /O/ in the phrase bonne tête /bOnt(Et)/ ‘good head’) showed a lack
of nasal airflow, whereas phonemic nasal vowels (e.g. /O/ in a word such as bonte
/bÕte/ ‘goodness’ or a phrase like bon nez /bÕn(e)/ ‘good nose’) presented significant
nasal airflow for most or all of its duration, with a rapid increase of nasal airflow and
plateau-like time-course. In American English, nasalization increased throughout the
duration of the segment. Pre-nasal vowels (in words such as dean /din/ and den /dEn/)
presented nasalization thoughout its duration, in a cline-like pattern (i.e. gradient).
Cohn concluded that English does not specify pre-nasal vowels for [nasal], and
therefore, English vowels receive no target, nasal or oral, and anticipatory nasalization
in the vowel is the result of co-articulation.
At first blush, the take-home message in Cohn (1993) is that when a language’s
vowels are not specified for the feature [nasal], anticipatory vowel nasalization is more
gradient. However, ‘phonological’ in Cohn (1993) is taken to be distinctive nasality
–nasal vowel phonemes– and not allophonic nasalization –nasalized vowel allophones.
These findings do not allow for the possibility that non-distinctive nasality (that is,
nasality in non-nasal vowels) can be either allophonic or phonetic. As previously
discussed, phonetic nasality is the result of co-articulation with an adjacent nasal
consonant, and in allophonic nasalization the vowel is intended as nasalized. Languages
may exhibit one of two types of [-nasal] vowels, those that in a pre-nasal environment
present allophonic nasality, as in the case of English, and those that present phonetic
31
nasality, like Spanish. Given the cross-linguistic results in Cohn (1993) and elsewhere
(e.g. Clumeck, 1976), one may expect that the behavior corresponding to allophonic
and phonetic nasality will be diverging as well. These are very much the findings in
Solé (1992, 1995), who compares nasality in Spanish and English.
Solé (1992) shows that for Spanish speakers9 the oral portion of the pre-nasal
vowel varied with speech rate; for Spanish speakers, velic lowering began at roughly
the same point in time before the onset of the nasal, regardless of how fast or slow they
were articulating. Thus, at faster rates, the percentage of the pre-nasal vowel that is
nasalized changes. Speakers of English, on the other hand, showed a different pattern.
In their case, the pre-nasal vowel was fully nasalized, regardless of speech rate. Velic
lowering, in this case, began at onset of the vowel. In a follow-up study with data from
the same pool of participants as in her 1992 study, Solé (1995) found that the velocity
of velum opening was unaffected in Spanish speakers, but it did vary for speakers of
American English, with faster speech rates showing higher velocity opening. The results
of these two studies lead Solé to conclude that Spanish speakers had an oral target,
and vowel nasality was the result of co-articulatory time constraints (i.e. phonetic
nasality). The American English speakers, on the other hand, systematically adjusted
the timing of anticipatory vowel nasalization and velocity of velar port lowering, which
suggested that the vowel was targeted as nasal (i.e. allophonic nasality).
In sum, anticipatory vowel nasalization varies within a language and across
languages. In terms of cross-linguistic variation, the phonological status of nasality
(distinctive, allophonic or phonetic) create different phonetic patterns. More to the
point, this work shows that there is language-specific organization of articulatory
gestures that account for cross-language co-articulatory variation. Extending this
viewpoint to consider how languages may phonologize nasality, we would expect that
if varieties of a language (regional or social) differ in terms of the phonological status
32
of nasality, they will exhibit differences in the phonetic realization of anticipatory
vowel nasalization, specifically in the time-course of nasal airflow. Spanish, thus, is
an ideal language with which to theorize about how co-articulatory overlap leads to
anticipatory vowel nasalization becoming allophonic since there is ample evidence
that some dialects of Spanish present extensive anticipatory vowel nasalization, which
has lead some researchers to claim that these dialects have developed nasalized vowel
allophones. The next chapter presents this body of work. But before, the next and final
section summarizes the attested patterns of diachronic development of phonological
vowel nasality.
2.4
Diachronic evolution of vowel nasalization
For distinctive and allophonic nasalization to occur, vowel nasality has to have
phonologized. In broad terms, phonologization refers to a situation in which a phonetic
property becomes phonological.10 Nasality can become phonological in two ways: a
language may acquire nasalized vowel allophones (i.e., allophonic vowel nasality), or it
may acquire nasal vowel phonemes (i.e., distinctive vowel nasality).
Descriptions of the phonologization of vowel nasality have mostly focused on the
historical development of distinctive vowel nasality, especially in Romance languages
(e.g. M. Chen, 1975; Foley, 1975; Hajek, 1997; Hombert, 1986, 1987; Lightner, 1973;
Ruhlen, 1973; Sampson, 1999; Schourup, 1973). The goal of this body of work has been
to identify universal patterns of change in the development of distinctive nasalization.
But the corpora used in these studies differ in terms of the languages sampled,
variability in size, quality and reliability, which lead to many disagreements with regard
to the proposed patterns and directions of change. Nevertheless, some observations
with regard to potential paths of variation and change involving nasalization can be
33
gleaned from this body of work.
There is general agreement that phonologization of vowel nasality is the result of
anticipatory (or regressive) nasalization in two stages.11 During the first stage, vowels
adjacent to a nasal consonant, in a sequence such as (C)VN, present some degree of
nasalization due to co-articulation. That is, in anticipation to the nasal consonant,
the velum lowers during the production of the vowel. As a result, a portion of the
vowel becomes nasalized. Over time, the very low level of nasality in the signal is
stepped up until the vowel is perceived as nasal rather than as oral. Thus, in stage 1
the vowel that is targeted as nasalized (VN > ṼN) and the perceived nasalization is
now a secondary cue to the upcoming nasal consonant. At a second stage of sound
change, the conditioning nasal consonant is elided and the nasality in the signal is
no longer associated with a nasal consonant. It is in this second stage where vowel
nasality has become distinctive (i.e., it has phonemized; ṼN > Ṽ).
There is less agreement with regard to the conditions that constrain sound change,
and the two stages are often conflated (see Hajek, 1997, for an notable exception).
Among the contraints that this body of work proposes there is much debate about
the effect of the quality of the vowel being nasalized. Some universalists, like M. Chen
(1975); Hombert (1986, 1987); Lightner (1973); Ruhlen (1973); Schourup (1972, 1973),
propose that vowel nasalization occurs preferentially in the context of low vowels
before spreading to mid and, finally, to high vowels. Hajek (1997), however, finds
little evidence of the diachronic interaction between vowel height and development
of nasality (see Hajek, 1997 for a review of cross-linguistic counter-examples to the
proposed effect of vowel height, pp. 123–125). Vowel backness, on the other hand,
has not been debated as much. Some authors do not make claims with regard to
its potential role in the development of nasalization (Foley, 1975; Lightner, 1973).
Others, Hombert (1987, p. 360), and Ruhlen (1973, pp. 11-12), claim that nasalization
34
develops first with front vowels and then spreads to back vowels. Schourup (1973, p.
192) considers an effect of back vowels, but only for carryover nasalization (i.e. NV
sequences). Finally, stress has also been claimed to have played a role as well. While
Foley (1975) found that distinctive nasalization developed in unstressed syllables, the
overwhelming majority observes the opposite trend: nasalization developed first in
stressed syllables (Hajek, 1997; Schourup, 1973), and some authors have related it to
the greater duration of stressed syllables (Hombert, 1986, 1987).
With regard to the nasal consonant, this body of work has examined consonant
weakening in the context of stage 2, i.e. loss of the conditioning nasal. In this literature,
nasal weakening surfaces as shifts in place of articulation12 and shortening of the
nasal consonant. With regard to changes in place of articulation, M. Chen (1973) and
Ruhlen (1973, p. 12) posit that all final nasal consonants gradually merge to [N] and
are finally lost in certain environments, in line with the literature on effacement of
Spanish word-final nasal consonants (e.g. Terrell, 1975 and Cedergren & Sankoff, 1975;
see section 3.2.1 in chapter 3). However, Hajek (1997, p. 192) challenges velarization
as a pre-requisite to deletion of the nasal consonant. Chen’s model fails to report for
dentalization and labialization of [N] in some dialects of Chinese (Zee 1985, cited in
Hajek, 1997). Additionally, such a model fails to account for develoments in languages
like English, where a velar nasal has never been claimed to develop in the process of
weakening and deletion of the conditioning nasal consonant –‘can’t’ [kãnt] > [kãn t] >
[kã:t], but not *[kãNt] or *[kãN] (Hajek, 1997, p. 65). In sum, while some languages
present velarized variants, velarization does not appear to be a universal process in
the phonologization of vowel nasalization.
Shifts in place of articulation aside, another aspect of nasal consonant weakening
that this body of work has noted, to a greater or lesser degree, is temporal reduction.
M. Chen (1973) and Foley (1975) refer to weakened variants and transcribe them as
35
[N ]. But rest assured, nasal supression is a gradual process. Beddor (2009) presents
experimental evidence of an intermediate stage in which the duration of the nasal
consonant is inversely related to the temporal extent of vowel nasalization: the shorter
the nasal consonant the more extensive the nasalization on the vowel. These results
were also obtained in perception as well. In an identification task, a second group of
participants were presented with stimuli (‘bet’, ‘bed’, ‘bent’ or ‘bend’), with varying
durations of nasal consonant and vowel nasalization. The results indicate that ‘bend’
and ‘bent’ were elicited more frequently if (a) nasalization in the vowel increased and
the duration of the nasal consonant decreased, or (b) nasalization decreased in the
vowel and the nasal consonat was longer. In other words, vowel nasalization and nasal
consonant duration were in a trading relationship in perception as well.
Hajek’s observations, and the experimental findings in Beddor (2009), raise the
question of whether there is a causal relationship between weakening of the nasal
consonant and the extent of vowel nasality. Hajek’s position is that phonologization
of vowel nasality is not dependent on weakening of the nasal consonant (he labels
his position “the V-NAS model”, as opposed to “the DEL-NAS model”) and that
each stage of sound change presents different contraints (Hajek, 1997). Such a view
would imply that phonologization of vowel nasality is not necessarily a compensatory
mechanism: while they function in tandem, nasal consonant effacement need not be
the cause of phonologization of nasality.
Furthermore, what the Hajek’s V-NAS model highlights the importance of examining both nasalization on the vowel and weakening of the nasal consonant, as this
hypothesis can be tested. Spanish, once again, provides an ideal language environment.
Below, I present statements in the literature in Spanish linguistics which unambiguosly
link nasal consonant weakening (including velarization) and vowel nasalization in
Caribbean dialects, to the exclusion of non-Caribbean dialects (emphasis my own).
36
“Nasalization also affects the perception of vowel quality (e.g. Ohala,
1974). Thus, it is also likely that varieties that have different processes
affecting nasals (e.g. higher rates of velarization) also differ in the phonetic
characteristics of the vocalic system.” (Colantoni, 2011b, p. 26)
“In most dialects that exhibit velarization this process shows some variation,
usually with the alveolar nasal [n] or elision coupled with nasalization of
the preceding vowel, but sometimes the nasal is totally effaced with no
traces on the preceding vowel; this occurs most frequently in final atonic
syllables.” (Lipski, 2011)
“Diachronically, a common source of phonemic nasalized vowels is the
weakening of syllable-final consonants, leaving nasalization of the preceding
vowel as residue (this is the origin of the nasalized vowels in French and
Portuguese). This path of development is observable in Caribbean and
Andalusian dialects where a word like /pán/ ‘bread’ may be produced as
[pãŋ] ~ [pã].” (Hualde, 2005, p. 123)
The next chapter delves into the Spanish nasal system and expands on this
literature before presenting the research protocol.
2.5
Summary
All in all, the literature reviewed in this chapter shows that the phonetic underpin-
nings of anticipatory vowel nasalization vary cross-linguistically, depending on whether
vowel nasality is phonetic, allophonic or distinctive. Additionally, anticipatory vowel
nasalization varies as a function of features within the segment (e.g. vowel quality)
as well non-segmental factors (e.g. stress). Environments that are prone to more or
37
earlier nasalization facilitate the sound change involving nasalization. Some of the
environments reported in the phonetics literature (e.g. vowel height and stress) find a
correlate in the historical literature and suggest that they may modulate variation in
the synchronic signal as well.
38
Chapter 3
WORD-FINAL NASALS AND NASALIZATION IN SPANISH
Unlike other Romance languages, Spanish does not present an oral-nasal contrast
for vowels. The consonant system, however, includes three nasal phonemes that
contrast by place of articulation: the bilabial nasal (/m/), the alveolar nasal (/n/),
and the palatal nasal (/ñ/). The distribution of nasal phonemes in Spanish is said to
be ‘defective’; they are contrastive only in syllable-initial position, as in the words
ca[m]a ‘bed’, ca[n]a ‘white hair’ and ca[ñ]a ‘sugar cane’. The bilabial and alveolar
nasals appear word-initially and word-medially. The palatal nasal has a more limited
distribution: it occurs word-medially and rarely word-initially, usually with loan words,
such as ñame ‘yam’ or ñandú ‘rhea’. In the syllable-coda, the most common nasal
consonant is the coronal /n/. Other phonemes, such as /m/, may be also found
word-finally, but only with learned words (cultismos) like referéndum ‘referendum’ or
with recent loanwords, such as álbum ‘album.’
The coda coronal /n/ exhibits considerable contextual and regional variation,
and for this reason has garnered the attention of much research, from formal, variationist and experimental perspectives (Formal: Baković, 2000; Colina, 2009; Harris,
1984a, 1984b; Hualde, 1989a, 1989b; Piñeros, 2006, 2014; Ramsammy, 2013; Variationist: Cedergren & Sankoff, 1975; D’Introno & Sosa, 1988; Haché de Yunén, 1981;
López Morales, 1980; Michnowicz, 2008; Terrell, 1975; Experimental: Colantoni &
39
Kochetov, 2012; Honorof, 1999; Kochetov & Colantoni, 2011b; Ramsammy, 2011,
2013).13 In terms of contextual or allophonic varitation, in the word-internal syllable
coda and across word-boundaries, /n/ undergoes place assimilation in preconsonantal
environments. Nasals are homorganic with the consonant that follows. For example,
informe /infóRme/ ‘report’ is articulated as [iMfóRme], whereas ángel /anxel/ ‘ángel’
as [áNxel]. As a result, in the syllable-coda, the alveolar nasal has six pre-consonantal
allophones, presented in Table 3.1 (Hualde, 2014, p. 173).
Table 3.1: Preconsonantal Spanish nasal allophones with examples in orthographic form.
Articulatory description
Example
bilabial nasal
labiodental nasal
dental nasal
alveolar nasal
palatalized alveolar nasal
velar nasal
ca[m]po ‘field’
e[M]fásis ‘emphasis’
a[n
”]tena ‘antenna’
a[n]sia ‘anxiety’
a[nj ]cho ‘wide’
ma[N]go ‘mango’
Word-finally, /n/ exhibits a wide range of regional variation. Dialects of Spanish
can be categorized as ‘preferring’ a coronal or a velar realization. Dialectological
compendiums of Spanish have proposed the realization of word-final nasals as one of
the features that distinguish Spanish dialects (Canfield, 1981; Lipski, 1994; Resnick,
1975).14 It is in dialects with production of the velar variant that allophonic nasalization
has been reported (Vaquero, 1996; Colantoni, 2011b; Campos-Astorkiza, 2012; Hualde,
2014; Lipski, 2011; see section 3.2.1), and has been linked to patterns of variation
attested in the evolution of Romance languages (D’Introno, del Teso Martín, & Weston,
1995, p. 310; Lipski, 2011, p. 73). The remainder of this chapter presents an overview
of the historical evolution of the Spanish nasal system, in section 3.1, followed by work
investigating synchronic variation in Spanish word-final nasals from variationist and
experimental perspectives, in sections 3.2.1 and 3.2.2, respectively.
40
3.1
Historical development of the Spanish nasal system
An overview of the development of the Spanish nasal system shows that sound
changes have affected the consonantal system and the distribution of nasal phonemes.
The three-way contrast in Modern Spanish nasals has its origin in two singleton and
two geminate bilabial and alveolar nasals in Latin (m, n, mm and nn, respectively).15
One important difference between the phonemic inventories of Spanish and Latin
is the number of contrasts based on place of articulation. While Latin had bilabial
and alveolar nasal phonemes, the Spanish inventory includes bilabial, alveolar and
palatal. At the same time, we also observe a difference between paradigms in terms of
length. Latin included singleton and geminate phonemes, of which only singletons are
included in the Spanish inventory. In the interest of clarity, this section focuses on the
development in the syllable coda domain, though table 3.2, at the end of the section,
summarizes the development of nasal consonants in Modern Spanish.16
Latin allowed both singleton phonemes, m and n, in the syllable-coda. As the
reader may recall, in Spanish only the alveolar nasal survived in this domain. The loss
of the declension system and the simplification of verbal morphology conspired in the
loss of word-final m during the first century BC. Other times, word-final /m/ underwent
apocope, but mostly with words with a preceding /e/; words such as mărginem were
reduced to margen ‘margin’, with word-final /n/. And when /m/ was retained —in
frequent monosyllables— place of articulation shifted to alveolar, e.g. cum became
/con/ with ‘with’. A different fate was reserved for the Latin word-final alveolar nasal:
it was either retained or the intervocalic alveolar nasal became word-final in Spanish
through the loss of a final vowel, usually /e/.
41
Table 3.2: Summary of the historical development of Spanish nasal consonants.
Latin
Onset #m-m-mm#n-n-nn-n+j-gnCoda
Modern Spanish
/m/
Example
mare >mar ‘sea’
remu >remo ‘oar’
flamma >llama ‘flame’
noviu >novio ‘fiancé’
cena >cena ‘dinner’
/n/
ø
annu >año ‘year’
vinea >viña ‘vineyard’
ligna >leña ‘timber’
iam >ya ‘now’
/n/
cum >con ‘with’
marginem >margen ‘margin’
/ñ/
-m#
-n#
-n-
in >en ‘in’
pane >pan ‘bread
Given the impact of nasality in the development of the vowel system in other
Romance languages, we conclude this section by briefly commenting on what we know
about nasality and the historical development of the vocalic system in Spanish. In
Spanish, nasality did not exert a major influence in the development of the Spanish
vowel system. The vowel inventory in Latin included long and short midvowels, which
not only differed in terms of duration, but also in terms of ‘aperture’ (i.e. height).
Long midvowels were realized as [e:] and [o:], whereas short midvowels as [E] and [O]
(Penny, 2002, p. 46). In French, one context that evidences the role of vowel nasality
in the evolution of the vowel system was the neutralization of midvowel pairs /e/-/E/
and /o/-/O/ before a nasal coda consonant. Spanish shows no such neutralizing effect
—/E O/ underwent diphthongization and therefore did not merge with /e/ and /o/ .17
See the examples in (1), where the word to the left includes a long midvowel (/e/ and
/o/), whereas the word to the right, includes a short vowel (/E /O/). These examples
who that in nasalizing contexts, short and long vowels did not merge into a single
42
phoneme and suggest that nasalization did not exert the same effect in Spanish, as it
did in French, in the development of the vocalic system (Sampson, 1999, p. 168-169).
(1)
a. vēndit >vende ‘s/he sells’ vs. tĕnda >tienda ‘store’
b. corōna >corona ‘crown’ vs. bŏna >buena ‘good’
Despite these observations, we do find some exception to the evidence presented
above. Some words actually have escaped the diphthongization explained above,
presumably due to a syllable-final nasal that raised /O/ to /o/. For example, the
modern reflex of mŏntem is monte ‘mountain’, and not *mounte (Menéndez Pidal,
1962, §13). Additionally, Sampson (1999) presents additional evidence that suggests
that nasality may have played a role in development of Spanish vowel system.18
There were two phonological contexts where high vowels developed in Spanish when
midvowels would have been expected, and suggest that nasality may be responsible for
vowel raising. Consonant clusters with nasals is the first phonological context. Before
the cluster [N] + velar consonant + consonant, we find that Spanish retains the high
vowel (see the example 2) . With the sequence [N] + velar consonant, we observe a
midvowel, as expected (see 3).
(2)
a. pŭnctum >punto ‘point’
b. pĭnctare >pintar ‘to paint’
(3)
a. trŭncum >tronco ‘trunk’
b. fŭngum >hongo ‘mushroom’
The second context where high vowels were developed in a context where a mid
vowel was expected involves the sequence -nj-. At a prior stage to the palatal nasal,
the midvowel raised to a high front vowel or a palatal glide, which in turn led the way
43
to the palatal nasal (see examples in 4). Taken together, what the examples in 2 and
4 suggest is that heightened nasality may have played a role in these outcomes.
(4)
a. tĭnea >tiña ‘ringworm’
b. cŭnea >cuña ‘wedge’
In sum, lenition and assimilation contributed to the development of the Spanish
nasal system. Most sound changes in the nasal system are attested in the consonantal
domain. Interestingly, the Spanish word-final nasal has originated not only from
Latin coda nasals (/m n/), but also from onset nasals. Recall that the diachronic
literature (see section 2.4 in chapter 2) agrees in that VN# (as opposed to NV) is the
preferred environment for phonologization of vowel nasality. Thus, the development of
word-final nasals in Spanish is important development because the number of words
that presented VN# environments increased in the evolution from Latin to Spanish.
If these new pre-nasal environments exhibit greater co-articulatory overlap, then the
way to phonologization of vowel nasality has been paved.
3.2
Synchronic variation
Traditional descriptions of Spanish phonetics and phonology detail the phonemic
inventory in Spanish and provide idealized descriptions of their phonetic realization
(e.g. Navarro Tomás, 1980[1918]). The aim of these accounts of Spanish phonetics
and phonology can be described as prescriptive, whether it is deliberate or not. One
of the overarching generalizations that these accounts make is that Spanish nasal
consonants “are produced with complete blockage of the airflow through the oral
cavity” (Hualde, 2005, p. 174). Any variation in the realization of these idealized
phonemes is represented as a deviation from a norm, often the speech of Castillian
educated speakers, and labeled as ‘regional.’
44
Navarro Tomás (1980[1918]) is an obligatory starting point for any account of
Spanish phonetics and phonology. Many traditional acccounts, such as Cárdenas
(1960); D’Introno et al. (1995); Malmberg (1965); Real Academia Española (2011),
echo his descriptions. Navarro Tomás (1980[1918]) acknowledges vowel nasality, but
emphasizes that it is not accompanied by loss of the nasal consonant (Navarro Tomás,
1980[1918], §38). Thus, productions such as [tãto] for tanto /tanto/ ‘so much’ are
evaluated as defective, and attributed to interference from a second language with
distintictive nasalization, such as French. One important observation in Navarro Tomás
(1980[1918], §100) is the mention of [N] as a regional variant of /n/, which shows that
this variation is by no means recent –in fact, Cline (1972, p. 77) observes velarized
variants in sixteenth century Andalusian Spanish.
Other descriptions of the Spanish phonemic inventory take a different route. Quilis
and Fernández (1973) extend the oral-nasal binary distinction to the vowel system as
well. For the authors, Spanish vowels can be oral, with a raised velum, or oronasal,
which surface as allophones between two nasal consonants, e.g. mano /máno/ [mãno]
‘hand’, and in word-initial position followed by a tautosyllabic nasal consonant, e.g.
insaciable /insasiáble/ [ĩnsasjable] ‘unsatiable’ (Quilis & Fernández, 1973, pp. 53-54).
This approach recognizes the effect of co-articulation. However, whether they interpret
anticipatory vowel nasalization as allophonic or phonetic is not explicit in the text.
Also not clear in the text is how, if at all, nasal consonant lenition contributes to the
production of nasalized vowels.
Thus, a reader of these phonological descriptions of Spanish will likely be left with
the impression that nasal consonants, unlike voiced obstruents for example, are not
subject to weakening patterns. While this may be true of many varieties of Spanish,
the remainder of this section presents the research that challenges these descriptions,
at least for some varieties of Spanish. The common thread running through the work
45
in the next section is that nasal weakening (and compensatory vowel nasalization) is
prevalent in some dialects of Spanish.
3.2.1
Variationist studies
Velarized realizations of word-final /n/ have been reported since early dialectological work in Latin American, as well as in some Iberian varieties of Spanish
(Quilis, 1965; Robe, 1960; Salvador, 1957; Zamora, 1985[1965]). However, it was
early variationist sociolinguistic work which made observations of nasalized vowels.
Variationist work has focused exclusively in the syllable-coda domain (word-internal
and word-final) and on dialects of the Caribbean region. An important contribution
of this body of work has been to describe the envelope of variation and the frequency
of occurrence of variants across speech communities. One of the recurring findings is
that the velar variant is often the preferred variant among Caribbean dialects, and
that nasal consonants are prone to weakening (c.f. Hualde, 2005).
Sociolinguistic investigations have uncovered a range of variants that authors
have analyzed as part of a continuum of lenition. This body of work shows that the
(standard) alveolar nasal ([n]) alternates with a velarized variant ([N]), and a deleted
variant (ø), often accompanied by nasalization of the preceding vowel ([Ṽ]). Table 3.3
below presents a summary of the envelope of variation reported in the sociolinguistic
literature. These variants have been analyzed as steps in a continuum of lenition with a
‘fully’ occluded alveolar nasal consonant on one end of the continuum, and the deleted
variant with nasalized pre-nasal vowel in the opposite end. The underlying implication
at work is that nasalization is the result of weakening of the nasal consonant, and as
such a compensatory mechanism. These observations parallel some of the observations
in the historical literature presented in chapter 2 and follow the DEL-NAS model that
Hajek (1997) criticizes.
46
Table 3.3: Variants of syllable-coda (word-internal and word-final) nasal consonants in the
sociolinguistic literature.
Study
Variants
Domain(s)
Terrell (1975)
Word-internal
and word-final
Cedergren and
Sankoff (1975)
Word-final
López Morales
(1980)
Word-internal
and word-final
Haché de Yunén
(1981)
Word-internal
and word-final
D’Introno and
Sosa (1988)
Word-final
Symbol(s)
Description
[n]
alveolar
[N]
velarized
[Ṽ]ø
deleted with nasalization
j
[n
” nmMn]
assimilated
[N]
velarized
ø
deleted
j
[n
” nmMn]
assimilated/standard
[N]
velarized
ø
deteled
[n
” n m M nj ]
assimilated
[N]
velarized
[Ṽ]ø
deleted with nasalization
ø
deteled with no nasalization
[n]
alveolar
[N]
velarized
[N]
weakened velarized
[N]
neutralized (i.e., assimilated)
[m]
bilabial
[Ṽ]ø
deleted with nasalization
ø
deteled with no nasalization
Another recurring finding in this body of work is that nasal consonant weakening
is more prevalent word-finally than word-internally, though findings are inconsistent.
Table 3.4 summarizes the information with regard to frequency of occurrence of variants
according to each prosodic domain in a number of study. The symbol “>” means
“greater frequency than”. Relevant to the present dissertation, Haché de Yunén (1981)
finds the same ordering of variants in terms of their frequency of occurrence in the
word-internal and word-final domains for Dominican Spanish (velarization being more
frequent than deletion). However, there is a difference between the word-internal and
the word-final domains with regard to the rate of application of [Ṽ]ø. The results in
47
this study indicate that word-internally, the rate of [Ṽø] is 19.5%, whereas word-finally
climbs to 44.1%.
Table 3.4: Frequency of syllable-coda nasal variants per domain in the sociolinguistics
literature.
Study
Domain
Frequency of variants
Terrell (1975)
Word-internal
Word-final
[assimilated]>[Ṽø]
[N]>[Ṽø]
Cedergren and Sankoff (1975)
Word-final
[ø]>[N]>[assimilated]
Word-internal
Word-final
Word-internal
Word-final
[assimilated]>[ø]>[N]
[assimilated]>[N]>[ø]
[N]>[Ṽø]>[assimilated]
[N]>[Ṽø]>[assimilated]>[ø]
Word-final
[N]>[N]>[N]>[n]>[m]>[Ṽø],[ø]
López Morales (1980)
Haché de Yunén (1981)
D’Introno and Sosa (1988)
In addition to variability as a function of syllable structure, the patterns of nasal
deletion are subject to the phonological environment (i.e., the preceding and following
segments). The early sociolinguistic work has examined rates of deletion as a function
of the phonological environment: pre-consonantal ( C), pre-vocalic ( V) or prepausal ( P), summarized in Table 3.5. The findings, however, present diverging
results. Some studies (Cedergren & Sankoff, 1975; Haché de Yunén, 1981; Terrell,
1975) indicate that deletion is more frequent in pre-consonantal position, whereas
other studies report that the favoring environment is pre-pausal (D’Introno & Sosa,
1988; López Morales, 1980). The pattern for pre-vocalic nasals is not clear. Nasal
effacement in this environment is reported as less frequent (Cedergren & Sankoff, 1975;
Haché de Yunén, 1981) or as frequent as with a following consonant (Terrell, 1975).
48
Table 3.5: Frequency of deletion of word-final nasals per phonological environment in the
sociolinguistic literature.
Study
C
V
P
Terrell (1975)
39%
38%
38%
Cedergren and Sankoff (1975)
74%
58%
69%
López Morales (1980)
6.2%
7.4%
8.1%
Haché de Yunén (1981)
47.9%
44.1%
24%
D’Introno and Sosa (1988)
1%
1.6%
3.3%
One of the main contributions of this body of work are the observations of nasal
weakening in the syllable coda in the Spanish of the Caribbean region. The story here
is that nasal effacement in Spanish is coupled with nasalization of the preceding vowel,
especially in the word-final domain.
Despite the important contributions that these studies have made to our understanding dialectal variation with word-final nasals, they point to potential avenues for
research that will expand our understanding of nasal consonant lenition and nasalization processes in Spanish. Firstly, sociolinguistic work has not examined dialects
outside of the Caribbean region, and as a result, our understanding of variation in
the realization of coda-nasals in dialects where the preferred coda-nasal is alveolar is
limited. Additionally, the impressionistic analyses that sociolinguistic studies employ
has the disadvantage of relying on the perceptual skills and biases of the researcher.
The further back the oral constriction is during the articulation of the nasal consonant
(i.e. velar nasal vs. alveolar nasal), the more likely the nasal segment will resemble
a nasalized vowel (for details, see Ohala & Ohala, 1993). As a result, the researcher
may have mistakenly reported velarization instead of nasalized vowels, and viceversa.
Additionally, in many cases, the researchers were not native speakers of Spanish or the
49
dialect under analysis, which also may have biased some of the reported frequencies.
Recent research bypases these shortcomings by utilizing instrumental methodology.
Additionally, they capture important differences in occlusion magnitude according to
contextual factors, as well as phonetic differences between phonetic and allophonic
vowel nasality. The next section presents an overview of this body of work in Spanish
phonetics and phonology.
3.2.2
Instrumental studies
As with all research, instrumental studies present advantages and disadvantages.
The undeniable advantage is that instrumental research is able to quantify phonetic
processes accurately. In doing so, this work has uncovered patterns of variation that
were otherwise unnoticed to the naked ear. As advantageous as this is, it is also true
that these studies often use speakers that may or may not be representative of the
Spanish variety under study, as they often take place in a university setting and in
English-speaking communities. These issues do not diminish their contribution, but
in order to contextualize their findings, I will detail participant information whenever
available.
Instrumental research has examined the production of word-final nasals, vowel
nasality and the perception of nasal-oral vowels provide show patterns in support of
the findings of the sociolinguistics literature reported in the previous section. And
unlike variationist research that has only focused on Caribbean Spanish, instrumental
work has also addressed non-Caribbean dialects. This section presents the findings in
this body of work.
50
3.2.2.1
Word-final nasal consonants
Three recent studies examine the production of word-final nasal consonants by
means of Electropalatography (Colantoni & Kochetov, 2012; Ramsammy, 2011, 2013,
EPG; ). EPG monitors the degree of contact between the tongue and the hard-palate
during the articulation of speech. For data collection, participants are fitted with
a custom made artificial palate that contains electrodes on the lower surface that
makes contact with the tongue. When the tongue makes contact with the (artificial)
hard palate, the electrodes activate. The electric signal is recorded and is used for
visualization in real-time of the location and timing of tongue contact against the
palate.
EPG has been an invaluable method in research examining place of articulation of
nasal consonants, as acoustic data is often unreliable for this type of work. Colantoni
and Kochetov (2012) and Ramsammy (2011, 2013) each compare two dialects of
Spanish, one representing a velarizing dialect, and the other one representing an
alveolarizing dialect. These studies have shown that a given phonological process —in
this case, neutralization— can be implemented in different ways by, not only different
speaker groups, but also speakers within one group (see also Honorof, 1999; Kochetov
& Colantoni, 2011b).
In Ramsammy (2011, 2013) data from native speakers of two Peninsular Spanish
dialects, Galician and Castillian Spanish, is compared. All participants were in their
thirties and residents in Manchester, UK at the time of data collection. Galician
Spanish speakers were males, from Vigo (Galicia) and Monforte de Lemos (Galicia),
while Castillian Spanish speakers were a male and a female, from Albeos (Galicia)
and Manzanares (Ciudad Real). The study was designed to compare six neutralizing
environments, of which three are of particular interest to the present dissertation:
51
nasals in word-final pre-vocalic environment (/-N#V/), nasals in absolute phrase-final
position (/-N##/), and nasals in word-final pre-obstruent position (/-N#C-/).19
The findings indicate that speakers of Galician Spanish produce word-final nasals
with dorso-velar constriction, whereas the speakers of Castillian Spanish, with a
linguo-alveolar constriction. Additionally, results indicate across- and within-regional
variation in patterns of assimilation and magnitude of occlusion during co-articulation.
One Castillian Spanish speaker (speaker identified as A1) realized word-final nasals
consistently as alveolar and presented full closure of the nasal consonant. The fact that
there was no evidence of dorsal realizations, for example in N#k, was interpreted by
the researcher as indicating that A1 did not present signs of assimilation of the nasal
consonant to the phonological environment. The other Castillian Spanish speaker (A2),
on the other hand, does present evidence of gestural blending. For A2, in N#t and
N#k, EPG showed closure throughout the full duration of the nasal consonant, with
place of constriction varying as a function of the following consonant: linguo-alveolar
contact in N#t, and linguo-velar in N#k.
Galician Spanish speakers (speakers V1 and V2) presented a different pattern.
With N#p, EPGs showed progressive loss of contact during the realization of the nasal
consonant. In other words, Galician Spanish speakers never reached maximum contact
during the nasal consonant. With N#t, the peak of linguo-alveolar contact was late
(that is, approaching the coronal consonant), which indicated co-articulation with the
following coronal obstruent. With N#k, the peak in linguo-velar contact aligned with
the nasal consonant and not with the following consonant, which indicated gestural
blending.
The comparison of pre-vocalic and phrase-final positions (/-N#V/ vs./-N##/)
also revealed differences in terms of degree of constriction, but only for speakers
of Castillian Spanish. Specifically, Ramsammy (2011, 2013) found that the non52
velarizing speakers, who consistently produce word-final alveolar nasals, presented
greater occlusion in prepausal than in prevocalic location.
Colantoni and Kochetov (2012) also compare the realization of the word-final
nasals, across three tasks, but in a different alveolar-velar dialect pair: in Argentine
Spanish and Cuban Spanish. Participants were three Argentine females and three
Cuban females, who ranged in age between 23 and 42 years-old, had university
education, and were residing in Toronto, Canada at the time of data collection.
Tasks included a carrier phrase task, reading of a passage and a story (a series of
pictures eliciting ‘Little Red Riding Hood’). Unlike Ramsammy (2011, 2013), only one
phonological environment was examined: VN#N environment, where V was stressed
/e a o/ in the carrier phrase task, stressed /i e o/ in the passage reading and stressed /e
a o u/ in the story task. For this last task, utterance final nasal (i.e. VN#) consonants
were also analyzed in this last task, and only data from two participants was analyzed
–one speaker of each dialect group.
Overall, and as expected, their findings indicate that Argentine Spanish speakers
produced more coronal, while Cuban Spanish more velar, word-final nasals. The
Argentine data presented less variability than the Cuban speaker data. The findings
indicated that both speaker groups produced weakened variants (i.e., less constricted)
in the less-controlled tasks (i.e. the story), but the degree of weakening was higher
for the Cuban Spanish speakers, who produced variants with incomplete closures or
‘vocalizations’. A comparison across tasks and environments revealed effects of the
task, context, vowel, and duration. With regard to context, the results showed more
weakening of word-final nasals in VN#V than in VN#. Between back vowels (i.e.,
/o N o/), word-final nasals were more retracted, less constricted and shorter, which
suggested incomplete closure or ‘vocalized’ realizations in this environment. In terms
of duration, there was a significant correlation between duration of the word-final
53
nasal consonant and degree of constriction for the Argentina data —the longer the
nasal consonant, the more constricted it was—, but this correlation was not significant
with the Cuban data, probably due to the variability observed in this dialect. Taken
together, these findings indicate that in Cuban Spanish the weakening process may be
more advanced than in Argentine Spanish (even among educated speakers who are
alledgedly more conservative). What is unclear, however, is the degree to which nasal
weakening was accompanied by nasalization in the preceding vowel.
All in all, these studies present articulatory confirmation of some production
differences between alveolarizing and velarizing dialects of Spanish. This work offers a
detailed account of the location and degree of constriction in the vocal tract and shows
that word-final nasals in both dialect groups weaken, though lenition may be greater
with velarizing dialects. However, due to the methods employed, we do not know the
details of how the velic gesture may operate in Spanish and how different dialects
may differ in terms of anticipatory nasalization. Though the research presented in the
next section does not specifically address this issue, it does provide some evidence of
phonetic differences in terms of vowel nasality between dialects of Spanish.
3.2.2.2
Vowel nasality
Chapter 2 showed that lowering of the velum before a nasal consonant results in
a certain degree of overlap between velic and vocalic gestures. An important generalization that can be gleaned from the literature on coarticulatory vowel nasalization is
that the degree of overlap between the realization of the vowel and the adjacent nasal
consonant is related to the phonological interpretation that vowel nasality has in a
given language. Some of the work presented in chapter 2 has provided experimental
confirmation that Spanish vowel nasality is phonetic (i.e. Solé, 1992, 1995). This
section presents additional work in Spanish that shows phonetic variation across
54
linguistic environments and that upholds the claims that some dialects of Spanish
may have progressed towards allophonic nasalization.
Fails (2011) is the only acoustic study (to my knowledge) examining Spanish
nasalization. This studies investigates how the patterns of nasality vary according to
the phonological status of nasality in a given language. He compares Spanish and
Portuguese vowels in a variety of phonological contexts adjacent to nasal consonants,
i.e. CV$N, #VN, NV$C, NV$N, CVN$C, NVN$C and VN#. For data collection, he
uses a nasometer, the same equipment employed in this dissertation. In addition to
showing that overall Portuguese presented higher rates of nasality, the findings indicate
that the production of nasality was gradient and that some phonological contexts may
be more susceptible to nasality than others. And unlike the trends reported in chapter
2, where there was an effect for vowel quality, it was high vowels which presented
higher rates of nasality. With respect to stress, stressed vowels present higher rates of
nasality. These findings can be taken to suggest that even a language with phonetic
nasality is prone to the patterns exhibited in language with phonological, allophonic
or phonetic.
Fails interprets these results as indicating that nasalized vowels are better described
as oro-nasal vowels, following Quilis and Fernández (1973). However, there are several
issues with the data presented in this study. These patterns represent the speech of
only one participant. Additionally, the presented data is the result of one repetition
per phonological context, though the methods section is confusing in this regard. Thus,
because the data belongs to one participant and one repetition, it is not clear how
these results may be generalizable across different participants or groups.
Experimental work on Spanish nasality has also uncovered dialectal differences:
Cuban Spanish has production patterns of nasality reminiscent of a language with
allophonic nasality. Lederer [2000, cited in Lederer 2003], who presents a three-way
55
comparison between English, Peninsular Spanish and Cuban Spanish, pre-nasal vowels
in Cuban Spanish pattern with other languages, such as (American) English, in terms
of the scope and the degree of nasality (c.f. Solé, 1992). However, all information with
regard to Lederer (2000) constitutes a brief paragraph in Lederer (2003). Thus, we have
no additional information to further evaluate the analysis and findings. The original
study only exists in hardcopy and it has been misplaced (Lederer, priv. comm.).
These studies provide experimental evidence that the phonetic underpinnings of
vowel nasality in Spanish differs from that languages such as Portuguese and English.
More importantly, they also provide experimental evidence that some dialects of
Spanish may have phonologized a pattern of pre-nasal vowel nasalization. One issue
with this research, however, is the small pool of participants. Thus, the extent to
which these patterns represent dialect groups is a question open to research. Another
layer to these (potentially) phonological differences in vowel nasality is related to the
effect these patterns have on perception. Recall that one of the effects of nasalization
is misperception of vowel height (see section 2.2.2 in chapter 2). It is well known
that the listener’s phonological inventory plays a role in the types of contrasts she
may or may not perceive (Strange, 1995). Thus, if vowel nasality has different
phonological interpretations, perceptual behaviour should reflect them. In other words,
an examination of perceptual patterns may provide evidence to phonologization claims.
The next section presents research on the perception of vowel nasality and what this
research may tell us about the phonologization claims.
3.2.2.3
Perception of oral-nasal vowels
Acoustic modeling indicates that nasalization adds an extra peak in the region
associated with vowel height, and as a result, vowel height is misinterpreted (House
& Stevens, 1956). Beddor et al. (1986), which was presented in chapter 2, provided
56
experimental confirmation of this acoustic modeling by testing if native speakers
of English were able to attribute nasalization to co-articulation with an adjacent
tautosyllabic nasal consonant. An important finding of this study was that listeners
judged vowel height accurately when nasality was present in the signal, as long as the
co-articulatory source (i.e. the nasal consonant) was presented to them. These results,
however, do not tease apart a language specific or a universal perceptual mechanism.
In other words, were the native speakers of English able to factor out the effects of
nasal coupling in a conditioning environment because nasality is allophonic in their
native language, and therefore part of the listeners’ grammar? Or was it due to a
universal mechanism?
Goodin-Mayeda (2011) followed up on these two possibilities and tested speakers
of Brazilian Portuguese and Castilian Spanish in a forced choice identification task with
three sets of stimuli: oral condition ([gus-gos]), contextual nasal condition ([gũns-gõns]),
and non-contextual nasal condition ([gũs-gõs]). To create stimuli, Goodin-Mayeda
used natural tokens of the endpoints [gos], [gus], [gons] and [guns] as the basis for
synthesizing, but it is not clear in the study whether native speakers of Spanish or
Portuguese (or a third language) provided the natural tokens. The results in this
study suggest that the ability to compensate for co-articulatory effects of nasalization
may be due to linguistic experience. The results indicate that, for Spanish speakers,
the rate of responses of /u/ was higher in both nasal conditions (i.e., contextual and
non-contextual) as compared to the oral condition. That is, they perceived the nasal
vowels as /u/ more often than their oral counterpart. For Portuguese speakers, on
the other hand, the perception of nasal vowels did not differ from that of oral vowels.
Therefore, the results of this study suggest that Spanish speakers were unable to undo
the effects of nasal coupling. That is, speakers of Spanish appear to be unable to
interpret the effects of nasalization as such, and instead, they interpret them as changes
57
in vowel height. Because of their linguistic experience, Portuguese speakers were able
to factor out the effects of nasalization. As a result, they accurately perceived the
height of oral and nasal vowels. Based on these results, the author concludes that
compensating for the acoustic effects of nasal coupling is not an universal mechanism,
but rather that it is connected to language experience.
In her 2016 book, Goodin-Mayeda expands the previous study by increasing the
participant pool and by adding a Cuban Spanish listener group, predicting that this
listener group would perceive vowel height with greater accuracy than Peninsular
Spanish listeners (Goodin-Mayeda, 2016). Methods were the same as in GoodinMayeda (2011). The results confirm earlier findings regarding differences betweeen
Portuguese and Spanish speakers, with the former participant group appropriately
judging vowel height in the presence of nasality. Interestingly, the results revealed
no statistical differences between the Spanish-listener groups, which contradicted the
predictions. Both Peninsular and Cuban Spanish speakers perceived /u/ more often,
even when there was a nasal consonant in the signal. The author attributes the
findings to linguists overreporting of nasalization for Caribbean dialects and calls for
more research to examine the issue. Nonetheless, it is important to consider how
certain methodological considerations in Goodin-Mayeda (2011, 2016) may account
for the results. One key issue has to with the use of synthesized speech. Typical
synthesis models, like the Klatt synthesizer used in these experiments, use LPC
analysis to reconstruct the signal, which is based on the assumption of a singleresonator tube. Nasal(ized) segments, however, can only adequately be modeled with
a dual-resonator tube. Thus, synthesizing and manipulating the nasality parameter
becomes an issue. As a result, it is not clear how the stimuli used in Goodin-Mayeda
(2011) and Goodin-Mayeda (2016) connect to real speech.
More recently, Martínez (2015) bypasses the issue of synthetic speech by using
58
natural tokens recorded with a Glottal Enterprises nasometer, the same that will be
used in this dissertation project (see chapter 4). Thus, nasality in the stimuli was
not synthesized but naturally produced. In this study, Martínez (2015) examines
the naïve perception of the Brazilian Portuguese contrast /i/-/ĩ/ by native speakers
of French, English, Caribbean Spanish and what the autor labels as ‘conservative’
Spanish (meaning non-Caribbean Spanish). None of these languages present the
contrast /i/-/ĩ/. The crucial difference between them is the phonological status of
vowel nasality: distinctive (French), allophonic (English and Caribbean Spanish) and
phonetic (‘conservative’ Spanish). The aim of this study was to test whether vowel
features, such as nasality, can be redeployed to create a novel vowel category following
the predictions of the Feature Model (C. A. Brown, 1998). The experimental protocol
was designed to test whether phonological differences between language groups in
terms of phonological status of nasality and syllable structure. The investigated
contrasts include: CV-CṼ, CṼ-CVN, CṼ-CṼN, CVN-CṼN, Cĩ-Cẽ and VĩN-CẽN. The
contrast /e/-/ẽ/ for comparison purposes.20 Due to the complexity of the research
design, a full account of the results falls beyond the scope of this chapter. For this
reason, Table 3.6 summarizes the findings for the English (EN), Caribbean (CS) and
‘conservative’ Spanish (SP) groups. The Caribbean Spanish listener group and the
‘conservative’ Spanish group only differ in one contrast (CV-CṼN), highlighted in the
table with bold typeface.
59
Table 3.6: Summary of findings in Martínez (2015).
Contrast
Vowel
CS
SP
EN
i
yes
no
no
e
yes
yes
yes
i
yes
yes
no
i
no
no
no
Statistical difference
between CS and EN
SP outperforms
EN and CS.
All groups are
e
Cĩ-Cẽ
no
yes
no
yes
no
yes
performing at chance level
No group exhibitied
CĩN-CẽN
yes
yes
yes
difficulty
CV-CṼN
CṼ-CVN
CṼ-CṼN
CVN-CṼN
Comments
No statistical
difference between groups
The fact that Caribbean speakers were able to discriminate /i/-/ĩ/, unlike the
‘conservative’ Spanish group, was interpreted by Martínez as providing evidence that
vowel nasality is allophonic in Caribbean Spanish, but phonetic in ‘conservative’
Spanish. In other words, nasality in Caribbean Spanish is part of the phonological
system, while it is a by-product of co-articulation for ‘conservative’ Spanish speakers.
Unexpectedly, however, Caribbean speakers did not exhibit similar patterns to English.
For the author, English speakers were not patterning like Caribbean Spanish speakers
in the perception of /i/-/ĩ/ due to differences in the nasal elision process of each
language: whereas in Caribbean Spanish the nasal coda can be elided syllable or
word-finally, in English the nasal consonant cannot be absorbed if it is not followed by
a tautosyllabic voiceless stop. These differences would suggest that while in Caribbean
Spanish a nasalized allophone can occur in open syllables, in English it is only licensed
in closed syllables. This may be a plausible interpretation, but one issue with Martínez
(2015) is that it examines naïve perception of non-native contrasts. The aim in this
study is to better understand the causes of L2 perception difficulties and how they
play out in setting up an L2 phonological system. Thus, it is not clear whether or
60
not we can take differences in perception of non-native categories as definite proof
of allophonic variants in the native language. At the same time, these post-hoc
explanations make the assumption that the extensive anticipatory vowel nasalization
reported for Caribbean dialects is the result of nasal consonant effacement and that
non-Caribbean dialects do not exhibit weakening.
3.3
Summary
The literature in Spanish linguistics reports that dialects in the Caribbean region
have developed nasalized allophones as a consequence of weakening of the nasal
consonant. In this body of work, velarization of word final nasals is conceptualized
as a form of weakening. Thus, the narrative presented in this research is that it is
in dialects with velarized variants where extensive vowel nasalization has developed.
One issue with this work is that it does not go across the aisle to examine patterns of
weakening and nasalization in non-velarizing dialects.
Recent experimental work does compare across dialects. But because of research
focus and/or methodological restrictions, either nasal consonant weakening or vowel
nasalization are examined. We are then left with several questions. How do dialects
with a preference for velarized nasal variants (e.g. Caribbean) compare to dialects
with alveolar variants (non-Caribbean) nasal consonant weakening and anticipatory
vowel nasalization? What is the relationship between anticipatory vowel nasalization
and nasal consonant weakening? Is the pattern for Caribbean dialects different from
the pattern for non-Caribbean dialects? These are the general wonderings at the
heart of this doctoral dissertation. The next chapter details the research questions,
hypothesis, and research protocol of this study reported here.
61
Chapter 4
THE RESEARCH PROTOCOL
The previous research described in chapters 2 and 3 motivates the need for further
investigations of anticipatory vowel nasalization and word-final nasal consonant weakening in Spanish. This work highlights that (a) anticipatory vowel nasalization is more
extensive in dialects with velar word-final nasal variants, such as Caribbean dialects of
Spanish, as opposed to non-Caribbean, which exhibit alveolar variants, and (b) that
deletion of the nasal consonant is a pre-requisite for anticipatory vowel nasalization.
The reported dialectal (and phonological) differences may be characterized in terms
of the time-course of nasality during the pre-nasal vowel, weakening of the nasal
consonant (i.e. degree of constriction and duration). At the same time, some work
has challenged descriptions where anticipatory vowel nasalization is the result of nasal
effacement (Hajek, 1997) . Thus, an examination of the relationship between the
time-course of vowel nasality and weakening of the nasal consonant is warranted. To
this end, an acoustic experiment was designed and carried out with speakers of Buenos
Aires (Argentina; BAS) and Santo Domingo (Dominican Republic; SDS) Spanish, an
alveolarizing and a velarizing dialect of Spanish respectively. The specific research
questions, and their corresponding hypotheses, for this study are as follows:
RQ1 Do SDS and BAS differ in terms of lenition of the word-final nasal consonant?
62
(H1) The SDS data will present greater degrees of word-final nasal weakening
than BAS (Colantoni & Kochetov, 2012). Thus, the word-final nasal
consonant will exhibit a higher ratio of oral energy and shorter durations.
RQ2 Do SDS and BAS differ in terms of the time-course of nasality?
(H2) SDS will present earlier onset of nasality, when compared to the BAs data
(Lederer, 2003; Solé, 1992). This scenario would indicate that SDS is further
advanced in phonologization of vowel nasality than BAS. That is, SDS is
progressing towards the ṼN stage (or possible even a ṼN stage), whereas
as Argentine Spanish is still in the VN stage.
RQ3 How do syllable type, environment, vowel type and lexical stress affect lenition
of word-final nasal consonants?
Syllable type
(H3) There is no evidence in the previous literature to hypothesize that CVN
or NVN will have an impact on the production of the word-final nasal
consonant. No differences are expected.
(H4) At the same time, we may encounter a repetition effect with NVN: we
might expect weakening of the nasal consonant in the syllable coda, when
there is a nasal consonant in the onset. The prediction here would be that
NVN syllables would exhibit higher consonant-to-vowel ratio and shorter
duration than CVN.
Environment
(H5) Work by Colantoni and Kochetov (2012) predicts that speakers of both
dialects will produce less lenited nasal consonants (longer and more con63
stricted) in the pre-pausal environment, than in the pre-vocalic environment. This was also the finding in Ramsammy (2011, 2013), though only
for speakers of the alveolarizing dialect (Castillian in Ramsammy, 2011,
2013, Argentine Spanish in the study reported here.)
(H6) In pre-consonantal position, there will be diverging results according to the
place of articulation of the following consonant (coronal vs. dorsal) and,
consequently, dialect group. If the nasal and oral consonants are produced
with the same articulator, there will be gestural blending. Argentine
speakers, who have been described as producing word-final alveolar nasal
consonants (Colantoni & Kochetov, 2012), will blend articulatory gestures
when the following consonant is coronal. For the Dominican speakers,
gestural blending should obtain when the following consonant is dorsal,
given that this dialect group is predicted to produce word-final velar nasals
(Colantoni & Kochetov, 2012; Haché de Yunén, 1981; Ramsammy, 2011).
With gestural blending, the prediction is that the likelihood of the nasal
consonant being fully realized is greater. With different articulators, the
prediction is that (oral) articulatory gesture for the nasal consonant will be
reduced (or completely deleted) and the gesture corresponding to the oral
consonant is fully extended. The prediction is that BAS will exhibit nasal
consonants with larger ratio of oral energy and shorter durations in the
pre-dorsal than in the pre-coronal condition. For SDS, on the other hand,
the opposite trend is expected: larger ratio proportions of oral energy and
shorter nasal consonants in the pre-coronal than in the pre-dorsal condition.
Vowel type
(H7) High vowels have been reported to favor deletion of the word-final nasal
64
consonant, whereas mid and low vowels inhibit it (Cedergren & Sankoff,
1975). Given that the corpus does not include high vowels, no difference is
hypothesized across vowel types.
Lexical stress
(H8) Word-final nasal will be sensitive to the effect of stress and there will be
greater tendencies to reduction in unstressed position, akin to Spanish
/s/-lenition (E. K. Brown, 2009). In unstressed position, nasal consonants
are expected to exhibit more weakening in terms of degree of constriction
and duration of the word-final nasal consonant.
RQ4 How do syllable type, environment, vowel type and lexical stress affect the timecourse of nasality?
Syllable type
(H9) CVC syllables, which serve as control stimuli, do not incorporate nasal
segments, and therefore will present no evidence of nasalization, neither in
the consonant nor in the tautosyllabic vowel.
(H10) If the vowel is not specified for [nasal] (as in Cohn, 1993), a preceding nasal
will cause the whole vowel to be nasal. In other words, there should not be
velum lowering during the vowel, as it takes place earlier in the consonant
and it is not raised again.
Environment
(H11) Any hypothesis with regard to anticipatory nasalization and environment
(pre-pausal, pre-vocalic, pre-consonantal) is the result of a model in which
nasalization is a compensatory mechanism. That is, we would expect
65
anticipatory vowel nasalization to be present an earlier onset if the nasal
consonant is weaker (see research question 3 above). Earlier onset of
the velum will take place pre-vocalic condition. In the pre-consonantal
condition, the prediction is that dialect and environment will interact. For
BAS, earlier onset will obtain in pre-dorsal position. For SDS, in pre-coronal
condition.
Vowel type
(H12) Nasalization will be greater with /a/ than with all other vowels (M. Chen,
1975; Clumeck, 1976; Hombert, 1986, 1987; Lightner, 1973; Ruhlen, 1973;
Schourup, 1972, 1973). As a result, the low vowel should exhibit earlier
onset of nasalization than the mid vowels.
(H13) At the same time, experimental evidence for different patterns of nasality
based on frontness and backness is inconclusive. For this reason, vowel type
will not have an effect in the production of nasality in the pre-nasal vowel.
Lexical stress
(H14) Nasality in the pre-nasal vowel will be greater in stressed syllables (Hajek,
1997, p. 99; Hombert, 1986, 1987; Krakow, 1999, p. 28; Schourup, 1973;
Vaissiére, 1988). The prediction is that stressed syllables will present the
velic gesture earlier than unstressed syllables.
(H15) The effect of stress will be more enhanced with some vowel types: the
stressed low vowels will present more nasality (Krakow, 1993, p. 10). Thus,
stressed low vowels will exhibit earlier onset of nasalization than unstressed
low vowels and mid vowels.
66
(H16) We can also hypothesize an interaction with dialect. If one dialect (i.e.
SDS) has phonologized nasality, then the effect of stress will enhance the
gesture and velum lowering will take place earlier for this dialect group.
RQ5 Is there a trade-off between the time-course of nasality and lenition of word-final
nasal consonants?
(H17) As the nasal consonant weakens, anticipatory vowel nasalization becomes
more extensive (Beddor, 2009). Given that dialects with velar variants have
been reported to show weakening of the nasal consonant, this relationship
is expected to obtain with SDS and not BAS.
The remainder of this chapter presents the research setting, participants, speech
elicitation tasks, experimental procedures and analysis of the data for the present
study.
4.1
Research sites
Data was collected in two research sites: Santo Domingo (Dominican Republic)
and Buenos Aires (Argentina). Figure 4.1 locates Santo Domingo (in red) and Buenos
Aires (in blue) within Latin America. The Dominican Republic is the second largest
Caribbean nation in area (only after Cuba) and third by population. It occupies
48,445 square kilometres and has a population of approximately 10.5 million people
(United Nations, 2015, p. 14). Santo Domingo is the capital city and one of the oldest
cities in the Caribbean region. It is the site of the first cathedral, castle, monastery,
and fortress built in all of the Americas, located in the Colonial Zone, an area declared
as a World Heritage Site by UNESCO (UNESCO, 2016). Including its metropolitan
area, population of Santo Domingo is estimated at 2,000,000 people (Oficina Nacional
de Estadística, 2015).
67
Argentina has a mainland area of 2,780,400 square kilometres and an estimated
population of 43.4 million people (United Nations, 2015, p. 13). Buenos Aires, the
capital city, has the largest popultation in the country, and is the fourth-most populous
city in the Americas. Including the greater Buenos Aires area, population of Buenos
Aires is estimated at 17,000,000 people. It is a top touristic destination: Buenos Aires
is the most visited city in South America and the second most in Latin America.
Figures 4.2 and 4.3 zoom into the Dominican Republic and Argentina, and locates
the capital city within each map.
N
0km
500km 1000km
Figure 4.1: Map of Latin America highlighting the location of Santo Domingo (red) and
Buenos Aires (blue).
68
N
0km
500km 1000km
Figure 4.2: Zoomed map highlighting the location of Santo Domingo (blue).
N
0km
500km
1000km
Figure 4.3: Zoomed map highlighting the location of Buenos Aires (blue).
69
While Santo Domingo and Buenos Aires differ in terms of their territorial and
population size, they were chosen as research sites because they constitute the cultural,
financial, political and commercial center of their country and, more generally, their
region. For this reason, the two research sites can be considered as comparable for the
purposes of this study.
4.2
Participants
4.2.1
Recruitment
Participants were recruited in two ways. Because I am originally from Buenos
Aires and have contacts in the city, in Buenos Aires participants were recruited using
the friend-of-a-friend method. I reached out to people who I knew taught at the
university level and asked them to forward an invitation email to their students. In
some cases, I was also invited to visit classes to recruit participants. Once participants
had been identified and contact initiated, I met individually with them. Recordings
took place at a quiet location of the participants’ choosing, most times in their home,
though in some cases it was at their workplace.
In Santo Domingo, I employed a different method for participant recruitment.
Unlike in Buenos Aires, I was not a local and did not have the same network of
contacts. For this reason, I reached out to a local university, Pontificia Universidad
Señora Madre y Maestra campus Tomas de Aquino (PUCMM-STA), to collect data.
In PUCMM-STA, participants were recruited via class visits, or by referral of other
participants. To conduct the recordings, the Department of Spanish at PUCMM-STA
facilitated a space.
While recruitment methods may seem to diverge, they are similar in two regards.
First, participants were recruited in an academic setting. Second, I had not met
70
participants previous to data collection. In both dialectal regions, I established contact
with participants through an acquaintance, be that one of my contacts in Buenos
Aires or the university staff/students in Santo Domingo. Thus, there was no previous
relationship between participants and myself at the time of data collection nor did
participants have any information about my research that I did not disclose myself
after the experiment was over.
4.2.2
Demographic information
Participants recorded in this study included 31 native speakers of Santo Domingo
Spanish and 28 speakers of Buenos Aires Spanish –all university students, such that
educational background could be consistent across dialectal areas. Due to issues with
recordings, five partipants were excluded from the sample (n = 2 from the BAS sample,
n= 3 from the SDS sample).
The Buenos Aires sample includes 18 females and 8 males. Age ranged between 19
and 28 for Buenos Aires Spanish speakers. Seventeen participants reported speaking
another language other than Spanish: French (n= 5), English (n= 19), Brazilian
Portuguese (n=5), Guaraní (n=1), Welsh (n = 1), Dutch (n= 1), Slovenian (n= 1)
and German (n= 1).
In the Santo Domingo sample, 18 speakers were female and 10 males. Age
ranged between 18 and 27. Some participants in this group are also multilingual,
ten participants reported speaking a language other than Spanish: French (n=3),
English (n=9) and Portuguese (n= 1). Additional demographic data can be found in
Appendix B on page 190.
71
4.3
Instrument
4.3.1
Stimuli design
Stimuli in this dissertation were real words. Nonce words would have allowed to
control nicely for phonological environment and for possible effects of word-frequency
and segmental-frequency, which has been important in previous work (e.g. Ramsammy,
2011). However, nonce words have the disadvantage of driving speakers into hyperarticulation. Because the equipment used in this dissertation (see Section 4.4) already
subtracts spontaneity from speech samples, it was important that the choice of stimuli
did not additionally encourage participants to be more careful in their speech than
they already were.
This issue was a concern when designing the study for two reasons. First, the
dialects under study differ in terms of the linguistic security their speakers exhibit
(see chapter 6 for a discussion on the topic and how it may have impacted the results).
Dominicans have been shown to evidence linguistic insecurity in relation to other
varieties of Spanish (Alba, 2004, 2009b; Büdenbender, 2010; Bullock & Toribio, 2014;
Toribio, 2000). Argentines, on the other hand, can be described (anecdotally) as
manifesting great linguistic security.
21
Second, I was part of the speech community
in one dialectal area (i.e. Buenos Aires), but an outsider in the other, which may
have led speakers in each dialect group to behave differently around me. For these
reasons, the use of real words was prioritized and, as a result, the stimuli presented
to participants did not present target segments in ideal contexts for segmentation.
Section 4.5.2 below details how data annotation was addressed.
Stimuli were designed to include words with a final syllable of the form: CVN#,
NVN#, NV#, where N= nasal consonant, V= vowel, C= oral (coronal) consonant, #
= word final. Additionally, CVC# and CV# were included as control environments.
72
Since the focus of this dissertation is on word-final nasals and pre-nasal vowels, only
environments for anticipatory nasalization –i.e. CVN#, and NVN#, and control
CVC#– are examined in the present study. Carry-over nasalization will be examined
in a future study. Table 4.1 presents example words of words included in the stimuli.
Table 4.1: Example words for phonological contexts under study
Phonological context
Example
CVN#
NVN#
CVC#
visiten ‘(may they) visit’
limón ‘lemon’
peces ‘fish (plural)’
The stimuli were also designed to explore the interaction with three additional
independent linguistic variables: vowel type (/e a o/), stress (tonic, atonic) and environment (pre-vocalic, pre-pausal, pre-consonantal [coronal, dorsal]). The independent
variable for vowel type explores non-high vowels due to lacunae in the Spanish lexicon.
Specifically, finding stressed and unstressed syllables with word-final /n/ preceded by
high vowels was difficult (with /i/) if not impossible (with /u/). Even though this has
been an important context in recent work in nasalizaiton in French and Picard (Dow,
2014), only non-high vowels are explored in the present study. Future research could
(and should) expand the contexts under analysis to consider nasalization with high
vowels.
With regard to the effect of environment, a few observations are in order. In
pre-vocalic and pre-consonantal environments, the word following the target was
a three-syllable paroxitone. In the pre-vocalic environment, the following segment
was unstressed /a/ in all cases (e.g. azul ‘blue’, alegre ‘joyful’) . With regard to
the pre-consonantal environment, recall that the dialects under study differ in the
preferred place of articulation of the word-final nasal consonant. As explained in
Chapter 3, in Dominican Spanish, the preferred word-final nasal is dorsal, whereas
73
for Argentine Spanish, coronal (Colantoni & Kochetov, 2012; Haché de Yunén, 1981).
In order to avoid glossing over dialectal differences in assimilatory processes, the
pre-consonantal environment includes two conditions: one where assimilation is to
a consonant that shares place of articulation features, and one where it is does not
–pre-coronal consonant (N#[t]) and pre-dorsal consonant (N#[k]). Finally, since velars
are known to present diverging articulations depending on the adjacent vowel (Clark,
Yallop, & Fletcher, 2007, p. 85, Ladefoged, 2005, p. 34), the dorsal stop was followed
by a low vowel, /a/, in all cases. That is, the syllable following the word-final consonant
in the pre-dorsal condition was unstressed /ka/. A full list of examples and carrier
phrases is presented in Appendix C on page 193, but examples are included in the
next section (see Table 4.2)
To summarize, stimuli were designed in order to examine four linguistic independent variables: target syllable (CVC# vs. CVN# vs. NVN#), vowel type (/e/ vs.
/a/ vs. /o/), stress (tonic vs. atonic), and environment (pre-vocalic vs. pre-coronal
consonant vs. pre-dorsal consonant vs. pre-pausal.)22
4.3.2
Elicitation task
Participants completed a carrier phrase task. Instead of ‘traditional’ carrier
phrases of the type Digo [target word] para ti ‘I say [target word] for you’, the
disseration employed NPs and VPs (NPs: determiner + noun + adjective; or VPs:
verb + NP or PP). ‘Traditional’ carrier phrases can be repetitive and tiresome for
participants, especially when large numbers of tokens are being elicited, and often
produce effects of ‘list reading’ when not varied. Table 4.2 presents examples of the
carrier phrases elicited.
The use of both NPs and VPs in the carrier phrase task, as opposed to only one
or the other, is rooted in the fact that in Spanish word-final nasals are not distributed
74
evenly among parts of speech. For example, Spanish words that end in unstressed -an,
-en and -(r)on are verbs, given that these sequences are verb 3rd person plural markers
(i.e., present indicative, present subjunctive, and preterite, respectively), whereas
stressed -on are nouns.
Table 4.2: Example carrier phrases
Target syllable
Example
pre-vocalic
Cuando tiraron azucar ‘When they threw sugar’
pre-coronal Cuando tiraron tabaco ‘When they threw tabaco’
CVN#
pre-dorsal
Cuando tiraron cadenas ‘When they threw chains’
pre-pausal
Digo que tiraron ‘I say they threw’
pre-vocalic
Pero ganan alegres ‘But they win happy’
pre-coronal
Cuando ganan terceros ‘When they win third’
NVN#
pre-dorsal
Siempre ganan cantando ‘They always win singing’
pre-pausal
Dicen que cocinan ‘They say they cook’
pre-vocalic
Los peces azules ‘The blue fish’
pre-coronal
Los peces torcidos ‘The twisted fish’
CVC#
pre-dorsal
Los peces carnosos ‘The fleshy fish’
pre-pausal
Los coloridos peces ‘The colorful fish’
4.3.2.1
Preparation of materials
In order to wash out potential list effects and avoid introducing systematic biases,
stimuli were randomized anew for each speaker. A macro was written for Powerpoint
in VBA code and was run three times, one per repetition. A total of 3 repetitions of
120 target words rendered 360 tokens of analysis per speaker.23 Each repetition took
approximately 6 minutes to read and was saved in its own sound file (a stereo .wav
file). Participants took a break in between repetitions.
Stimuli were displayed in Microsoft PowerPoint on the same laptop used in recording (a MacbookPro with a 13” screen). In order to accomodate to the participants’
native dialects, the PowerPoint had instructions using vos ‘you (singular)’ for the
75
Buenos Aires speakers, while tú ‘you (singular)’ for the Santo Domingo speakers.
Participants first saw a black cross, followed by the carrier phrase. Slides automatically moved forward after 2.5 seconds. Automated slide transitions liberated the
participants’ hand to hold the nasometer, and standardized across repetitions –in my
experience, when participants are allowed to self-time themselves, the second and
third repetions are much faster than the first one. In order to ease participants into
and out of the task, five non-target carrier phrases that were not randomized were
added at the beginning and end of the task.
Figure 4.4: PowerPoint sample slides with carrier phrases. On the left, the target word is
capitán ‘captain’; on the right, the target word is tiraron ‘they pulled’
Given that I was controlling for vowel, as well as adjacent segments, some of the
carrier phrases did not read naturally. For this reason, the carrier phrase task was
contextualized (Figure 4.5). Before stimuli were presented, participants were told
that Estanislao, a very peculiar and unsuccessful poet, had misplaced all of his notes
and writings (subfigures 4.5a and 4.5b). As a result, the words that make up his
poems had been jumbled up. It was the participants’ job to read out loud the verses
back to Estanislao so that he could take note of his own poems (subfigures 4.5c and
4.5d). While the carrier phrases were not any less unconventional, it was expected
that this contextualization would ease participants into the task. In fact, a number of
participants added a funny note in the background questionnaire (see below), when
prompted to add any final comments.
76
(a) ‘This is Estanislao. He is a clumsy and (b) ‘He’s never been a successful poet
excentric poet’
but he finally has the opportunity to publish’
(c) ‘The problem is that he wrote hundreds of(d) ‘And his poems are all scattered
lines, but... he’s lost all of his
Your job is to help Estanislao
notebooks and notes!’
rebuild his verses’
Figure 4.5: PowerPoint slides depicting context of the carrier phrase task, with translations.
4.3.3
Background questionnaire
Participants also completed a background questionnaire that elicited information
regarding (a) their native language and educational background and that of their
parents, (b) their foreign language background, (c) their residential history, (d) contact
with speakers of other dialects, and (d) their intuitions about the study. The responses
to the written questionnaire ensured that participants with diverging backgrounds
could be identified, as well as potential issues with the elictation task. Appendix A,
on page 183, includes the full questionnaire.
77
4.4
Experimental Procedure
4.4.1
Equipment
The data collected and analyzed in this study were chosen with the purpose
of conducting an acoustic analysis of vowel nasalization in Spanish. The choice
of acoustics (as opposed to articulatory or aerodynamic) was two-fold. First, my
aim beyond the scope of this dissertation is to examine the perception of nasalized
vowels across dialects of Spanish. An acoustic study mediates between production
and perception, and thus, is key to understanding the factors that contribute to
phonologization of nasalization as they pertain to the role of the listener in such
changes (Ohala, 1981, 1996, 2012). Additionally, it was important to collect the data
within each dialectal region, as speakers that reside outside of their dialectal area
of origin often accommodate their speech as a result of interactions with speakers
of other dialects. Additionally, collecting acoustic data allows working with larger
sample of participants. Articulatory and aerodynamic studies are typically impossible
and costly to conduct in the absence of a specialized laboratory.
With these issues in mind, data was collected by means of nasometry. A nasometer consists of a splitchannel set of microphones separated by a removable plate,
that record oral and nasal signals simultaneously yet separately. In other words, a
nasometer incorporates one microphone that records activity through the nose and
one microphone that records the mouth; a plate that rests above the lip separates
these two microphones.24 For this reason, recordings are made in stereo.
Nasometry provides a quantitative assesment of nasality. In addition to the
advantages mentioned above, the main advantage of nasometry is that it avoids
invasive and cumbersome procedures (e.g., nasography) while still providing fine-tuned
phonetic data. Even though there are considerable advantages when compared to
78
other instrumental measurements of nasality, nasometry does present limitations of
its own. The main disadvantage of using this instrument is that it restricts the type
of tasks that can be presented to speakers. For example, a sociolinguistic interview
would not be possible as the speaker would have to hold the nasometer throughout
the interview. Nevertheless, the nasometer can be used for cross-linguistic and crossdialectal investigations in which subjects are asked to read items (Thomas, 2010, p.
247).
This dissertation uses a Glottal Enterprises nasometer (NAS-1 SEP Clinic).
This nasometer is handheld by the participant and positioned above the upper lip,
between the mouth and the nose. The removable plate is inserted between the two
microphones and comes in three sizes (small, medium and large) to accommodate
to participants’ anatomy. Figure 4.6 illustrates how a participant would handle the
nasometer (photograph was provided in the nasometer’s literature). Figure 4.7, on
the other hand, presents the waveform and spectrogram produced by Praat (Boersma
& Weenink, 2016), showing nasal and oral channels (Channel 1 and Channel 2,
respectively).
Figure 4.6: Participant holding the nasometer, from the equipment manual.
79
Figure 4.7: Waveform and spectrogram of the word ‘bend’, produced by a native speaker
of English.
The Glottal Enterprises nasometer connects to the laptop computer through the
USB port. Recordings were done in Praat, in stereo and sampled at 44.1 kHz.
4.4.2
Procedure
Sessions were conducted by myself and I was present for the entirety of the
session. In Buenos Aires, I met with participants individually. Participants first
completed the recording and then the background questionnaire. In Santo Domingo,
data collection took place on university grounds while school was in session. Thus,
some accommodations had to be made so as not to disrupt classes (e.g. meeting with
two participants at a time. While one was recorded, the other participant completed
the background questionnaire.) In all cases, recording sessions took around 40 minutes.
At the start of each session, I informed participants, in non-specific terms, of the
nature of the study and obtained consent. I then showed them the nasometer, which
I had previously described to them as a ‘special microphone’, and how it had to be
held. I emphasized that contact between the plate and the face was essential, and I
80
demostrated the ideal angle of 10° from perpendicular, as suggested by the equipment
manual. After identifiying the correct plate size, I desinfected it and inserted in the
nasometer. I finally handed the nasometer to the participant, asked them to show me
how they would hold it and corrected posture if needed.
Before recording, I instructed participants to read phrases “naturally”, as they
would if they were reading something to a friend and in a pace that was neither slow
nor fast for them. I also emphasized “naturally” by saying that the task was not a
reading test and that at no time would I correct them or intervene. Finally, I invited
participants to correct themselves if they so desired at any point during the recording.
4.5
Data processing
4.5.1
Data management
In order to process the data efficiently, while at the same time guaranteeing low
incidence of human error during annotation and data management, the process was
streamlined using Praat and Python scripts. First, raw recordings were annotated in a
single-tier Textgrid. Intervals indicated carrier phrases and were left blank. Irrelevant
intervals (for example, at the beginning and end of the recording; or in between carrier
phrases) were marked with ‘xxx’. Next, labels were assigned automatically to the
blank intervals using a Praat script (Lennes, 2002a). A detailed explanation of how
labels were created is included in Appendix D, in page 202.
The assignment of labels was verified at random points of the raw recording,
as well as at the start and end of each file. These labels not only identified the
carrier phrase, but also the target’s type of syllable (e.g., CVC, CVN, etc.), word,
vowel, stress, and following segment. By including the coding of the linguistic internal
variables in the interval labels, human error in coding was greatly diminished.
81
Once intervals were labeled, another Praat script was run to extract each carrier
phrase into an individual .wav file (Lennes, 2002b). Proper extraction was verified
by checking random files of the directory where they were stored (and after finishing
data processing, I can verify no audio file was extracted incorrectly.) Finally, a third
Praat script was run to create TextGrids for each .wav file (adapted from Braver,
2012). Annotation was performed manually on each individual token file following the
criteria described in the next section.
4.5.2
Data annotation
The data was annotated in a Praat TextGrid using multiple acoustic events on a
token-by-token basis. The purpose of annotation was to identify (i) onset of the target
consonant, (ii) offset of the target consonant, and (iii) onset of the pre-consonantal
vowel. Additionally, annotation indicated whether the token was included or excluded
from analysis, as well as comments (e.g. why the token was excluded, if the participant
seemed to be reading faster or if they had laughed at the beginning of the phrase,
etc.), in different tiers. In the following subsections, I detail annotation procedures, as
well as criteria for exclusion. The description that follows deals with ‘well-behaved’
tokens. Example spectrograms (and waveforms) of segmentation are presented in
pages 84–86 and will be referenced throughout the next section. Tokens for which
additional considerations were made are discussed in section 4.5.2.3.
4.5.2.1
Annotating the target consonant
The target consonant, oral or nasal, was estimated by spectrographic reading as
well as waveform discontinuities. The target nasal segment was defined by the visual
presence of an abrupt change and damping of formant frequencies, in the spectrogram,
as exhibited by the telltale ‘smearing’ at the time of the formation of the articulatory
82
closure. The waveforms in the oral and nasal channels provided additional information:
a sudden absence of audio in the oral channel and a strong presence of audio in the
nasal channel were taken to indicate the nasal segment (see figures in the next section
for illustration, figures 4.8 to 4.11.)
Oral consonants, on the other hand, were always /s/. Fricatives present aperiodic
noise in the waveform and high frequency noise in the spectrogram due to the partial
passage of air through the oral cavity. The onset of the oral consonant was identified
based on the decrease in intensity in the oral channel in the waveform, as well as
a shift from periodic wave to aperiodic noise. In the spectrogram, loss of formant
structure was taken as an indication of onset of the oral consonant.
4.5.2.2
Annotating the pre-consonantal vowel
There were three different target syllable types: CVC, CVN and NVN. Because
the oral consonant in the onset of the syllable was different (/t/, /d/, /R/, /s/ or a
nasal), the criteria followed to identify the onset of the tautosyllabic vowel differed
depending on the onset consonant. With oral consonants (in CVC and CVN syllables),
first, the release of the consonant was identified in the acoustic signal. The onset of
the vowel was marked at the beginning of perioditicy in the waveform and consistent
formant structure in the spectrogram, be it following a burst and voice onset time
(VOT, as with stop consonant in a word such as sartén ‘skillet’; see Figure 4.8) or after
an occlusion (with /R/ as in a word such as tiraron ‘they threw’; see Figure 4.9, in page
84). In other cases, the oral consonant was /d/, as in the word veintidos ‘twenty-two’,
the oral consonant surfaced as an approximant (Martínez-Celdrán, 2008; see Figure
4.11, in page 86). With /d/, the onset of the vowel was defined as a decreased of
intensity in the waveform and weakening of formant structure in the spectrogram.
Finally, in NVN syllables, which include an onset nasal consonant, the beginning of
83
the vowel defined by the start of consistent formant structure and the breaking up
and ‘smearing’ in the spectrogram (see Figure 4.10, in page 85). Figures 4.8, 4.9,
4.10 and 4.11, which have been referenced to throughout this section, present sample
spectrograms of tokens that meet all requirements.
Frequency (Hz)
5000
0
sartén
e
0
n
k
0.6
Time (s)
Figure 4.8: Sample spectrogram of target word sartén / saRtén / ‘skillet’ in the phrase la
sartén caliente ‘the hot skillet’ produced by speaker DRF02 (Santo Domingo sample)
84
Frequency (Hz)
5000
0
tiraron
o
n
t
0
0.655
Time (s)
Figure 4.9: Sample spectrogram of target word tiraron / tiRáRon / ‘they threw’ in the
phrase cuando tiraron tabaco ‘when they threw tobacco’ produced by speaker DRF13 (Santo
Domingo sample)
Frequency (Hz)
5000
0
limón
o
0
n
0.619
Time (s)
Figure 4.10: Sample spectrogram of target word limón / limón / ‘lemon’ in the phrase un
ácido limón ‘an acid lemon’ produced by speaker BAF03 (Buenos Aires sample)
85
Frequency (Hz)
5500
0
veintidos
o
s
0
a
0.809
Time (s)
Figure 4.11: Sample spectrogram of target word veintidos / beintidós / ‘twenty two’ in
the phrase los veintidos aliados ‘the twenty two allies’ produced by speaker DRF04 (Santo
Domingo sample)
4.5.2.3
Less than picture perfect tokens
As the reader may very well imagine, not all tokens were picture perfect or wellbehaved. For example, in some cases, especially with onset /d/, onset /R/ and nasals
(e.g., the words veintidos ‘twenty two’, esperan ‘they wait’ and ganan ‘they win’), the
extent of lenition of the consonant complicated annotating the onset of the vowel.
This section addresses how these cases were handled.
With onset /d/, and in particular in the BAS corpus, there were considerable
amounts of consonantal weakening. In particular, one target word of the CVC stimuli
challenged annotation, veintidos /beintidos/ ‘twenty two’. In Figure 4.11 above we
see a /d/ with a well defined constriction: there is weakening of formant structure in
between vowels, accompanied by loss of energy in the waveform and intensity in the
spectrogram (shades of grey). Overall, the majority of BAS /d/ realizations do not
follow this description. Instead, BAS speakers produce lenited variants, in which the
86
constriction cannot be easily identified. Consider the spectrogram and waveform in
Figure 4.12.
Frequency (Hz)
5500
0
veintidos
s
0
0.867
Time (s)
Figure 4.12: Sample spectrogram of target word veintidos /beintidós/ ‘twenty two’ in the
phrase el número veintidos ‘the number twenty two’ produced by speaker BAF18 (Buenos
Aires sample)
In Figure 4.12 onset and offset of word-final /s/ is annotated. However, onset of
the vowel cannot be clearly established. There is a clear articulatory target for /i/, as
evidenced by the low F1 and high F2, which transitions into a higher F1 and low F2
for/o/. We see a correlate of this transition in the waveform as well, with an increase
in energy in the oral channel (channel 2) during /o/, which is in stressed position.
There is no evidence of an occlusion for /d/, except for a subtle narrowing of the
amplitude in the waveform at the time of formant movement.
In dealing with lenited variants of /d/, one option would be to discard these
tokens. This would be a less-than-desired approach, given that weakened intervocalic
/d/ was the case with most of the tokens in the BAS corpus (n= 361). Taking this
approach would mean that half of the data is discarded. In order to avoid discarding
87
this amount of data, an alternative approach was taken. First, the transition between
/i/ and /o/ was identified in reference to two acoustic events that would indicate
reduction in amplitude (as a result of greater constriction): (a) energy movements in
the waveform and (b) intensity contours. Once the transition was identified, onset of
/o/ was set at the midpoint of the transition. In the tier that annotates the comments,
tokens that were annotated following this guideline for reporting purposes –see the
example in Figure 4.13. While this approach means that some tokens present less
clear cut segmentation contexts, this is data that is not excluded.
Frequency (Hz)
5500
0
veintidos
o
s
tr
0
0.867
Time (s)
Figure 4.13: Sample spectrogram of target word veintidos /beintidós/ ‘twenty two’ in the
phrase el número veintidos ‘the number twenty two’ produced by speaker BAF18 (Buenos
Aires sample)
Some tokens in which the onset consonant of CVN target was /R/ were articulated
with a lenited variant (n = 70). A constriction could be identified on the basis of
spectrogram reading (i. e. clearer shades of gray) and a narrowing in the oral waveform
(see Figure 4.14). In the cases where it was not possible to establish a boundary
88
between the onset /R/ and the pre-nasal vowel, the token was excluded from analysis
(n = 8).
Frequency (Hz)
5000
0
esperan
a
n
k
0
1.118
Time (s)
Figure 4.14: Sample spectrogram of target word esperan / espéRan / ‘(they) wait’ in the
phrase esperan callados ‘they wait silently’ produced by speaker DRM08 (Santo Domingo
sample)
Finally, we address the cases with lenited variants of /n/ (n = 48). All tokens that
presented signs of weakening of the constriction of the nasal consonant were separated
for evaluation purposes. Two broad types of weakening patterns were identified. A
first set of tokens presented a somewhat weakened nasal consonant. Segmentation was
determined on the basis of spectrogram reading as well as waveform overall contours, in
the same vein as with lenited /R/. Another group of tokens showed no oral constriction
at all for the nasal consonant (nasality, however, was observed in the nasal channel).
In such cases, no boundary was established between the nasal consonant and the
pre-nasal vowel. Figure 4.15 presents a token with no signs of an oral constriction;
nasality, however, is not absent from the signal. Duration of the nasal consonant was
established as zero, and all measurements were taken during the production of the
89
vowel.
Frequency (Hz)
5500
0
capitan#a
cvn#a
0
0.818
Time (s)
Figure 4.15: Sample spectrogram of target word capitán / kapitán / ‘captain’ in the phrase
el capitán astuto produced by speaker DRF08 (Santo Domingo sample.)
With other tokens, on the other hand, it was clear that an oral constriction was
present in the signal, though it was not possible to establish an indisputable location
for the articulatory closure/release following the criteria described in section 4.5.2.1.
In these cases, the onset and offset of the target consonant was established in reference
to three acoustic cues: (a) a dip in frequency of formant contours (especially in the
F1 region), (b) a decrease in amplitude in the waveform in the oral channel, and
(c) change and damping of formant frequencies (‘smearing’ in the spectrogram). In
Figure 4.16, an oral constraint is observed in the oral channel waveform, albeit not
complete. Additionally, in the spectrogram, there is a dip in formant curves, and also
the formant is less clear.
90
Frequency (Hz)
5000
0
canon
o
n
0
a
0.471
Time (s)
Figure 4.16: Sample spectrogram of target word canón /kánon/ ‘cannon’ in the phrase el
canón alegre ‘the joyful cannon’ produced by speaker DRM06 (Santo Domingo sample.)
Another issue that at times complicated annotation was due to the recording
equipment itself. As with any microphone, the nasometer may pick up breathing
sounds which sometimes interact with the microphone being placed very close to the
speaker. In these cases, the offset of the (oral or nasal) consonant was marked as the
last regular periodic cycle in the waveform. That is, the last regular period before
frication, indicative of noise, begins. Additionally, the voicing bar in the spectrogram
was used as confirmation. Figure 4.17 presents one such example.
91
Frequency (Hz)
5000
0
tiraron
o
n
0
0.769
Time (s)
Figure 4.17: Sample spectrogram of target word tiraron / tiRáRon / ‘they threw’ in the
phrase digo que tiraron ‘I say they threw’ produced by speaker BAF02 (Buenos Aires sample)
4.5.2.4
Excluded data
Tokens were excluded for several reasons. The most common exclusion criterion
was participants misreading the target word. In some cases, participants misplaced
stress (for example, read the target word as [kanón], for /kánon/) or read the wrong
vowel (e.g., [espéRen] for /espéRan/). In other cases, tokens were excluded because the
participant misread the following word and the environment was non-target.
Other times, participants misread part of the carrier phrase. For example, participant BAF15 read Pasó en un santiamén ‘It happened suddenly’ as Pesó en un
santiamén, where a vowel in the first word was read incorrectly. In these cases, and
in all other cases where the target consonant and its environment were produced as
had been presented to the participant, the token was not excluded. The next section
specifies number of excluded tokens per environment.
92
4.6
Instrumental analyses
A total of 11,438 of tokens were extracted, of which 394 were eliminated from the
analysis during annotation, following the criteria described above. An additional set
of 72 tokens was eliminated from the analysis during data processing (I address this
issue in Section 4.6.2.) Thus, the total number of tokens analyzed was 10,972. Table
4.3 presents the count number and reason of excluded tokens per syllable-type. Of the
excluded tokens, speakers from the Dominican Republic produced the majority of the
tokens, with a total of 337 of the 394 excluded tokens (85.5%). I will come back to
this issue in the discussion of the results, in Chapter 6.
Table 4.3: Excluded tokens
n
Reason
misread target word
misread target vowel
misread stress placement
misread following word
segmentation issues
background noise
skipped token
recording issues
other
165
27
120
47
8
11
9
4
3
Two types of measurements were taken with Praat. First, the duration of the
word-final consonant was calculated. In addition, energy measurements were taken
at 20 equidistant time points over the course of the vowel and over the course of
the nasal consonant. Energy is measured in Pascal (Pa), a measure of air pressure
fluctuation. The script used the Praat’s standard “Get Energy...” command, which is
a squared measurement of amplitude over a given time interval. The formula Praat
uses to calculate energy is presented in formula 4.1, where x(t) is the amplitude.25
93
t2
t1
x2 (t)dt
(4.1)
Energy signals were then smoothed using a 5-point window and a number of
measurements were made using the resultant curves. The analysis of the data focuses
on weakening of the nasal consonant (section 4.6.1), the time-course of anticipatory
vowel nasalization (section 4.6.2). And given that these two have been linked (e.g.
Cedergren & Sankoff, 1975), the analysis also examines the relationship between the
two (section 4.6.3). Next, we describe each measurement scheme.
4.6.1
Weakening of the word-final nasal
Lenition of the nasal consonant can be manifested in two ways: in terms of
the duration of the consonant or as weakening in the consonantal constriction. The
duration of the nasal consonant was measured automatically in milliseconds in Praat,
from onset to offset following the segmentation criteria established in Section 4.5.2.
For tokens that have been labeled as exhibiting no oral constriction for the nasal
consonant (see above discussion of lenited variants of /n/), duration was set at 0.
For determining the degree of consonantal constriction, however, a series of
calculations were made. Fully constricted nasal consonants present a closure in the
oral channel, and an open velum. Thus, one way to examine weakening of the nasal
consonant is by looking at the oral channel during the production of the segment:
if there is a full constriction, the oral channel should present a flat waveform –full
closure equals zero energy. A vowel, on the other hand, presents the opposite pattern:
no constriction in the oral cavity leads to energy in the oral channel. To this end, we
calculate the ratio of oral energy in the nasal consonant to oral energy in the pre-nasal
vowel.
94
This ratio expresses the degree of constriction as a mathematical relationship
between the constricted segment (i.e. the nasal consonant) and an unconstricted
segment (i.e. the vowel). The inspiration for this measurement is taken from the
literature in weakening of /bdg/ (e.g. Hualde, Simonet, & Nadeu, 2011). For this
purpose, energy measurements in the oral channel were extracted from the vowel
and the nasal channel for each token. Next, the maximum oral energy reading was
extracted from the vowel and the minimum from the nasal consonant. Finally, the
oral energy in the consonant was divided by oral energy in the vowel. The higher the
value of this ratio, the less constricted the nasal consonant is. That is, the closer to
1, the more similar to the vowel the nasal consonant is; the closer to zero, the more
constricted it is. For tokens that have been labeled as exhibiting no oral constriction
for the nasal consonant (see above discussion of lenited variants of /n/), the ratio was
set at 1.
4.6.2
Time-course of nasality
While both nasal and oral energy measurements were collected in this study, only
nasal energy is considered to characterize the timing of the gesture. Nasal energy is
treated here as a proxy for the velic gesture. For this purpose, velic movement was
characterized in terms of three parameterized targets: (a) onset of the gesture, (b)
mid-gesture and (c) completion of the gesture.
In order to locate the three gestural landmarks, the minimum and the maximum
energy reading for each token were located. Figure 4.18 plots a random sample of
30 tokens. Black traces represent nasal energy traces, and blue and red points the
minimum and the maximum reading, respectively. Next, the range between the two
calculated. Onset of the gesture was operationalized as the point at which nasal
energy crossed a threshold of 15% of the range. Mid-gesture, on the other hand, was
95
operationalized as the time point at 50% of the range and completion of the gesture,
at 80% of the range (Delvaux et al., 2008; Honorof, 1999; Solé, 1992).
Figure 4.18: Sample of minimum and maximum readings for 30 random tokens.
After extracting the three landmarks, it was observed that some tokens (n=
72) still presented very jittery nasal energy signals even though the data had been
smoothed. Specifically, with these tokens the mid-point of the gesture (50% threshold)
would be extracted before the onset (15% threshold.) Upon closer examination of this
subset of the data, it was decided to exclude these tokens from the analysis.
A note should be made about control tokens, or CVC syllables. Per definition, the
gestural score of CVC syllables does not include a velic gesture. Nasal energy traces
could be extracted and analyzed following the aforementioned criteria (calculating
a range, determining the 15% cut off threshold, etc.), but this procedure would be
uninterpretable. For this reason, and in order to show that nasometer data does allow
characterizing the oral/nasal binary distinction, the CVC data will be chracterized in
terms of raw energy measurements, and in comparison to CVN and NVN in section
96
5.1 in the next chapter.
4.6.3
Relationship between time-course of nasality and weakening of the
nasal consonant
In order to express the relationship between the time-course of nasality and
weakening of the nasal consonant, onset of the velic gesture (as defined above) was
analyzed as a function of duration of the nasal consonant and consonant-to-vowel
ratio.
4.7
Statistical analyses
The quantitative analysis investigates the production of anticipatory vowel nasal-
ization and word-final nasal consonants according to the three dependent variables of
the present study and the influence of the various independent factors. Determining
which statistical analyses, and which variables and interactions to include in the
analyses, was not a straightforward process. Because different research questions are
answered in reference to different dependent variables, the analysis was carried out
using various statistical models. Each of the descriptive and inferential statistical
analyses run will be described in detail in chapter 5 with the corresponding results.
97
Chapter 5
RESULTS
Employing the methods described in the previous chapter, this chapter presents
the results of the analysis of the production of word-final nasals and anticipatory
vowel nasalization, and the relationship between the two, by the participants in the
present study. As described in the Methods chapter, a total of 54 speakers, 26 from
Buenos Aires (BAS) and 28 from Santo Domingo (SDS) were surveyed. A total of
11,438 of tokens were analyzed. This chapter is structured in the following manner.
The first section establishes differences between the control syllables, CVC, and the
target sequences, CVN and NVN. In section 5.2, the analysis of weakening of the
nasal consonant is addressed. Section 5.3 presents the analysis of the time-course of
nasalization. Finally, section 5.4 examines the relationship between the time-course of
nasalization and weakening of the nasal consonant. An overall summary of results is
presented in section 5.5. Taken together, these results provide the basis for answering
the research questions that guided this dissertation, which will be addressed more
explicitly in the discussion chapter.
5.1
Control and target items
CVC syllables, which serve as control stimuli, do not incorporate nasal segments,
and therefore the hypothesis in H9 is that they will present no evidence of nasalization,
98
neither in the consonant nor in the tautosyllabic vowel. A first analysis of raw energy
traces showed that CVC tokens presented flat nasality traces, compared to CVN and
NVN sequences. Figure 5.1 plots smoothed energy nasal traces for CVC, CVN and
NVN, per dialect, for comparison. In order to produce a more visually amenable
figure, 10% of the tokens in the corpus were randomly selected. Black traces represent
individual tokens, whereas red traces represent the mean. Negative time points take
place during the pre-nasal vowels. Zero (i.e. the dotted green line) represents onset of
the consonant (oral for CVC; nasal for CVN and NVN.) Thus, it illustrates the vowel
and the first fourth of the consonant.
cvc
cvn
nvn
0.5
0.4
BA
0.3
Smoothed nasal energy
0.2
0.1
0.0
0.5
0.4
DR
0.3
0.2
0.1
0.0
−15
−10
−5
0
5
−15
−10
−5
0
5
−15
−10
−5
0
5
Smoothed time point
Figure 5.1: Time-course of nasal energy for CVC, CVN and NVN of 10% of the tokens in
the corpus, per dialect.
Figure 5.1 shows a clear contrast between tokens with an oral and a nasal wordfinal consonant. Tokens with a word-final oral consonant, which do not include a
velic gesture, present overwhelmingly floor-level readings, confirming H9. Tokens
with a word-final nasal consonant, on the other hand, show traces of increasing nasal
99
energy. As a whole, these data are not of much interest, serving rather to confirm
the proper functioning of the nasometer. Having established that the lack of velic
movement during CVC syllables correlates with representations of nasal energy, control
tokens will not be considered further. We now turn to the analysis of syllables with
a word-final nasal consonant, CVN and NVN, and consider weakening of the nasal
consonant, the time-course of nasality and the relationship between the two.
5.2
Weakening of the nasal consonant
5.2.1
Duration of the nasal consonant
The first measurement of weakening of the nasal consonant is duration: shorter
nasal consonants are more weakened than longer nasal consonants. The hypothesis
regarding the effect of dialect was that SDS would exhibit more weakened variants
(H1). To that effect, the prediction was that this dialect group would exhibit shorter
nasal consonants. Table 5.1 includes the number of cases, mean, standard deviation,
median and range for the two dialect groups. Figure 5.2 presents box, jitter and violin
plots of the same data. The choice of three modes of visualization is to facilitate
comparison across levels of the variable while presenting the distribution of the data.
Box plots describe the full range of variation (from min to max), the likely range of
variation (the IQR), and a typical value (the median). The jitter plots permit assesing
the relative distribution of all points. They are similar to scatterplots, only that one
of the variables is categorical, and by adding jitter, we can see every individual data
point. Finally, the violin plots show the probability density of the data at different
values. SDS is shorter than BAS, though differences are very small with means of
90ms (BAS) and 87.2ms (SDS).
100
Table 5.1: Descriptive statistics on the duration of the nasal consonant (in ms), per dialect.
The values are raw averages (M ) and standard deviations (sd) that have not been adjusted
by the statistical model.
N
M
3493
90.0
BAS
sd median
41.9
81.0
range
N
M
316.4
3530
87.2
SDS
sd median
39.0
75.4
range
283.6
Figure 5.2: Box, jitter plot and density plot for duration of the nasal consonant (in ms),
per dialect. Diamonds represent the mean.
101
In order to determine the statistical significance of the patterns of duration
of the word-final nasal consonant, a repeated measures ANOVA was run with raw
duration of the nasal consonant (in ms) as the dependent variable, with syllable
type, vowel type, stress and environment as within-subject variables, and dialect
group as between-subjects variable (the current, and all statistical analyses, were
run in SPSS Corp., 2013). A full-factorial model was not used, as this would have
included too many factors and would have diminished statistical power. All main
effects were included in the model, as well as all simple interactions between dialect
and the linguistic variables (dialect×syllable, dialect×stress, dialect×vowel type, and
dialect×environment.) Because previous work does not show relationships between
other variables, no other interactions were included in the model in order to prevent
the model from over-fitting the data. Statistical significance was set at α = .05.
Given that Mauchly’s test of sphericity was significant, we use the Greenhouse-Geisser
correction for the test of overall main effects and the interactions (all results in section
5.2 will reporting using the Greenhouse-Geisser correct as per recommendation in
Howell, 2012).
Table 5.2 summarizes main effects and interactions. The lack of a main effect
for dialect group provides statistical confirmation that duration of the word-final
nasal consonant does not vary as a function of dialect group. What these findings
suggest is that weakening (or lack thereof) of the nasal consonant does not vary as
a function of dialect, at least with regard to duration of the nasal consonant in this
formal context. Duration of the nasal consonant does vary, however, according to
the linguistic variables. The results indicate that the nasal consonant is shorter (a)
in CVN syllables, (b) in unstressed syllables and (c) when preceded by /a/. Tables
5.3 to 5.5 include descriptive statistics for each dialect (mean, standard deviation,
median and range). Visualization of the data is presented in Figure 5.3. Despite the
102
statistical findings, differences in duration of the nasal consonant across levels of these
linguistic variables are small and do not amount to more than 5ms at most.
Table 5.2: ANOVA results for main effects and significant interaction for the analysis of
duration of the nasal consonant as dependent variable.
F statistic
Factor
Dialect
2.06
Syllable
9.24
Stress
56.36
Vowel
5.60
Environment
273.56
Dialect × Syllable
0.37
Dialect × Stress
0.38
Dialect × Vowel
1.42
Dialect × Environment
2.9
∗= statistically significant; n.s. = not
p
ηp2
n.s 0.06
∗ 0.23
∗ 0.64
∗ 0.15
∗ 0.90
n.s 0.01
n.s 0.12
n.s 0.04
n.s 0.08
significant
Table 5.3: Descriptive statistics on the duration of the nasal consonant (in ms), per syllable
type and dialect. The values are raw averages (M ) and standard deviations (sd) that have
not been adjusted by the statistical model.
M
sd
BAS
median
88.0
41.9
80.2
0
277.9
86.1
39.7
74.1
0
283.6
NVN 92.2
41.7
82.0
0
316.4
88.4
38.2
76.4
0
275.0
CVN
min
max
M
sd
SDS
median
min
max
Table 5.4: Descriptive statistics on the duration of the nasal consonant (in ms), per stress
condition and dialect. The values are raw averages (M ) and standard deviations (sd) that
have not been adjusted by the statistical model.
M
sd
BAS
median
92.9
45.7
83.4
0
277.9
90.0
41.0
75.9
0
267.6
Unstr. 87.0
37.2
78.9
0
316.4
84.2
36.5
74.8
0
283.6
Str.
min
max
M
103
sd
SDS
median
min
max
Table 5.5: Descriptive statistics on the duration of the nasal consonant (in ms), per stress
condition and dialect. The values are raw averages (M ) and standard deviations (sd) that
have not been adjusted by the statistical model.
M
sd
BAS
median
/e/
90.3
41.5
81.7
0
277.9
86.2
39.7
73.5
0
267.6
/a/
89.2
42.7
79.4
0
251.6
86.2
39.7
73.5
0
256.7
/o/ 90.6
41.4
81.8
0
316.4
85.4
38.7
74.8
0
283.6
min
max
M
104
sd
SDS
median
min
max
105
Figure 5.3: Box, jitter plot and density plot for duration of the nasal consonant (in ms) for syllable, stress and vowel type, per
dialect. Diamonds represent the mean.
Environment (that is, whether the word-final nasal consonant is followed by a
pause, a vowel or a consonant) also had a main effect, with the largest effect size in
the model (ηp2 = .90). In H5, we hypothesized that word-final nasals would exhibit
utterance-final lengthening and, therefore, they would be longer in the pre-pausal
(/n##/) than in the pre-vocalic condition(/n#V/). Figure 5.4 illustrates the data
and Table 5.6 summarizes the descriptive statistics (mean, standard deviation, median,
and range.) In pre-pausal condition (/n##/), both dialects exhibit longer nasal
consonants than in the pre-vocalic condition (/n#V/). In pre-pausal condition, the
nasal consonant has an average duration of 142.3ms (sd= 38.7; BAS) and 136.9ms
(sd= 39.1; SDS), compared to 50.4ms (sd= 21.4; BAS) and 60ms (sd= 22.7; SDS) in
pre-vocalic condition. As such, the findings confirm H5.
Table 5.6: Descriptive statistics on the duration of the nasal consonant (in ms) for prepausal and pre-vocalic environments, per dialect. The values are raw averages (M ) and
standard deviations (sd) that have not been adjusted by the statistical model.
M
sd
BAS
median
Prepausal
142.3
38.7
140.0
29.1
316.4
136.9
39.1
135.4
0
283.6
Prevocalic
50.4
21.4
48.8
0
245.7
60.0
22.7
57.3
0
212.0
min
max
M
sd
SDS
median
min
max
The pattern for pre-consonantal environments is different. The hypothesis was
that dialects would exhibit divergent patterns because of known regional variants at
different places of articulation: dorsal in SDS and coronal in BAS (H6.) In the condition
prone to gestural blending (i.e. same place of articulation of the nasal and following
oral consonant) word-final nasal consonants will be less weakened (and, therefore,
longer). For SDS, the predicted pattern is that nasal consonants will be longer in
the pre-dorsal condition (/n#k/). For BAS, the opposite pattern is expected –longer
106
Figure 5.4: Box, jitter plot and density plot for duration of the nasal consonant (in ms)
for syllable, stress and vowel type, per dialect. Diamonds represent the mean.
word-final nasal consonants in pre-coronal condition (/n#t/). Figure 5.5 presents box,
jitter and violin plots of pre-dorsal and pre-coronal enviroments, per dialect group.
Table 5.7 summarizes the descriptive statistics (mean, standard deviation, median,
107
and range.) The BAS data follows the predicted patterns: the nasal consonant in
pre-coronal (/n#t/) condition is on average longer than the nasal consonant before a
dorsal consonant (/n#/˛) . Differences across conditions are minimal in this case as well
(i.e. 83.7 vs. 77.2ms). For the SDS data, the predicted pattern does not obtain. The
nasal consonant does not present robust differences across pre-consonantal condition.
While on average, duration is greater before a coronal consonant, mean differences
across conditions are in the range of 2ms. The interaction term dialect×environment
was not significant. Thus, the results do not support the predictions.
Table 5.7: Descriptive statistics on the duration of the nasal consonant (in ms) for precoronal and pre-dorsal environments, per dialect. The values are raw averages (M ) and
standard deviations sd have not been adjusted by the statistical model.
M
Pre83.7
coronal
Predorsal
77.2
sd
BAS
median
sd
SDS
median
min
max
M
min
max
18.1
81.7
41.3
204.3
78.7
18.5
77.2
34.5
182.5
17.7
74.1
0
154.0
76.6
20.9
72.8
0
191.8
Taken together, the results for the effect of environment indicate that the weakening
(as indexed by duration) is greatest in the pre-vocalic condition (/n#V). Word-final
nasal consonants at the end of the utterance, in pre-pausal condition (/n##), are the
least weakened. The patterns for the pre-consonantal environment are more difficult
to interpret. These will be detailed further in the next chapter, when the effect of the
linguistic variables is discussed.
5.2.1.1
Summary
To summarize the findings in this section, as indexed by duration, word-final
nasal consonants are shorter for SDS, but differences across dialectal groups were
108
Figure 5.5: Box, jitter plot and density plot for duration of the nasal consonant (in ms)
for pre-coronal and pre-dorsal environments, per dialect. Diamonds represent the mean.
not significant. The analysis of the linguistic variables, however, revealed statistical
trends in the data under analysis. Nasal consonants tend to be shorter when the
syllable is CVN, in unstressed position and when the vowel in the syllable is /a/.
109
Additionally, in pre-vocalic condition (/n#V/) nasal consonants are the shortest. And
comparing across pre-consonantal conditions, /n#t/ or /n#k/ reveals that before a
dorsal consonant, word-final nasal consonants are the shorter than before a dorsal
consonant. In the next section, we present the results for the consonant-to-vowel oral
energy ratio.
5.2.2
Constriction in the nasal consonant
The second measurement of weakening of the nasal consonant is a measurement
of constriction. Degree of constriction is expressed here as a ratio of the minimum
oral energy reading in the nasal consonant to the maximum in the pre-nasal vowel.
The higher the value of this calculation, the less constricted (and more similar to the
vowel) the nasal consonant is.
The hypothesis under examination is that SDS exhibits more weakened variants
than BAS (H1). As such, this dialect group is expected to exhibit greater values of
the ratio than BAS. Table 5.8 includes the number of cases, mean, standard deviation,
median and range. Figure 5.6 presents box, jitter plot and violin plots.26 SDS shows
a (slightly) higher consonant-to-vowel oral energy ratio than BAS, with a mean of 0.1
(sd= 0.1) for SDS and 0.08 (sd= 0.5), and thus, are in line with the hypothesis (and
prediction.)
Table 5.8: Descriptive statistics of consonant-to-vowel oral energy ratio, per dialect. The
values are raw averages (M ) and standard deviations (sd) that have not been adjusted by
the statistical model.
N
M
3693
0.08
BAS
sd median
0.50
0.12
range
N
M
0.99
3681
0.10
110
SDS
sd median
0.13
0.50
range
0.99
Figure 5.6: Box, jitter plot and density plot for consonant-to-vowel oral energy ratio, per
dialect. Diamonds represent the mean.
In order to examine the statistical significance of the patterns of consonant-tovowel energy ratio, a repeated measures ANOVA was run with raw oral energy ratio
as the dependent variable, and syllable type, vowel type, stress, and environment
as within-subjects variables and dialect group as between-subjects variables. Much
like with duration of the nasal consonant, it was decided not to use a full-factorial
model. Instead, all main effects were included in the model, as well as all simple
interactions between dialect and linguistic variables (dialect×syllable, dialect×stress,
dialect×vowel type, and dialect×environment.) Statistical significance was set at α=
.05.
111
Table 5.9 summarizes the results of the statistical analysis. Even though SDS
presents a higher consonant-to-vowel oral energy ratio, dialect did not have a main
effect. These findings indicate that dialects are not different in terms of consonantto-vowel oral energy ratio in addition to duration. Thus, weakening is not different
across dialect-groups in this corpus. The analysis does reveal significant main effects
for other independent variables. These findings indicate that nasal consonants in (a)
NVN syllables, (b) unstressed condition and (c) with a preceding /e/ vowel are more
weakened, as indexed by the constriction ratio. In the same vein as with duration of
the nasal consonant, these variables exhibited significant main effects, even though
mean differences are small. Tables 5.10 to 5.12 include descriptive statistics (mean,
standard deviation, median and range) for each of these independent variables, per
dialect. Visualization of the data is presented in figure 5.7.
Table 5.9: ANOVA results for main effects and significant interaction for the analysis of
oral energy ratio as dependent variable. ∗= statistically significant; n.s. = not significant
Factor
F statistic
Dialect
Syllable
Stress
Vowel
Environment
Dialect × Environment
Dialect × Syllable
Dialect × Stress
Dialect × Vowel
∗= statistically significant;
112
2.18
19.83
64.37
8.71
104.81
7.32
2.81
1.57
0.08
n.s. = not
p
ηp2
n.s 0.45
∗ 0.30
∗ 0.58
∗ 0.15
∗ 0.69
∗ 0.14
n.s 0.06
n.s 0.03
n.s 0.00
significant
Table 5.10: Descriptive statistics of consonant-to-vowel oral energy ratio for syllable type,
per dialect. The values are raw averages (M ) and standard deviations (sd) have not been
adjusted by the statistical model.
M
sd
BAS
median
0.08
0.1
0.4
0
1.0
0.1
0.12
0.05
0
1.0
NVN 0.09
0.1
0.05
0
1.0
0.1
0.1
0.05
0
1.0
CVN
min
max
M
sd
SDS
median
min
max
Table 5.11: Descriptive statistics of consonant-to-vowel oral energy ratio for stress condition,
per dialect. The values are raw averages (M ) and standard deviations (sd) have not been
adjusted by the statistical model.
M
sd
BAS
median
0.07
0.1
0.04
0
1.0
0.09
0.1
0.05
0
1.0
Unstr. 0.09
0.1
0.05
0
1.0
0.1
0.1
0.05
0
1.0
Str.
min
max
M
sd
SDS
median
min
max
Table 5.12: Descriptive statistics of consonant-to-vowel oral energy ratio for vowel type,
per dialect. The values are raw averages (M ) and standard deviations (sd) have not been
adjusted by the statistical model.
M
sd
BAS
median
min
max
M
sd
SDS
median
min
max
/e/
0.09
0.1
0.04
0
1.0
0.1
0.1
0.06
0
1.0
/a/
0.08
0.1
0.04
0
1.0
0.09
0.1
0.04
0
1.0
/o/ 0.08
0.1
0.04
0
1.0
0.1
0.1
0.04
0
1.0
113
Figure 5.7: Box, jitter plot and density plot for consonant-to-vowel oral energy ratio for
syllable, stress and vowel type. Diamonds represent the mean.
Environment also had a statistical main effect and was part of a significant
interaction with dialect. In order to explore the significant interaction we used a
line plot to examine differences in mean across environment conditions and dialect
groups (figure 5.8). In pre-prepausal and pre-vocalic condition, BAS exhibits a lower
consonant-to-vowel oral energy ratio. Additionally, BAS exhibits different means per
condition, whereas for SDS we see that the pre-coronal and pre-pausal conditions have
114
similar means.
Consonant−to−vowel ratio
0.20
0.15
0.10
0.05
BAS
SDS
Dialect group
pre−coronal
pre−pausal
pre−dorsal
pre−vocalic
Figure 5.8: Line graph for interaction of environment with dialect group for consonant-tovowel oral energy ratio.
In H5, we hypothesized more weakening in the pre-vocalic condition (/n#V/)
than in the pre-pausal (/n##/), and thus predicted a higher consonant-to-vowel oral
energy ratio in the former than in the later. Findings confirm the hypothesis. While
still very constricted segments, in pre-vocalic condition, nasal consonants exhibit a
higher consonant-to-vowel oral energy ratio, with a mean of 0.2 (sd= 0.2) compared
to the mean of 0.04 (sd= 0.07) for pre-pausal. Table 5.13 summarizes mean, standard
deviation, median and range for each dialect. Figure 5.9 presents box, jitter and violin
plots.
115
Table 5.13: Descriptive statistics of consonant-to-vowel oral energy ratio for pre-pausal
and pre-vocalic, per dialect. The values are raw averages (M ) and standard deviations (sd)
have not been adjusted by the statistical model.
M
sd
BAS
median
Prepausal
0.03
0.05
0.01
0
1.0
0.05
0.08
0.03
0
1.0
Prevocalic
0.2
0.05
0.01
0
1.0
0.2
0.2
0.2
0
1.0
min
max
M
sd
SDS
median
min
max
Figure 5.9: Box, jitter plot and density plot for consonant-to-vowel oral energy ratio for
pre-pausal and pre-vocalic environments, per dialect. Diamonds represent the mean.
Finally, we consider differences across pre-consonantal conditions. The hypothesis
116
was that weakening would be greater in conditions where the articulator of the nasal
consonant was different from the articulator of the oral consonant (pre-dorsal for
BAS and pre-coronal for SDS; H4.) In as much as a higher consonant-to-vowel oral
energy ratio value indicates that the nasal consonant is more similar to its preceding
vowel, the prediction was that BAS would exhibit greater ratio values in the pre-dorsal
(/n#k/) condition than in the pre-coronal condition. (/n#t/) The opposite trend
is expected for SDS: greater ratio values for the pre-coronal condition. Table 5.14
includes descriptive statistics for both dialects (mean, standard deviation, median
and range). Figure 5.10 illustrates the data. The results reveal that both dialects
present the same trend: a higher consonant-to-vowel oral energy ratio in the pre-dorsal
condition than in the pre-coronal, albeit with small differences. These findings support
the hypothesis for BAS, but reject it for SDS.
Table 5.14: Descriptive statistics of consonant-to-vowel oral energy ratio for pre-dorsal
and pre-coronal, per dialect. The values are raw averages (M ) and standard deviations sd
have not been adjusted by the statistical model.
M
Pre0.06
coronal
Predorsal
0.09
sd
BAS
median
min
max
M
0.06
0.04
0
0.06
0.08
0.06
0
0.05
117
(sd)
SDS
median
min
max
0.05
0.07
0.03
0
1.0
0.08
0.09
0.05
0
1.0
Figure 5.10: Box, jitter plot and density plot for consonant-to-vowel oral energy ratio for
pre-coronal and pre-dorsal environments, per dialect. Diamonds represent the mean.
5.2.2.1
Summary
In terms of consonant-to-vowel oral energy ratio, word-final nasal consonants are
not different across dialect groups. Regarding the linguistic variables, the statistical
analysis revealed that nasal consonants are less constricted with NVN syllables, in
unstressed position, and when the vowel is /e/. The word-final consonant also had
118
the highest consonant-to-vowel oral energy ratio in the pre-vocalic condition. The
difference between pre-pausal and pre-vocalic conditions are larger for SDS than for
BAS. A comparison of the pre-consonantal environments reveals that before a dorsal
consonant (/n#k/), the energy ratio is higher, and this trend holds across dialect
groups. Having described the findings for weakening of the word-final nasal consonant,
prior to discussing them, we present the analysis of the time-course of nasalization in
the following section.
5.3
Time-course of nasality
In this section, we present the analysis of the time-course of nasality. The structure
of this section will be different from the previous one. As will become clear in short,
the variable of interest (onset of the velic gesture) is not normally distributed. Thus, in
order to determine the statistical significance, a different statistical approach had to be
taken. Ideally, the statistical analysis would examine all relevant factors (main effect
and interaction terms) in one single model, therefore providing the ability to compare
across dialect groups while minimizing type I and type II errors. But parametric tests,
such as a factorial ANOVA, are discouraged in this case given that the data does not
meet the assumption of normality.
Non-parametric statistics do not require that the data be normally distributed,
and are thus compatible with the data set under analysis here. A caveat of nonparametric tests is that there is no factorial-ANOVA counterpart. This approach has
two disadvantages, though. The first is that interactions cannot be examined. The
second issue has to do with familywise error: every time we run a statistical test, we
increase the possibility of Type I error. In other words, multiple tests increase the
possibility of observing a (significant) difference when in truth such difference is not
119
statistically significant.
Despite these issues, a series of non-parametric tests were performed: three MannWhitney U and three Kruskal-Wallis tests, which will be referenced in the relevant
sections in the exposition of the results. In order to correct for familywise error, the
Bonferroni familywise p-value correction to interpret the results will be used. Thus,
statistical significance was set at α = .003. We now turn to the analysis of each
independent variable, starting with the analysis of the effect of dialect.
The time-course of nasality had been characterized in this study in terms of three
gestural landmarks: onset of the velic gesture, mid-gesture and completion of the
gesture. Onset of the gesture is the time point at which nasal energy crosses the 15%
range; mid-gesture is operationalized at 50% of the range; finally, completion of the
gesture is at 80% of the range (see section 4.6.2 in chapter 5). The hypothesis is that
dialect groups will be different with regard to the onset of the gesture. The prediction
was that SDS would present an earlier onset of the velic gesture. Table 5.15 includes
the mean, standard deviation, median and range. The data is visualized in figure
5.11, which presents a histogram with the distribution of the data for each gestural
landmark, per dialect group. The top panel represents the first landmark, onset of
the gesture (threshold at 15%). The mid-gesture (threshold at 50%) is in the panel
in the middle. The bottom panel plots completion of the gesture (threshold at 80%).
Dashed lines represent the mean for each dialect group (yellow for SDS, blue for BAS).
Time point zero corresponds to the onset of the nasal consonant, represented in the
figure as a dotted line. Thus, negative time points take place during the pre-nasal
vowel, whereas positive time points during the nasal consonant.
In H2, we hypothesized that SDS would show earlier onset of the velic gesture than
BAS. For SDS, onset of the gesture takes place at time point -7.0 (sd= 4.8), whereas
it is at time point -4.9 (sd= 4.7) for BAS. We also see that SDS achieves the midpoint
120
gesture also earlier, at -2.0 (sd= 4.3) compared to BAS at -0.8 (sd= 3.6). In terms of
completion, both dialects achieve it at roughly at the same time point: 1.2 (sd= 4.1)
for SDS and 1.7 (sd=3.5) for BAS. These findings confirm the hypothesis and indicate
that, for SDS, the pre-nasal vowel exhibits a greater extent of anticipatory vowel
nasalization. A Mann-Whitney test with onset of the gesture as dependent variable
revealed that the differences across dialects is statistically significant, U =4,474,085,
p=.000.
Figure 5.11: Histograms for time-course of nasality per dialect. Dashed lines represent the
mean.
121
Table 5.15: Descriptive statistics on time-course of nasality per dialect group. The values
are raw averages (M ) and standard deviations (sd) that have not been adjusted by the
statistical model.
M
Onset
-4.8
of gesture
sd
BAS
median
SDS
median
min
max
M
min
max
4.5
-3.0
-17.0
11.0
-7.0 4.8
-7.0
-17.0
10.0
sd
Midgesture
-0.5
3.0
0
-16.0
16.0
-1.8
4.0
-1.0
-16.0
17.0
Completion
of gesture
2
2.8
2.0
-16.0
18.0
1.4
3.8
2.0
-16.0
18.0
Concerning syllable, the purpose of including NVN to compare with CVN syllables
was to test whether the velic gesture for the word-final /n/ is separate from the gesture
of the vowel. The hypothesis was that NVN would exhibit no velic movement, as it
had been lowered for the preceding nasal consonant and not raised again (H10.) In
other words, if the velum is lowered for the onset nasal consonant, will it stay lowered
throughout the vowel, or will it be raised for the vowel and lowered again before
articulating the word-final nasal consonant? If the former, then we would expect to
see flat nasal energy traces. If the latter, we would expect to see changes in the nasal
energy traces. The results reject the hypothesis: NVN syllables do exhibit lowering
of the velum. The data shows movement in velic aperture between the two nasal
consonants (figures 5.12 and 5.13 plot randomly selected 100 tokens for CVN and
100 for NVN). Specifically, nasal energy increases as time point 0 is reached (i.e. the
boundary between the nasal consonant and the vowel). A comparison of dialects also
shows that the effect is much more pronounced for BAS, where we often see a turning
point in nasal energy traces (the mean trace places it at time point -3.0). The overall
trend in the SDS data, on the other hand, is a steady cline throughout the segment.
122
BAS
SDS
smoothed nasal energy
0.20
0.15
0.10
0.05
0.00
−15
−10
−5
0
5
−15
−10
−5
0
5
Time Point
Figure 5.12: Time-course of nasal energy for CVN in 100 randomly selected tokens in the
corpus, per dialect. Red line represents the mean.
BAS
SDS
smoothed nasal energy
0.20
0.15
0.10
0.05
0.00
−15
−10
−5
0
5
−15
−10
−5
0
5
Time Point
Figure 5.13: Time-course of nasal energy for NVN in 100 randomly selected tokens in the
corpus, per dialect. Red line represents the mean.
123
Having established velic movement, we move on to present the analysis of the
onset of velum lowering per syllable type, CVN vs. NVN. Figure 5.14 presents the
distribution of the data per gestural landmark, syllable type and dialect. Dashed
vertical lines indicate mean values per syllable type (grey for CVN and orange for
NVN). Histograms to the left belong to BAS data, whereas those in the right represent
SDS. Descriptive statistics are presented in Table 5.16. Overall, both dialects present
the same pattern, and the trend of SDS exhibiting earlier onset of nasality obtains
here as well. The data shows that onset of the gesture takes place earlier with CVN
than with NVN syllables. For SDS, mean onset of the gesture takes place at time
point -8.7 (sd= 4.8) for CVN, whereas -5.2 (sd= 4.1) for NVN. With the BAS data,
we find that mean onset of velic gesture occurs at -6.4 (sd= 4.8) and at -3.0 (sd= 3.3)
for NVN. A Mann-Whitney U with onset of the gesture as dependent variable revealed
that the differences across syllable type are statistically significant, U = 8,747,563.5,
p= .000.
124
Figure 5.14: Histograms for time-course of nasality per syllable type and dialect.
Table 5.16: Descriptive statistics on time-course of nasality per syllable type, BAS data.
The values are raw averages (M ) and standard deviations (sd) that have not been adjusted
by the statistical model.
M
Onset
-6.4
of gesture
sd
CVN
median min
max
M
sd
NVN
median min
max
4.8
-6.0
-17.0
11.0
-3.0
3.3
-2
-17.0
10.0
Mid-gesture
-1.0
3.3
0
-16.0
16.0
0
2.5
0
-16.0
11.0
Completion
of gesture
1.8
2.9
2.0
-16.0
18
2.2
2.5
2
-16.0
14.0
125
Table 5.17: Descriptive statistics on time-course of nasality per syllable type, SDS data.
M
Onset
-8.7
of gesture
sd
CVN
median min
max
M
sd
NVN
median min
max
4.8
-9
-17.0
10.0
-5.2
4.1
-4
-17.0
8.0
Mid-gesture
-2.5
4.3
-1.0
-16.0
17.0
-1.0
3.5
0
-16.0
14.0
Completion
of gesture
1.0
4
2.0
-15.0
18.0
1.7
3.7
2.0
-16.0
15.0
Next, we examine the distribution of the data according to stress condition:
stressed vs. unstressed syllables. As stated in Section 2.3 in Chapter 2, articulatory
research revealed that in stressed syllables the velum lowers earlier. Additionally, the
literature examining universals of sound change in nasalization makes similar remarks:
nasalization spreads from stressed pre-nasal vowels to unstressed pre-nasal vowels. For
this reason, H14 hypothesized that stressed syllables would exhibit an earlier lowering
of the velum. Since for SDS nasality is hypothesized to be an intended property of
the vowel, H17 makes the hypothesis that earlier velum lowering would be enhanced
with this dialect group. Figure 5.15 presents histograms with the distribution of
the data per gestural landmark, stress condition and dialect. Descriptive statistics
for each dialect are presented in Tables 5.18 and 5.19. Results show that the mean
onset of the gesture is earlier with stressed syllables in agreement with the hypothesis
(H14), and the trend of SDS presenting earlier onset of the gesture obtains here as
well. Differences between stressed and unstressed targets are subtle in both dialect
groups, with mean difference between conditions being less than one time point. In
the stressed condition, BAS presents mean onset of the gesture at time point -5.0
(sd= 4.6) and SDS at time point -7.6 (sd= 4.8). In unstressed condition, mean values
are at time point -4.6 (sd= 4.4) for BAS and time point -6.4 (sd= 4.7) for SDS. A
Mann-Whitney U with onset of the gesture as dependent variable revealed that the
126
differences across stress conditions are statistically significant, U = 6,742,096.5, p=
.000. In order to compare across dialects and explore H17, a second Mann-Whitney U
was run on the stressed data, which revealed that differences across dialects in the
stressed condition are significant, U = 1,129,278.5, p= .000.
Figure 5.15: Histograms for time-course of nasality per stress condition and dialect.
Table 5.18: Descriptive statistics on time-course of nasality per stress condition, BAS data.
The values are raw averages (M ) and standard deviations (sd) that have not been adjusted
by the statistical model.
Stressed
median min
M
sd
-5
4.6
-3
Mid-gesture
-0.5
2.9
Completion
of gesture
2
2.6
Onset
of gesture
Unstressed
median min
max
M
sd
-17
11
-4.6
4.4
-3
-17
7
0
-16
12
-0.5 3.1
0
-16
16
2
-16
14
2
-16
18
127
2
2.9
max
Table 5.19: Descriptive statistics on time-course of nasality per stress condition, SDS data.
The values are raw averages (M ) and standard deviations (sd) that have not been adjusted
by the statistical model.
M
Onset
-7.6
of gesture
sd
Stressed
median min
max
M
Unstressed
sd median min
max
4.8
-8.0
-17.0
10.0
-6.4
4.7
-6.0
-17.0
10.0
Mid-gesture
-2.2
4.1
-1.0
-16.0
14.0
-1.4
3.8
0
-16.0
17.0
Completion
of gesture
1.2
4.1
2.0
-16.0
15.0
1.6
3.6
2.0
-16.0
18.0
Vowel type (/a/ vs. /o/ vs. /e/) has also been hypothesized to affect the realization
of anticipatory vowel nasalization. A recurring observation in the previous literature
(historical and instrumental) has been that there is greater velum lowering during
the production of low vowels due to the muscles involved during tongue movement.
Thus, one prediction would be that /a/ will exhibit greater nasality (H12.) H13, on
the other hand, predicted no effect of frontness and backness. Thus, no differences
between mid vowels were expected. Figure 5.16 presents the data per vowel type and
dialect, and Table 5.20 details descriptive statistics. Both dialects show overlapping
distributions per vowel type, with the mid back vowel presenting a subtly earlier onset
of the gesture than the other two vowel types. For BAS, /o/ has a mean onset of
the gesture at -6.8 (sd= 5.2), whereas -9.0 (sd=4.8) for SDS. A Kruskal-Wallis with
onset of the gesture as dependent variable revealed that the differences across vowel
conditions are significant, U = 51.938 p=.000. Both H12 and H13 are rejected. The
implications of these findings are discussed further in the next chapter.
128
Figure 5.16: Histograms for time-course of nasality per vowel condition and dialect.
129
130
sd
2.2
Completion
of gesture
sd
2.9
3.0
-1.6
1.5
Mid-gesture
Completion
of gesture
3.9
3.9
Onset
-6.8 4.7
of gesture
M
-0.4
Mid-gesture
Onset
-4.4 4.3
of gesture
M
2.0
0
-6.0
/e/
median
2.0
0
-3.0
/e/
median
-16
-16.0
-17.0
min
-16.0
-16.0
-17.0
min
18.0
17.0
7.0
max
16.0
13.0
10.0
max
1.5
-1.5
-6.7
M
2.2
-0.3
-4.6
M
3.8
4.0
4.9
sd
2.5
2.8
4.4
sd
2.0
0
-6.0
/a/
median
2.0
0
-3.0
/a/
median
-16.0
-16.0
-17.0
min
-16.0
-16.0
-17.0
min
15.0
14.0
8.0
max
16.0
12.0
11.0
max
-1.0
-2.3
-7.6
M
1.6
-1.0
-5.4
M
3.9
4.0
4.7
sd
2.8
3.1
4.7
sd
2.0
-1.0
-8.0
/o/
median
2.0
0
-4.0
/o/
median
-16.0
-16.0
-17.0
min
-15.0
-16.0
-17.0
min
15.0
14.0
10.0
max
18.0
16.0
6.0
max
Table 5.20: Descriptive statistics on time-course of nasality per vowel type. BAS data is presented in the top panel; SDS, in the
bottom. The values are raw averages (M ) and standard deviations (sd) that have not been adjusted by the statistical model.
Last, we examine the role of environment in anticipatory vowel nasalization:
pre-pausal, pre-vocalic and pre-consonantal. The hypothesis in H11 is that in the
environments where nasal consonant weakening obtains, onset of the velum lowering
starts earlier, and is the result of a model in which nasal weakening is a pre-condition
for anticipatory vowel nasalization. Thus, if the pre-pausal condition (/n##/) is
expected to be less weakened, it is also expected to exhibit later onset of the velum
lowering, in comparison to the pre-vocalic condition (/n#V/). Figures 5.17 presents
the data for pre-pausal and pre-vocalic environments for BAS and SDS, respectively.
Descriptive statistics are included in Tables 5.21 and 5.23. For both dialects, the 15%
threshold is crossed later in the pre-pausal than in the pre-vocalic condition. The
BAS data shows a mean onset of the gesture at -4.3 (sd= 4.6) for the pre-pausal
environment and at -7.1 (sd= 6.2) for the pre-vocalic. In SDS, on the other hand,
the mean onset of the gesture for the pre-pausal environment is -8.2 (sd= 5.2) and
-10.1 (sd= 4.3) for the pre-vocalic. These results confirm the hypothesis –the context
more prone to word-final nasal consonant weakening exhibits earlier onset of the velic
gesture and this trend obtains with both dialect groups.
131
Figure 5.17: Histograms for time-course of nasality for the pre-pausal and pre-vocalic
conditions, BAS data.
Regarding pre-consonantal conditions, because of inter-dialectal differences in
the preferred nasal consonant variant word-finally ([n] for BAS, [N] for SDS), the
hypothesis was an interaction between dialect group and environment (H11.) The
prediction here was that for BAS, the pre-dorsal environment would show earlier onset
of nasality because this is the environment where nasal consonants are expected to be
weaker. SDS is predicted to exhibit these patterns in the pre-coronal environment.
Results show that both dialects exhibit an earlier mean onset of the velic gesture
before a coronal stop (/n#t/). For BAS, the onset of the gesture is -3.4 (sd= 3.6)for
pre-coronal and -7.0 (sd= 3.7) for pre-dorsal. In the case of SDS, on the other hand,
we find that onset of the gesture has a mean of -5.2 (sd)= 4.7) in pre-coronal and -7.3
(sd= 4.1) in pre-dorsal. These findings confirm the prediction for the BAS data, but
132
reject it for SDS.
Figure 5.18: Histograms for time-course of nasality for the pre-coronal and pre-dorsal
conditions, BAS data.
Given that the hypothesis predicts an interaction between dialect and environment
condition, two Kruskal-Wallis test were run, one per dialect. The dependent variable
in both analysis was onset of the gesture. The results reveal that the differences across
environment conditions are significant for both dialects (BAS, H 2= 595.5 p=.000;
SDS, H = 280.4 p=.000.)
133
Table 5.21: Descriptive statistics on time-course of nasality for pre-pausal and pre-vocalic
environments, BAS data. The values are raw averages (M ) and standard deviations (sd)
that have not been adjusted by the statistical model.
M
Onset
-4.3
of gesture
Pre-pausal
sd median min
M
Pre-vocalic
sd median min
max
max
4.6
-2.0
-17.0
8.0
-7.1
6.2
-6.0
-17.0
11.0
Mid-gesture
0.2
2.5
0
-17.0
16.0
-2.2
5.7
0
-17.0
12.0
Completion
of gesture
2.2
2.8
2.0
-17.0
18.0
1.0
5.8
3.0
-17.0
16.0
Table 5.22: Descriptive statistics on time-course of nasality for pre-coronal and pre-dorsal
environments, BAS data. The values are raw averages (M ) and standard deviations (sd)
that have not been adjusted by the statistical model.
M
Onset
-3.4
of gesture
Pre-coronal
sd median min
M
Pre-dorsal
sd median min
max
max
3.6
-2.0
-16.0
7.0
-7.0
3.7
-7.0
-17.0
2.0
Mid-gesture
0.2
1.5
0
-15.0
9.0
-1.5
2.7
-1.0
-16.0
8.0
Completion
of gesture
0.4
3.0
0
-16.0
9.0
-1.7
3.3
-1.0
-15.0
13.0
Table 5.23: Descriptive statistics on time-course of nasality for pre-pausal and pre-vocalic
environments, SDS data. The values are raw averages (M ) and standard deviations (sd)
that have not been adjusted by the statistical model.
M
Onset
-6.8
of gesture
sd
Pre-pausal
median min
max
M
sd
Pre-vocalic
median min
max
5.3
-6.0
-17.0
10.0
-8.8
4.3
-9.0
-17.0
7.0
Mid-gesture
-2.0
4.7
0
-16.0
17.0
-3.2
4.4
-3.0
-16.0
9.0
Completion
of gesture
0.9
4.4
2.0
-16.0
15.0
0.9
4.9
1.0
-16.0
15.0
134
Table 5.24: Descriptive statistics on time-course of nasality for pre-coronal and pre-dorsal
environments, SDS data. The values are raw averages (M ) and standard deviations (sd)
that have not been adjusted by the statistical model.
M
Onset
-5.2
of gesture
Pre-coronal
sd median min
M
Pre-dorsal
sd median min
max
max
4.7
-3.0
-17.0
4.0
-7.3
4.1
-7.0
-17.0
10.0
Mid-gesture
-0.4
3.0
0
-16.0
9.0
-1.7
3.3
-1.0
-15.0
13.0
Completion
of gesture
2.2
2.6
2.0
-16.0
12.0
1.4
3.0
2.0
-12.0
15.0
5.3.0.1
Summary
This section described the time-course of vowel nasalization in pre-nasal vowels.
The findings reveal that dialect groups differ in terms of lowering of the velum, with
SDS presenting an ealier onset of the gesture. In terms of the linguistic variables,
the analysis indicates earlier nasalization with CVN syllables and the vowel /o/.
The analysis of stress reveals that while unstressed syllables show an earlier onset of
nasalization, differences between stressed and unstressed conditions are greater for
SDS than for BAS. In terms of environment, time-course of nasalization is earlier with
the pre-vocalic condition (/n#V/). Comparing the pre-consonantal environments
reveals earlier onset of the velic gesture with pre-dorsal nasal consonants for both
dialect groups.
Overall, the conditions where we see earlier lowering of the velum coincide (broadly
speaking) with the environments where the word-final nasal consonant is weakened.
The next section examines the relationship between the the time-course of nasality
and weakening of the nasal consonant.
135
5.4
Relationship between the time-course of nasality and weakening of
the nasal consonant
In order to examine the relationship between onset of the velic gesture and
weakening of the nasal consonant, duration of the segment and consonant-to-vowel
oral energy ratio were analyzed as a function of the other. The previous work reports
that Caribbean dialects present high frequencies of velarized and deleted variants of
/n/. At the same time, velarized variants have been analyzed as a weakened variant.
Thus, following the literature in Spanish linguistics we would make the prediction
that only the Caribbean dialect, Santo Domingo Spanish, would show a relationship
between the aforementioned variables. This relationship would not necessarily obtain
with the non-Caribbean dialect, Buenos Aires Spanish (H14.) But as the reader
may remember from the previous section, dialects are not different with regard to
weakening of the nasal consonant. Thus, in light of the findings thus far the alternative
prediction is that there is no relationship between the variables.
Figure 5.19 presents a scatterplot with duration of the nasal consonant in the
y-axis and onset of the gesture in the x-axis. The figure shows that, as the nasal
consonant gets longer, onset of the velic gesture takes place later, and thus, there is a
relationship between the two variables. This pattern obtains for both dialect groups
and rejects the hypothesis in H14. To test whether this relationship was significant,
two simple correlations were carried out, one per dialect. The results show statistical
significance for these patterns for both dialects (BAS data: r= .25, p < .05 [.23, .27];
SDS data: r= .30, p < .05 [.25, .30]).
136
BAS
SDS
Duration of the nasal consonant (in ms)
300
200
100
0
−10
0
10
−10
0
10
Onset of the gesture (Threshold at 15%)
Figure 5.19: Scatterplot for duration of the nasal consonant (in ms) as a function of onset
of the gesture, per dialect.
In figure 5.20, consonant-to-vowel oral energy ratio was plotted as a function of
onset of the velic gesture. The findings reveal that as the nasal consonant decreases in
ratio (i.e. it is more constricted), onset of the gesture takes place later. In other words,
the relationship between consonant-to-vowel oral energy ratio and onset of the gesture
also shows that as the nasal consonant weakens, anticipatory vowel nasalization takes
place earlier. The results show statistical significance for these patterns for both
dialects (BAS data: r= .17, p < .X05 [-.04, -.02]; SDS data: r= .13, p < .05 [-.04,
-.02]).
137
BAS
SDS
1.00
Consonant−to−vowel ratio
0.75
0.50
0.25
0.00
−10
0
10
−10
0
10
Onset of the gesture (15%)
Figure 5.20: Scatterplot for consonant-to-vowel oral energy ratio as a function of onset of
the gesture, per dialect.
5.5
Summary of the chapter
Sections 5.2, 5.3 and 5.4 provided an explanation of the results for the analyses
of weakening of the word-final nasal consonant, time-course of nasalization and
the relationship between the two. All in all, the results show little evidence of nasal
consonant weakening, as most participants produced very constricted segments. Unlike
the reports in the previous literature (sociolinguistics and instrumental), dialect groups
did not differ in terms of weakening of the nasal consonant. Linguistic variables were
shown to have a significant influence on the properties of word-final nasal consonants,
with cvn syllables, the unstressed condition and non-low vowels facilitating weakening
of the nasal consonant. In pre-vocalic condition, both dialects exhibited the most
138
weakened variants, both reflected in the durational and consonant-to-vowel oral energy
ratio. Regarding the time-course of nasalization, the findings show that velum lowering
takes place earlier in the SDS data. The linguistic variables were also shown to have
an effect in the patterns of variation. The analysis the gestural landmarks showed
that earlier onset of velum lowering obtained with CVN syllables, stressed syllables,
the back vowel and in pre-vocalic condition. Some of these environments coincide
with the environments where weakening of the nasal consonant is greater (i.e. cvn
and pre-vocalic). With the stress condition and vowel type, however, findings differ
from weakening of the nasal consonant.
Finally, exploring the relationship between weakening of the word-final nasal
consonant and velum lowering has revealed that these two acoustic cues work in
tandem, and that this pattern holds across dialect groups. In other words, as the nasal
consonant is weakened, onset of the velic gesture takes place earlier, both for the dialect
group described as preferring velar variants of /n/ and for BAS, an ‘alveolarizing’
dialect.
Taken together, these analyses portray anticipatory vowel nasalization as differing
considerably from traditional descriptions of the Spanish nasalization. Characteristics
that are typically described of Caribbean dialects were found for the non-Caribbean
dialect, and viceversa. A more in-depth analysis and explanation of the effects
described in the results chapter will be presented in the following chapter, which
contains the discussion.
139
Chapter 6
DISCUSSION AND CONCLUSION
The goal of this final chapter is to revisit the research questions that guided the
investigation, explain the extent to which the hypotheses were confirmed or refuted
by the acoustic analyses, and discuss the implications of the findings. Following the
discussion of findings by each acoustic measure and independent variables, a general
discussion of findings is offered that relates the results more broadly to important
issues in Spanish sociophonetics and laboratory phonology. A critical evaluation of
the research reported in this dissertation is then presented, along with suggestions for
future research. The final section of this chapter outlines the main conclusions drawn
from this dissertation study.
6.1
Regional variation in weakening of the nasal consonant
The first research question guiding this study is presented below:
RQ2 Do SDS and BAS differ in terms of lenition of the word-final nasal consonant?
The hypothesis was that SDS would present greater degrees of word-final nasal
weakening than BAS. The prediction was that this hypothesis would be borne out
as higher ratio of consonant-to-vowel-energy and shorter duration. There are no
considerable differences between dialect groups in terms of duration, but SDS presents
140
a higher consonant-to-vowel ratio. Inferential statistics support the null hypothesis
that groups are not different, neither in terms of degree of constriction (as indexed by
the consonant-to-vowel-energy ratio) nor temporal extent (as measured by duration).
The analysis of the production of word-final nasal consonants in this study reveals that
dialect groups are not different in terms of weakening of the segment. Figures 5.2 and
5.6, presented in the results chapter, are replicated below as figure 6.1, and depicts the
overall distribution of the data. These results, thus, support the hypothesis that SDS
weakens word-final nasal consonants just as much (or just as less, given the findings
reported here) than BAS.
Figure 6.1: Duration (in ms) and consonant-to-vowel ratio of the nasal consonant, per
dialect. Diamonds represent the mean.
The amount of evidence informing the hypothesis prior to the analysis was robust.
As I went through the results and came to understand the data, there were a number
of recurring questions. In light of the sociolinguistics literature, why does the corpus
under study present little evidence of weakening of the word-final nasal consonants?
141
Why are the Dominicans not leniting more? The remainder of this section seeks
answers to these questions. We start, first, by connecting the results to the previous
literature. In doing so, we will consider three aspects: (a) the reported differences
between Caribbean and non-Caribbean dialects, (b) the patterns of weakening of
observed in previous work and (c) methodological differences across studies. Finally,
we discuss the acoustic characterization of nasal consonant weakening.
The findings here contrast starkly with previous work. The first difference is
in terms of the reported frequencies of weakened variants. Some studies in the
sociolinguistics literature report deletion rates at 31% (others at 1.9%, though). For
the one study conducted in the Dominican Republic (in Santiago de los Caballeros, in
the Cibao region, not in Santo Domingo), Haché de Yunén (1981) had found that in
word-final position, elided variants accounted for 45% of the data (n= 489; 44.1% for
elided accompanied by nasalization and 1.2%, with no nasalization). Instrumental
work, on the other hand, does not quantify weakening in terms of frequencies of
variants, but rather characterize the degree of constriction as a function of articulatory
variables. Specifically, Colantoni and Kochetov (2012) and Ramsammy (2011, 2013)
do so by quantifying contact with the tongue and the different areas of the palate.
Even though these studies survey speakers from different dialectal areas, they have a
commonality: they both find that it is among speakers of the dialect that exhibits
velarized variants that weakening was greater(Cuban in Colantoni & Kochetov, 2012;
Galician in Ramsammy, 2011, 2013.) Specifically, Colantoni and Kochetov (2012) find
that Cuban Spanish not only exhibits a more dorsal articulation, but more weakening,
and greater speaker variability as well. These findings connect with the sociolinguistics
literature, which was reporting velarization of the word-final nasal consonant at rates
of 66% to 85% (e.g. Cedergren & Sankoff, 1975; D’Introno & Sosa, 1988; López Morales,
1980).
142
Colantoni and Kochetov (2012) also found a relationship between the duration of
the nasal consonant and the degree of constriction: the longer the nasal consonant,
the more constricted the nasal consonant is. For this reason, I went back to the
corpus under analysis, and plotted duration of the nasal consonant as a function of
consonant-to-vowel ratio both in BAS and in SDS (figure 6.2). In the present study,
the relationship obtains for both dialects. In addition, we also observe that at longer
duration values (150ms and longer), variability in consonant-to-vowel-ratio decreases
greatly, for both dialects. These patterns indicate that longer nasal consonants are
unequivocally constricted. Shorter nasal consonants show more variability.
BAS
SDS
Duration of the nasal consonant (in ms)
300
200
100
0
0.00
0.25
0.50
0.75
1.00 0.00
0.25
0.50
0.75
1.00
Consonant−to−vowel−ratio
Figure 6.2: Duration of the nasal consonant (in ms) as a function of consonant-to-vowel
oral energy ratio, per dialect.
The research reported here also contrasts with previous work in terms of the
elicitation techniques. The present study falls within what we can call ‘instrumental
143
fieldwork’ or research employing laboratory techniques to the study of sociolinguistic
variation. More broadly, we could also consider this work falling under the purview
of laboratory phonology due to the treatment of the data (i.e. acoustic measures),
but also to the design of the elicitation task.27 The early sociolinguistics literature,
on the other hand, gathered the data via sociolinguistic interviews. These two
ways to build a corpus sit on opposite ends with regard to how much attention
participants pay to speech. Thus, one issue in connecting the results reported here
to the sociolinguistics literature is to what extent are they really comparable due
to speaking style differences. The findings of this study, however, do follow findings
in the instrumental work in Colantoni and Kochetov (2012) and Ramsammy (2011,
2013) who report more weakening for the dialects with velarized variants. Though in
this study, dialectal differences in nasal consonant weakening did not reach statistical
significance.
Concerning speaking style, it is also important to consider how the task may
have elicited different behaviors among dialect groups, and thus, impacted the results.
One dialect group (i.e. SDS) may have been paying more attention to the task,
and consequently their speech, than the other (i.e. BAS). Even though it had been
emphasized that the task was not a reading test and that participants should speak
as normally as possible, some speakers of SDS were more careful in their speech (see
below for more comments on this issue.) Some ways in which this ‘attentive’ linguistic
behavior surfaced include more mistakes in reading. More tokens were discarded for
SDS for misreading the target word or stress pattern than for BAS (SDS, n =327
[2.85% of the corpus] vs. BAS, n= 41 [0.35%]). Also, this dialect group exhibited some
instances of dialect inappropriate variants, like the use of the interdental fricative, [T],
for graphemes <c, z>: a total of 13 tokens (0.1% of the corpus), produced by four
speakers (three women, and one man; DRF13, n=1, DRF16, n=4, DRF19, n= 5 and
144
DRM09, n= 3.)
Other ways in which we see heightened attention surface relate to hyperaticulation.
Hyperarticulation involves pronouncing words more clearly than they are “normally”
produced, and is associated with various acoustic-phonetic features of enhanced speaker
effort, such as longer durations and enlarged vowel spaces (Lindblom, 1990). One
indicator of hyperarticulation in the data set is the production of the word-final /s/.
While both SDS and BAS are /s/-weakening dialects, SDS is at a more advanced stage
than BAS (Hualde, 2014, pp. 157–162). Recall that word-final oral consonants were
all /s/. It was expected that both dialects would produce lenited variants (aspirated or
elided, for example), with SDS exhibiting more elided variants than BAS. Figure 6.3
presents box, jitter and violin plots for duration of /s/ for both. During annotation,
elided variants were labeled and duration was set at 0ms. Mean differences between
dialects are not different, but SDS presents more fully elided variants (n= 85 for SDS
[1.5% of the SDS corpus] vs. n= 25 for BAS [0.45% of the BAS corpus]), as well as
longer variants. In fact, these productions are reminiscent of hypercorrected variants
described for journalists in Dominican television (Alba, 2009a).
145
Figure 6.3: Duration of word-final /s/ (in ms), per dialect. Diamonds represent the mean.
The lack of weakening, especially for SDS, can be attributed to speakers’ heightened
attention to speech. As a result, they were articulating each segment clearly and
produced well constricted nasal consonants. In section 6.5, we discuss the relationship
between laboratory speech and sociolinguistics and comment further on this issue.
But before doing so, we will consider the causes of these differences in behavior. On
one hand, dialect groups may diverge in terms of their linguistic (in)security (Alba,
2004, 2009b; Büdenbender, 2010; Bullock et al., 2014; Toribio, 2000). Added to these
cross-dialectal differences, the task itself –i.e. reading and the use of the nasometer–
influenced their behavior. On the other hand, I was a local in one community and not
146
the other. Argentines have been reported in previous research to use the interdental
fricative as well (specially in tasks involving reading), but in this work the researcher
was a not local himself (Rohena-Madrazo, 2015, p. 302 and p. 304). In the study
reported here, the same pattern surfaced with the dialect group where the researcher
(i.e. myself) and participants mismatched, which suggests a trend for speakers to
attempt what they see as a supraregional norm (probably influenced by alphabetization
practices). Even though care had been taken during the development of the protocol
by, for example, not employing “traditional” carrier phrases such as “I say [target
word] for you” to avoid repetitive reading and to distract participants from the purpose
of the study,28 these findings do show a tension between experimental approaches
and sociolinguistic methods and point to the need to probe the linguistic system via
multiple elicitation tasks. Adding a comparison of speaking styles is warranted and
constitutes an important adjustment in future research. Additionally, when working
with different dialect groups it will be important to either have non-local researchers
interact with participants at all times, or recruit a local in the community when the
researcher is an outsider, regardless of their L1.
Having made these considerations on the drawbacks of instrumental work, it is
also important to consider how these findings connect with previous work. In light
of the lack of weakening in the corpus under study, we may also wonder to what
extent the previous literature may not have over-reported nasal consonant weakening
to some degree. We have two reasons to consider this alternative interpretation
of the results. In the sociolinguistics literature, the researchers were relying on
perceptual skills to accurately characterize productions. This is not to say that this
work is not valuable. But rather, it is well established that instrumental analyses
remove “the problem created when binary classifications are imposed on gradient
data” and “reduces possible experimenter bias, and allows the results of a study to be
147
more easily replicated by other researchers.” (Eddington, 2011, p. 2). Additionally,
nasal consonants are notoriously known for their perceptual consequences. Work in
perception has not only well documented misperception of vowel height (Beddor, 1993;
Beddor et al., 1986), but also misperception of place of articulation (Goodin-Mayeda,
2016; Herrera, 2002; Liberman, Delattre, Cooper, & Gerstman, 1954; Malécot, 1956;
Narayan, 2008). A key aspect about the perception of nasal(ized) segments as it relates
to the research reported here is that the fact that the further back the articulation of
the nasal consonant (e.g. alveolar vs. velar) the more it resembles a nasalized vowel
in perception (Ohala & Ohala, 1993). The sociolinguistics literature may have been
reporting deoccluvized nasal consonants (i.e. weakened variants) as nasalized vowels
when in fact they were constricted (c.f. Goodin-Mayeda, 2016, p. 56). Instrumental
research may also have been contributing to the percept of pervasive weakening. While
EPG can accurately depict place of articulation and degree of contact, its precision
wavers when the constriction is located towards the posterior end of the vocal tract.
Colantoni and Kochetov (2012, p. 21) in fact make the disclaimer that “velar closure
is not always detected by the EPG palate” in reference to the vocalized variants they
report. All in all, the findings in this study challenge the frequency of weakening of
word-final nasal consonant as reported in previous work.
Nevertheless, we also must consider if and how this study under-reporting wordfinal nasal consonant weakening. Methods such as electropalatography capture degree
of contact with greater precision (in addition to place of articulation) than acoustic
measurements. But acoustic characterizations of weakening have long been used in
Spanish and Romance linguistics (Colantoni & Marinescu, 2010; Cole, Hualde, &
Iskarous, 1999; Dalcher, 2008; Hualde et al., 2011; Kingston, 2008; Lavoie, 2001, among
others) for example with /b d g/. Collectively, this work has established duration was
a reliable cue to weakening and so was the relative consonant-vowel intensity-ratio.
148
While work examining /b d g/ employs these methods consistently and systematically,
one issue with this measurement, and its extension in the research reported here, is
that it does not permit to tease apart consonant reduction from vowel reduction. That
is, is the ratio the result of the nasal consonant weakening or the vowel reducing?
The analysis reported here does not allow to answer this question. Future work will
benefit from combining methodologies such as EPG and nasometry or airflow, such
that articulatory data is combined with acoustic/airflow data. This approach will not
be able to address some of the issues raised above related to formal speech, but it
would permit to further test some of the claims in the study reported here.
Concerning the validity of the acoustic characterization of consonantal weakening,
during annotation of the corpus two trends of weakened tokens of /n/ were observed:
cases with no oral constriction at all, and cases with a ‘weaker’ constriction. If these
two types of tokens are combined, the total is 89 observations, which represents 0.8%
of the data (of these weakened variants, 20 correspond to the BAS data, and the
remaining 69 to SDS.) When compared to the findings for consonant-to-vowel-energy
ratio, these results are not surprising: the bulk of the data occupies the 0-0.3 region of
the distribution of the ratio (n= 6896 or 93.5% of the data). Thus, the measurements
of degree of contact go in hand with the qualitative annotation of the corpus.
The following section address the regional differences in the time-course of nasality.
6.2
Regional variation in time-course of nasality
The second research question guiding this study is repeated below:
RQ1 Do SDS and BAS differ in terms of the time-course of nasality?
The hypothesis was that SDS would present more anticipatory vowel nasalization.
As such, SDS was predicted to show earlier onset of the velic gesture. Descriptive and
149
inferential statistics indicate that SDS presents earlier onset of lowering of the velum.
Figure 5.11, presented here again as figure 6.4, illustrates the overall distribution of
the data. These findings mean that anticipatory vowel nasalization in SDS is more
extensive than with BAS, and thus confirm the hypothesis that SDS exhibits an earlier
onset of the velic gesture. In this section, we first connect the findings to previous
work in nasalization, such as Cohn (1993). We then comment on expected findings
uncovered, and finally, we consider some of the assumptions going into this research.
Some of the patterns reported here fit with a model like the one in Cohn (1993).
Cohn (1993) presents a model in which the phonetic realization of [nasal] follows a categorical phonetic realization (applies to most of the segment; exhibits a plateau-shaped
curve), or gradient (increases over time; exhibits a steady cline). This distinction
allows for a characterization of the phonetic realization of nasal vowel phonemes (as
in French) and nasalized vowel allophones (like in English). The French data in Cohn
(1993) showed that when the vowel is specified as [-nasal] (i.e. the vowel in bonne
tête ‘good head’ /bõnt(Et)/), there is no significant nasal airflow during the vowel.
Conversely, when the vowel is phonologically nasal ([+nasal], as with nasal in bon nez
‘good nose’ /bõn#(e)/), airflow was present for most of the duration of the vocalic
segment. With English, on the other hand, the vowel is not specified for [nasal]. As a
result, the phonetic realization was gradient, increasing throughout the duration of the
segment. As is, this model does not account for the dialectal differences in Spanish,
given that we have to account for the phonetic realization of two types of non-phonemic
nasal vowels: phonetic and allophonic. One way we can make the accommodation
is by requiring BAS to specify the vowel as [-nasal]. As such, SDS, which exhibits
earlier onset of nasalization, does not specify for [nasal], and by anticipating onset of
velum lowering, presents a cline-like pattern of nasalization. BAS, on the other hand,
specifies the pre-nasal vowel as [-nasal], with onset of velum lowering late such that
150
“non-nasalization” affects most of the segment. One way to test for this interpretation
in future research is by replicating, to the extent possible, the protocol in Solé (1992),
where rate of speech is used to differentiate between ‘unintended’ (i.e. phonetic) and
‘intended’ (i.e. allophonic) nasalization. Whether the proportion of the vowel that is
nasalized stays constant or varies would serve as additional evidence of the dialectal
differences we report here.
Figure 6.4: Time-course of nasality per dialect. Dashed lines represent the mean.
Even though the hypothesis was confirmed, there are a number of ways in which
the findings in this study were unexpected. First, the mean timepoint of onset of the
velic gesture for SDS does not align with the previous work. Lederer (2000, cited in
Lederer, 2003) reports that nasal airflow with Cuban Spanish begins at the release of
the onset and increases during the duration of the vowel until it peaks at the closure
of the nasal consonant. In the corpus under analysis here, the mean onset of the velic
gesture for SDS data takes place during the second half of the vowel, and not earlier.
Thus, we expected onset of the velum earlier for this dialect group.
151
Some of the findings for BAS were also unexpected. About 10% of the BAS data
(n = 1,116 tokens) exhibit onset of velum lowering at time point -7.0 (the mean for
SDS) or earlier. What this means is that a considerable portion of the BAS data is
patterning with SDS. This may be due to some linguistic variables (see next section)
or it may be that some BAS speakers are early “onsetters.” We are not analyzing
individual data presently, but it is an avenue for future research certainly worth
considering.
So, to recap, findings confirm the hypothesis that the Caribbean dialect (i.e. SDS)
would exhibit earlier onset of the velum than the non-Caribbean dialect i.e. BAS). At
the same time, SDS was expected to exhibit an even earlier onset of the gesture than it
exhibits here. And BAS was expected to present the bulk of the data clustering during
the last portion of the vowel, and not present such a wide range in the distribution
of onset of velum lowering. How do we reconcile the findings with the expectations
going into this research? Much of our predictions were informed by Lederer (2000, as
reported in Lederer 2003), but we have not been able do a critical reading of Lederer’s
work because it was not possible to access the manuscript. Thus, the extent to which
our findings connect to Lederer (2000) is not clear.
Firstly, differences in methodological decisions may have also contributed to the
mismatch between predictions and findings. First, operationalization of the onset of
the gesture may be different across studies. In this study, it was defined as the time
point at which nasal energy crossed a threshold of 15%. We do not know how was
onset of the gesture operationalized in Lederer (2000). If we had been able to put
studies next to each other, we would be able to compare, for example, if thresholds
were equally conservative or equally sensitive. In the absence of this information, we
can only speculate.
Another layer to consider relates to the dialect-pairing. Lederer (2003) presents
152
differences between Cuban and Peninsular Spanish in the 2000 study as clear-cut:
one overlaps with English, the other one does not. We are left with no information
as to within-dialect variation, if any. The data presented here shows within-dialect
variation, especially for the BAS data. How does BAS compare to Peninsular Spanish?
The fact that BAS presents many instances of early onset of the velic gesture suggests
differences in anticipatory vowel nasalization with alveolarizing dialects of Spanish as
well, and not just between broadly defined regional varieties. Therefore, the study
reported here may have uncovered patterns of early nasalization that had not been
described before, much less accounted for in discussions of Spanish, nasalization and
sound change. The bulk of the previous work focused on Caribbean dialects and may
have glossed over patterns taking place elsewhere. We can now ask ourselves: is BAS
an innovative dialect, just not one with a preference for velar variants?
This last question additionally challenges how dialects of Spanish have been
described as ‘conservative’ and ‘innovative’. The point of departure in this doctoral
project had taken Caribbean as the ‘innovative’ region and elsewhere as maintaining
the status quo. But it is really so? In our conversation about Spanish, word-final
nasal consonants and nasalization (or at least how it has been framed here), have we
been conflating ‘innovative’ and ‘conservative’ with ‘velarizing’ and ‘alveolarizing’?
Ultimately, this is an empirical question. A way to test for differences between
velarizing and non-velarizing dialects, as well as ‘innovative’ and ‘conservative’ dialects
is to add dialect groups for comparison. For example, such an approach could
survey (in a series of studies) two Latin American velarizing dialects (e.g. Dominican
and Cuban Spanish), Latin American non-velarizing dialects (e.g. Argentine and
Colombian Spanish), Peninsular velarizing dialects (Galician and Andalucian Spanish)
and Peninsular non-velarizing dialects (Northern and Southern Castillian Spanish).
This work will not only further our understanding of patterns of nasalization in Spanish
153
but will also encourage conversations about how we talk about variation and whose
variation we chose to label, study and why.
6.3
The influence of linguistic variables
The third and fourth research questions pertain to the effects of four within-
subjects variables on the production of vowel nasalization and nasal consonant weakening: syllable type, vowel type, lexical stress, and environment. The goal of including
these variables was to examine the effect of linguistic factors that had been described
in the previous literature as influencing the patterns of variation in anticipatory vowel
nasalization and word-final nasal consonant weakening. These questions are presented
again below:
RQ3 How do syllable type, environment, vowel type and lexical stress affect the timecourse of nasality?
RQ4 How do syllable type, environment, vowel type and lexical stress affect lenition
of word-final nasal consonants?
The analysis of the linguistic variables revealed some patterns of variation, both
for nasal consonant weakening as well as the time-course of nasalization. Overall,
differences between means were small, yet statistically significant, which suggests that
these effects are small and very consistent across subjects. We will now comment on
each of the linguistic variables, starting with syllable type.
One hypothesis was that there would be no effect of syllable type on the weakening
of the nasal consonant (H3.) We also had hypothesized that NVN would be weakened
due to the presence of an onset nasal consonant (H4.) The findings show that
word-final nasal consonants in NVN syllables are indeed longer, but exhibit higher
consonant-to-vowel ratios.
154
The findings of higher consonant-to-vowel energy ratio for NVN than for CVN
may be an artifact of the calculation itself. In as much as the measurement is expressed
as a ratio, the value can increase or decrease by virtue of consonant reduction, vowel
reduction or both. Thus, with NVN, the consonant-to-vowel ratio may be higher
because the vowel is reduced. One way to test for this possibility is to consider
differences between CVN and NVN in terms of duration of the vowel. This analysis
reveals that, indeed, the vowel is shorter with NVN than it is with CVN. Figure 6.5
presents box, jitter and violin plots for the duration of the vowel in CVN and NVN.
With CVN, the vowel segment has a mean of 84.4ms, whereas with NVN, 79.1ms.
Thus, it seems more likely that we can reject the hypothesis in H4 and say that, with
CVN syllables, the nasal consonant is weakened, given the durational measurements.
Figure 6.5: Duration of the vowel (in ms) per syllable type. Diamonds represent the mean.
Regarding onset of the velic gesture as operationalized here, NVN exhibited
later onset of velum lowering than CVN. The hypothesis in H10 stated that, with a
155
preceding nasal consonant, velum lowering could take place during the onset nasal and
not be raised again, such that the whole vowel would be nasalized. This hypothesis is
connected with Cohn (1993) in that such an effect would be the result of the vowel
not being specified for [nasal], neither as [+nasal] nor as [-nasal]. Applied to the
data in this study, the findings for NVN suggest that vowels are indeed specified as
[-nasal] and onset of velum lowerinf is delayed to achieve an oral vowel. These patterns
are specially visible with the BAS data, where mean onset of the velum for NVN is
located at time point -3.0, and provide additional evidence to the dialectal differences
discussed in the previous section.
Concerning stress, the hypothesis was that stressed syllables would be less weakened than unstressed ones (H8.) In the case of the time-course of nasality, we had
hypothesized that stressed syllables would exhibit a later onset of the velic gesture
(H14.) The results confirm all hypotheses: in the stressed condition, the nasal consonant is less weakened and velum lowering starts earlier.
The fact that all hypotheses for stress are confirmed challenges narratives of
anticipatory vowel nasalization as a compensatory mechanism. The analysis of the
data shows more extensive nasalization when the nasal consonant is less weakened.
These findings suggest an alternative explanation. Stress functions as a “pointer” by
indicating which information in an utterance is most important. Articulatorily, a
stressed syllable is usually produced with greater physical effort. This means more
enhanced gestures and an increase in the activity of the respiratory muscles (Ladefoged,
2005; Raphael, Borden, & Harris, 2011). As a result, nasal consonants are longer
and more constricted in the stressed condition. At the same time, lowering of the
velum is also enhanced and onset of the gesture is elicited earlier. If we look at the
results of stress across dialects, these are larger for SDS than for BAS, and provide
substantiation that vowel nasality in the pre-nasal vowel has a different phonological
156
status for SDS and BAS. Since for this dialect group nasality is an intended part of
the vowel, and not a mechanical by-product of co-articulation, temporal differences in
velum lowering across conditions are magnified for SDS (and these differences across
dialect-groups are statistically significant.) Additionally, the results show that fully
constricted nasal consonants and earlier anticipatory vowel nasalization need not be
divorced and anticipate the discussion of the results regarding the relationship between
weakening of the nasal consonant and the time-course of nasality in section 6.4.
We will now consider the effect of vowel type. Regarding weakening of the
nasal consonant, we had hypothesized no effect on the production of word-final nasal
consonants (H7.) The analyses of the data show that, when the vowel is /a/, the wordfinal nasal consonant is the shortest, and when the vowel is /e/, the word-final nasal
consonant is longer and exhibits a higher consonant-to-vowel energy ratio. We will
connect these findings to the previous work and to issues related to the methodology
employed in this study.
Comparing the findings here to the previous work is complicated due to methodological choices. In the sociolinguistics literature, the findings in Cedergren and
Sankoff (1975) indicate that high vowels trigger greater frequencies of nasal consonant
effacement. However, this study did not include high vowels, due to lacunae in the
lexicon (i.e. few unstressed high vowels in pre-nasal word-final position). In the
instrumental literature, Colantoni and Kochetov (2012) found that, in the word-list
task, de-occluvized nasal consonants were more common in between two back rounded
vowels /o/. Given that the research design in this study does not include this context
(i.e. /on#o/), comparing across studies is not straightforward.
With regard to the methodological choices, the contradictory findings may point
to issues with the acoustic measurements. While taking durational measurements is
straightforward, one issue with the consonant-to-vowel energy ratio is that, as already
157
mentioned, it does not permit teasing apart the effects that are due to nasal consonant
reduction, vowel reduction, or both. This observation provides additional support to
my comment above that future work will be well served by combining articulatory
methods, such as EPG, with acoustic or airflow measurements, like nasometry or nasal
airflow, respectively.
For the time-course of nasality, the previous literature presented contradictory
information. We had hypothesized that the low vowel /a/ would be more nasalized
than the mid vowels, and for this reason, we had predicted that /a/ would exhibit
earlier onset of velum lowering (H12.) At the same time, in H13, we had predicted
no effect of vowel type due to inconsistent findings within and across studies. The
results reveal that /o/ is the vowel with earlier lowering of the velum, and differences
across vowel types are small (in the range of one time point or less, yet statistically
significant.) Therefore, both hypotheses are rejected. One possibility for these findings
is that the differences in nasalization patterns do not vary as a function or vowel
type per se, but that variation is better described as a distinction between high and
non-high vowels, which the present research protocol does not allow to examine as
mentioned above.
Finally, we consider the effect of environment, that is, whether the word-final
nasal consonant was at the end of an utterance (pre-pausal), before a vowel (prevocalic), or before a consonant (pre-consonantal). The purpose of manipulating the
environmental condition was to examine anticipatory vowel nasalization and weakening
of the nasal consonant in contexts that had been described almost exclusively in the
nasal consonant literature (e.g. Colantoni & Kochetov, 2012; Ramsammy, 2011).
A crucial difference between these environments is that with pre-pausal and prevocalic, dialect specific processes (e.g. alveolar vs. velar variant) are expected, and
in the pre-consonantal environments assimilated variants are expected, based on the
158
sociolinguistics literature and recent instrumental research (Colantoni & Kochetov,
2012; Ramsammy, 2011). Thus, this variable allows to test further the relationship
between processes of weakening and the extent to which they are accompanied by
vowel nasalization.
With pre-pausal and pre-vocalic environments, the hypothesis was that in the first
nasal consonants would exhibit less weakening than in the pre-vocalic environment.
The prediction established that nasal consonants would be longer and more constricted
before a pause than before a vowel (H5). Findings confirm the hypothesis: in pre-vocalic
condition word-final nasal consonants are shorter and exhibit a higher consonant-tovowel ratio. Overall, these effects are consistent with previous instrumental findings
on Spanish (Colantoni & Kochetov, 2012; Ramsammy, 2011).
With regard to the time-course of nasality, the previous literature did not provide
a basis to make predictions. Thus, we can only hypothesize that it is in the contexts
in which nasal consonant weakening was greater, following a model of a compensatory
mechanism, that anticipatory vowel nasalization would be greater. As such, the
prediction was that the pre-vocalic environment would present earlier onset of velum
lowering than the pre-pausal environment. The results show that in the former lowering
of the velum takes place earlier than in the latter, and thus confirm the hypothesis.
In the case of the pre-consonantal environment, it was hypothesized that the
nasal consonant would exhibit more weakening when place of articulation of the
following consonant was different from the nasal consonant (H6). For BAS, nasal
consonants would be shorter and less constricted in the pre-dorsal environment. The
same results would obtain for SDS in the pre-coronal environment. The predictions
for pre-consonantal environments were confirmed for one dialect group but not the
other. In terms of duration, consonantal environments are not different. However, in
the pre-dorsal environment, BAS data shows a slightly greater consonant-to-vowel
159
ratio which suggests that in this environment nasal consonants may in fact be more
weakened for this dialect.
The patterns for SDS, on the other hand, are the same as with the BAS speakers. However, we predicted that it would be in the pre-coronal environment where
weakening of the nasal consonant would be greater. The possibility of this dialect
group hyperarticulating may bear on these patterns (or lack thereof). Heightened
attention to speech may have encourage speakers to produce more ‘standard’ variants.
As a result, we observe the same patterns across dialects, and not the interaction
we had anticipated. The acoustic analysis performed here allows to characterize
weakening of the nasal consonant and the time-course of nasalization, but not place of
articulation. While instrumental work examining place of articulation does document
that ‘alveolarizing’ and ‘velarizing’ dialects assimilate the nasal consonant to the place
of articulation of the consonant that follows (Colantoni & Kochetov, 2012; Ramsammy,
2011), the exact articulatory details and how they connect to the patterns of nasalization cannot be determined with the present research design. Future research can
expand on this idea by combining data collection with EPG and nasometry.
All in all, some of the findings for the effect of the linguistic variables under
analysis here provide further support to the dialectal effects reported in the previous
section. Namely, differences across dialects in stress condition and type of syllable
provide evidence that anticipatory vowel nasalization does have a different phonological
status in each dialect.
6.4
The relationship of time-course and duration of the nasal consonant
The final research question under examination in this study concerns the relation-
ship between the time-course of nasality and weakening of the nasal consonant. The
160
fifth research question in reiterated below:
RQ5 Is there a trade-off between the time-course of nasality and lenition of word-final
nasal consonants?
Yes, there is. One of the main findings of this study concerns the relationship
between the time-course of nasalization and the weakening of the nasal consonant.
In order to examine this co-variation, we analyzed onset of velum lowering as a
function of duration of the nasal consonant and consonant-to-vowel energy ratio. Since
Caribbean dialects had been described as presenting the weakened variants that lead
to the aforementioned nasalization patterns, it was expected that only SDS would
show that as the consonant weakened, nasalization in the vowel started earlier (c.f.
Beddor, 2009). The findings revealed that the trend of shorter/more weakened nasal
consonants leading to earlier nasalization (and vice-versa) obtained not only for SDS,
but for both dialects, regardless of whether, in previous research, they exhibited a
preference for alveolar or velar variants. Had we not analyzed vowel nasalization and
nasal consonant weakening together and across dialects of Spanish, these patterns
would not have been uncovered.
These patterns speak more broadly as to how the continuum of lenition and the
envelope of variation has been operationalized in previous work. In Spanish linguistics,
allophonic vowel nasalization has been conceptualized as a compensatory mechanism.
Under this view, anticipatory vowel nasalization is extended after weakening of
the nasal consonant. In many of these descriptions, be they the primary source
(e.g. Lederer, 2003; Piñeros, 2006) or review chapters (e.g. Campos-Astorkiza, 2012;
Colantoni, 2011b; Hualde, 2005), velarization is part of the process of weakening.
However, the data analyzed and presented here suggests otherwise. Along the lines
of Beddor (2009), nasalization in the vowel and weakening of the nasal consonant
161
do work in tandem. But, we can have nasalization with no weakening of the nasal
consonant or even without velarized variants. Thus, while the nasal consonant is
the co-articulatory source, nasalization in the vowel is need not be a compensatory
mechanism.
The observation of anticipatory vowel nasalization in dialects of Spanish with a
preference for alveolar variants, and the patterns of co-variation with nasal consonant
weakening extending to them, had not been reported before. Why is it so? There
are probably several conflating reasons. The work that established the envelope
of variation (Cedergren & Sankoff, 1975; D’Introno & Sosa, 1988; Haché de Yunén,
1981; López Morales, 1980; Terrell, 1975) was only examining Caribbean dialects
of Spanish. To my knowledge, no such descriptions were made for other dialects.
At the same time, work comparing across dialects, such as Colantoni and Kochetov
(2012), Ramsammy (2011, 2013) and Lederer (2000, cited in Lederer, 2003), were
looking at the production of either nasalization or word-final nasal consonants, and
did not analyze them together. Taken together, these facts account for patterns going
unnoticed and suggest that future research must include the analysis of anticipatory
vowel nasalization and word-final nasal consonants as different variables.
One final aspect we will consider in this section concerns sound change. The
descriptions referenced above that present vowel nasalization as the result of nasal
consonant effacement fit into a model of a pull-chain mechanism (Martinet, 1952). In
a pull chain mechanism, one change creates a hole in the phonemic pattern which is
followed by another change which fills the hole by ‘pulling’ another sound into the
newly vacated space. Applied to anticipatory vowel nasalization, weakening of the
nasal consonant ‘pulls’ the change in nasalization in the vowel. The alternative is a
push chain mechanism. In this case, a sound starts encroaching the area of another
sound and sets off a chain reaction. Previous narratives of Caribbean vowel nasalization
162
fit into pull chain mechanism (as well as the model that Hajek, 1997, labels as DELNAS model). However, the data presented here supports the latter model: the nasal
consonant is the co-articulatory source that sets off anticipatory vowel nasalization in
the vowel and suggests that weakening of the nasal consonant (be that in duration,
weakening or change in place of articulation) is not a mandatory pre-requisite for
phonologization of anticipatory vowel nasalization. In the data analyzed here, earlier
onset of velum lowering obtains even with fully constricted nasal consonants. Thus, in
moving forward, we should consider (and continue to examine) vowel nasalization in
Spanish as a pull chain mechanism. In other words, I am suggesting to change the
approach and explore anticipatory vowel nasalization and word-final nasal consonants
as two different variables in future work.
6.5
Instrumental analyses and sociolinguistics
This study contributes to the conversation about laboratory approaches (or more
broadly, instrumental work) in the study of sociophonetic variation and the role of
cross-dialectal comparisons (e.g. Campos-Astorkiza, forthcoming; Colantoni, 2011b).
Laboratory speech has been described as unnatural, over-planned and “unreal.” As
such, laboratory research is in opposition to sociolinguistic research. From this point
of view, it is only by examining spontaneous speech that we can understand the nature
of language.
My position is that taking such an approach has the potential of hindering research.
This is not to say that laboratory research is and should be a replacement for collecting
vernacular data. My position is that sociolinguistic and laboratory approaches are, or
should be, in a symbiotic relationship. Laboratory approaches, whether taken in a
broader or more narrow sense, further our understanding of how universal and language-
163
specific patterns shape specific sound changes. The use of experimental techniques to
develop studies and treat data allow for accurate characterizations of phonetic and
phonological variation that would otherwise go unnoticed. Sociolinguistic research, on
the other hand, has an impact on the way experimental studies are designed. More
importantly, sociolinguistics connects the variation in the signal to people’s use of
language, for social differentiation, projection of one’s identity and/or categorization
of that variation. It is only by combining approaches that we will be able to probe the
linguistic system from multiple angles, and thus, better understand the underlying
mechanisms of human language, and how they interact in variation and change.
The research reported here examines the language-internal processes and compares
them across dialects. Studying parallel linguistic systems allows for an examination of
how systems may (or may not) diverge. Nasality provides a crucial example of the
contribution of an instrumental approach to the study of linguistic variation. Auditory
transcriptions were unable to capture the time-course of nasalization or accurate
differences between nasal consonant variants to the same extent that laboratory
approaches have, for example in Colantoni and Kochetov (2012), Ramsammy (2011,
2013). The comparison of Argentine and Dominican Spanish in this doctoral project
uncovered patterns that had not been reported before. For example, a portion of the
Buenos Aires Spanish data exhibits onset of the velic gesture similar to the Santo
Domingo data. Additionally, the pattern of nasalization with fully occluded or alveolar
nasal consonants was not part of previous conversations of sociolinguistic variation in
Spanish. Had it not been for the instrumental approach taken here, we would remain
in the dark. As such, the findings in this dissertation not only challenge previous
descriptions but also long held assumptions about how we talk about linguistic
innovation among Spanish dialects, at least when it comes to anticipatory vowel
nasalization.
164
The comparison of phenomena across dialects provides a window into the emergence or loss of a linguistic feature and the range of the variation that leads to it.
Small deviations in pronunciation may lead to changes in the sound structure of a
language. Within- and across-regional differences provide insights into the details of
variability itself in parallel systems. In doing so, they reveal variation patterns that
allow to test hypotheses about language differences and language change. Dialectal
variation provides critical information on the mechanisms underlying these processes.
6.6
Limitations
Despite the interesting and important findings described in chapter 5 and discussed
in the previous sections, the present investigation has some limitations. Some of these
were addressed in the relevant portions of this discussion. In this section, we address
other limitations that have not been discussed thus far, and describes how they might
be addressed in future studies.
The elicitation task utilized in this investigation was designed to examine the
production of anticipatory vowel nasalization and word-final nasals in different phonological environments. In order to control for a variety of linguistic variables (e.g. stress
and vowel type), less frequent words were used, such as canón ‘cannon’ and santiamén
‘in a flash’. Most speakers had no difficulties in reading these words.For other speakers,
lack of familiarity lead to reading mistakes with these two words, especially with canón
read it as [kanón] instead of [kánon], and were discarded from the analysis. There
were other instances of misreadings as well. For example, in some cases, there were
minimal pairs such as ganan ‘they win’ ganen ‘may they win’, and some speakers read
one for the other.
Another aspect to consider is whether, and to what extent, frequency of the lexical
165
item may have affected the patterns of variation. The instrument could have been
designed to control for lexical frequency by using nonce words. This was the approach
taken in Ramsammy (2011, 2013): he used nonce words that imitated verbs, such
as ditén. However, as explained in chapter 4, I prioritized providing ‘less repetitive’
carrier phrases to minimize the effects of “list reading” and to manage the differences
in linguistic behavior across dialects I already expected. As no formal measures of
lexical frequency or neighborhood density were calculated for the words included
in this study, it is not possible at this time to analyze the extent to which these
characteristics influenced production of anticipatory vowel nasalization and word-final
nasal consonants. Based on the results in studies such as Scarborough (2013) where
words with lower relative neighborhood frequency are produced with greater degree of
anticipatory vowel nasalization. The corpora utilized in future studies could carefully
control for frequency and density in order to compare the production of anticipatory
vowel nasalization and word-final nasal consonants in high and low frequency words
that belong to dense and sparse neighborhoods.
The experimental design in this study did not allow to test for differences across
different morphological classes. Word-final /n/ constitutes a plural verbal marker
in Spanish (i.e. third person plural, canta ‘s/he sings’ > cantan ‘they sing’.) In
future work, it would be also important to connect the processes of anticipatory vowel
nasalization and weakening of word-final nasal consonants with the morphological
category of the word. An extension of the work presented here is to examine whether
these processes show different patterns when the target word is a noun or a verb.
Thus, pursuing this work would connect variation in the phonological system with the
grammatical system.
At different points throughout the manuscript we have commented on dialectal
differences in attention paid to speech and how they may have had an effect on the
166
results. These different behaviors are probably a combination of linguistic (in)security
and the observers’ paradox (i.e. the use of the nasometer and my status as a member
–or not– of the community). For many readers, this fact may be a limitation. However,
all research suffers from the observer’s paradox. While we can try to minimize for
its effect, like it is the case for sociolinguists, we cannot totally avoid it (Thomas,
2010, p. 3). One way around it is to capitalize on it and elicit multiple speaking
styles, for example with multiple tasks (see section below). Additionally, it would be
worth considering the effect of the interlocutor and enlist the help of a local in the
community to elicit speech.
6.7
Evaluation and recommendations for future research
The study reported in this dissertation makes several contributions to the field of
phonetics and phonology, and Spanish in particular. Colantoni (2011b) calls to extend
empirical coverage of phenomena covered, diversify methodological approaches and
compare across dialects of Spanish. The current doctoral project accomplishes each
tenant: it explores a domain that has lagged behind in instrumental work (nasality),
employs innovative methodology (use of the nasometer) and compares across dialects
(Santo Domingo and Buenos Aires Spanish).
The cross-dialectal design of the study reported in this dissertation represents
another advantage, as it permits the observation and discussion of variation patterns.
This dissertation provides evidence in favor of the reports of extensive anticipatory
vowel nasalization for Caribbean dialects of Spanish. The findings regarding the
production of the word-final nasal consonant, on the other hand, challenge descriptions
of the process as compensatory. It is not my claim that Caribbean dialects do not
present weakening of the word-final nasal consonant. However, one of contributions of
167
this study is that it uncovered that early anticipatory vowel nasalization obtains as
well with fully occluded when speakers produce non-weakened variants, be that fully
occluded or alveolar segments.
Another asset of this doctoral project is the examination of anticipatory vowel
nasalization together with weakening of word-final nasal consonants. By combining
analyses, we were able to uncover patterns of nasalization in the absence of nasal
consonant weakening. We were able to challenge the implications at play in long
held narratives regarding of word-final nasals. Of course, there are still aspects of
the patterns of variation to uncover. For instance, how these patterns play out in
spontaneous speech is unknown. Future research may also benefit from the inclusion
of a variety of task types in their methodological design. A combination of tasks will
allow to resolve the tension between aiming for naturalness in speech and achieving
replicability (Thomas, 2010, p. 3). Lastly, research on more dialectal groups would
also be beneficial, as a detailed description of the full scope of variation in anticipatory
vowel nasalization is not possible with the speakers sampled for this study. As
mentioned in section 6.1 above, a study that compared dialects with a preference
for different word-final nasal variants from different regions provide opportunities to
examine patterns of innovation. Incorporating these lines of work, or as many of them
as possible, will further our understanding of dialectal differences in gestural overlap
and weakening processes regarding nasality.
Another avenue of work constitutes the examination of carry-over nasalization.
This dissertation only examined one context: when the nasal consonant is word-final.
Much less work has commented on nasalization when the nasal consonant occupies
the onset of the syllable. Work comparing anticipatory and carry-over nasalization
has uncovered that languages with limited anticipatory vowel nasalization tend to
exhibit larger carryover effects. This is true for languages like French, which has nasal
168
vowel phonemes, and a language like Greek, which has been described as exhibiting
phonetic nasalization (Delvaux et al., 2008; Diakoumakou, 2004). This body of work
has posited two hypotheses for this asymmetry. One is related to perceptual differences
between carryover and anticipatory nasalization. Delvaux et al. (2008) hypothesize
that anticipatory vowel nasalization over time is perceived as a property of the vowel,
thereby leading to phonologization. Carry-over nasalization, on the other hand, is not
attributed to the vowel and mainly contributes to the percept of the preceding nasal
consonant. The other option to explain these asymetrical patterns, argued for by
Diakoumakou (2004), concerns language-specific prosodic organization. Diakoumakou
(2004) also posits that languages with a preference for open syllables (such as Greek),
tend to exhibit more carryover nasalization than anticipatory vowel nasalization.
An examination of carryover and anticipatory vowel nasalization in Spanish would
contribute to this body of work. Like Greek, Spanish has been previously described
as having limited anticipatory vowel nasalization (i.e. phonetic, and not allophonic).
Since the data presented here confirms differences in anticipatory vowel nasalization
across dialects of Spanish, a comparison of SDS and BAS would permit exploring some
of these issues. This data has already been collected and constitutes a new avenue of
research.
Finally, future research will also benefit from considering the other side of the
coin: perceptual patterns. Evaluating the perception of the effects of nasal coupling
will extend the findings of this doctoral project in several ways. First, perception may
serve as a diagnostic tool to assess the dialectal differences in vowel nasality, which
would provide key evidence to substantiate the claims of extensive anticipatory vowel
nasalization in Caribbean dialects. The goal of perception studies is to examine if
and how native speakers of Spanish perceive the effects of nasalization. Nasalization
often leads to misperception of vowel height, which can be reverted under the right
169
conditions (i.e. when the co-articulatory source is present) for speakers of some
languages (e.g. English). If Caribbean listeners can compensate for vowel nasality,
it would provide perceptual evidence of the patterns reported in the sociolinguistics
literature of anticipatory vowel nasalization and uncovered in the present study.
In addition to examining differences in perception when the nasal consonant is
present or absent in the signal, perceptual protocols will benefit from incorporating
talkers from different dialectal regions. Lack of familiarity with a particular dialect
may lead to misapprehension of the speech signal. While it may not lead to sound
change per se, it does provide an opportunity to reveal loci for sound change. Thus,
research will be well served by incorporating the perception of native and nonnative
dialect stimuli in the design.
More broadly, extending the research reported here to include perception also
contributes more broadly to discussions of the relationship between production and
perception. Specifically, one key aspect that this dissertation did not explore relates to
individual variation. Some speakers are more attentive to phonetic cues than others,
which makes them ‘innovative’ listeners. What we still do not know, however, is how
‘innovative’ listeners of dialects at different stages of sound change behave similarly
or not. In other words, do innovative listeners of one dialect pattern with innovative
speakers of the other dialect, or do we observe gradience in listening innovation?
All in all, the combination of production and perception protocols will permit an
examination of broader scope of phonological variation and language change at the
individual level, as well as the group level, by incorporating comparisons of language
varieties, be them regional or social.
170
6.8
Concluding remarks
The present investigation has offered an analysis of the production of anticipatory
vowel nasalization and word-final nasal consonants in a Caribbean and a non-Caribbean
dialect of Spanish. The analyses examined the effects on the time-course of nasality
and word-final nasal consonant weakening, indicating that Santo Domingo Spanish
presents extensive anticipatory vowel nasalization, when compared to Buenos Aires
Spanish. Thus, the results confirm the reports of phonologization of anticipatory vowel
nasalization in Caribbean dialects of Spanish.
To summarize the general findings of this dissertation, the Caribbean dialect
produces earlier velum lowering than the non-Caribbean dialect. In terms of weakening
of the nasal consonant, dialect groups are not different. The analysis of co-variation
between the two variables revealed that, for both regional varieties, as the nasal
consonant weakens, onset of velum lowering takes place earlier. The analysis of the
linguistic variables also revealed group differences that contribute to the aforementioned
dialectal characteristics. Specifically, differences related to the stress condition, with
SDS enhancing earlier velum lowering, indicate anticipatory vowel nasality as an
intended property of the vowel for this dialect group, in comparison to BAS.
Taken together, the overall findings have important implications for our understanding of anticipatory vowel nasalization and its relationship to nasal consonant
weakening, as well as cross-dialectal differences. First and foremost, this investigation
contributes to the growing number of instrumental studies that diversify methodological approaches to the investigation of Spanish. The general description of the
time-course of nasalization together with nasal consonant weakening have implications
as to how we talk about the envelope of variation and the patterns of weakening.
Although the speech style elicited here is formal, the examination of co-variation
171
between the timing of velum lowering and the weakening of the nasal consonant
revealed that while the variables work in tandem, anticipatory vowel nasalization
need not be a compensatory mechanism. Instead of arguing for nasalization as a step
post-velarization or post-nasal consonant weakening, future descriptions of the system
will be well served by emphasizing the co-variation between the velic and the oral
gestures and comparing across dialect groups.
The acoustic analyses of vowel nasalization and nasal consonant weakening using
nasometry lays the foundation for future comparisons to other regional and social
varieties of Spanish using the same speech elicitation instruments employed here.
In-depth comparisons with other dialect groups, both within and across ‘velarizing’
and ‘alveolarizing’ dialects of Spanish, will provide further insight into variability and
divergence in related linguistic systems. Comparisons of social groups will permit
analyses of social uses of variation in anticipatory vowel nasalization and discussions
regarding the social dynamics of language change.
In conclusion, this doctoral project uncovered patterns of variation that had not
been part of previous conversations. The acoustic and statistical analyses presented
in this study have contributed to our overall understanding of how parallel linguistic
systems may diverge over time. Taken together, the findings depict patterns of
nasalization of a lesser-studied Romance language in this regard and emphasizes the
necessity to compare across regional varieties. All in all, this investigation highlights
the importance of acoustic analyses of nasality and demonstrates how detailed, acoustic
studies of speech production can offer insight into the system that is not achievable
by impressions alone.
172
Notes
1 The
borders of the city of Buenos Aires are, to the East, the River Plate, and to the
north, west and south, the highway “General Paz.”
2 In
this footnote, we take a brief detour to describe the physiology of the nasal cavity. The
nasal septum divides the nasal cavity into two passages, the right and left nares. Towards
the posterior end of the nasal cavity, the passages blend, via the choanae (or posterior
nares), into the nasopharynx. Towards the front, the nares emerge at the nostrils into the
open air. On the sides of the nasal cavity, there are three horizontal outgrowths called
nasal conchae (singular “concha”) or turbinates. The conchae are covered by many layers
of mucous membranes and blood vessels, which can swell when they become infected, like
in a cold, and leads to a “stuffed nose”. The mucosa are kept moisten by several glands.
One biological function of this complex pathway is to warm up, moist and filter the inhaled
air that we breathe, which prevents dry, cold, hot or dirty air from reaching the alveoli in
the lungs. The reader should be aware that even though the description above presents the
anatomy of the nasal cavity as homogeneous, research has shown large individual differences
in terms of volume and shape; but interestingly, not in terms of length. (Dang et al., 1994)
3 It
should be noted, however, that Bell-Berti and Hirose (1973) did find that one speaker
(subject BG) appeared to employ the palatoglossus muscle for the articulation of nasal
consonants, which lead them to conclude that “palatoglossus activity is primarily associated
wth tongue body movement, but may be implicated in the nasal manner of articulation in
some speakers” (p. 204).
4 As
a side note, the velopharyngeal port mechanism can be compared to a “trapdoor”
mechanism, in which the velum is lowered or raised like a gate. However, the velopharyngeal
port mechanism is a more complex process, that in fact, shows individual differences (Gick
et al., 2012). That is, even though speakers share the same set of muscles, they need not
operate them in the same way. The majority of speakers use the “trapdoor” method: the
levator palatini muscles are activated to raise the velum to form a complete seal against
173
the rear pharyangeal wall. Other mechanisms have been attested, namely the circular, the
circular with Passavant’s ridge and the sagittal methods. The main difference between these
alternative methods and the “trapdoor” method, aside from the number of speakers that
use them, is that the latter methods close the velophrayngeal port from multiple sides. For
this reason, they have been referred to as “drawstring” methods. Given that how speakers
close or open the velopharyngeal port is not relevant to the present dissertation, they will
not be discussed further here. For a review, see Gick et al. (2012, pp. 134-136). But they
are mentioned here because they show that the binary distinction between oral and nasal
can be implemented in different ways.
5A
nasograph is a device used to determine when the velopharyngeal port opens (or closes)
by detecting light coming up through the port (Ohala, 1971). A light source, embedded
in a thin flexible tube, is inserted through a nostril and positioned below the port. A
light-detector, also embedded in the tube, but positioned above the port, is used to pick
up the light as it passes through the open port. By determining the opening/closing of the
velopharyngeal port, the nasagraph can determine the extent to which a segment is nasalized
6A
force sensing bulb is a tear-drop shaped device that measures the force of the closure of
the velopharyngeal port in grams. The bulb is inserted transnasally (Kuehn & Moon, 1998).
Electromiography, on the other hand, is a technique that measures muscle movement and
provides information on the control of the velopharyngeal mechanism. Electrodes are placed
(in pairs) in the vicinity of the muscle under examination to detect and record time-varying
electrical activity (Krakow & Huffman, 1993)
7 Recasens
(1983) reports values of a single male speaker, as well mean values of a sample
of 12 male speakers. Values reported in in Table 2.1 represent the mean values of the 12
male Catalan speakers.
8 These
observations have been confirmed in perceptual studies. Beddor et al. (1986)
tested if native speakers of English were able to attribute nasalization to co-articulation
with an adjacent tautosyllabic nasal consonant. For this purpose, their perception of vowel
174
nasality was tested in three conditions: oral ([bEd-bæd]), contextual nasal ([bẼnd-bæ̃nd]),
and non-contextual nasal ([bẼd-bæ̃d]), with synthetically designed stimuli. Listeners were
asked to judge the height of the vowel, by indicating if it was a word with [E] (e.g., ‘bed’
or ‘bend’) or [æ] (e.g. ‘bad’ or ‘band’). Their hypothesis was that in the absence of the
co-articulatory source (i.e. the nasal consonant), listeners would judge [æ] as [E] (i.e. [æ]
would be perceptually raised). And indeed, this was their finding. In the non-contextual
nasal condition, they judged [æ] as [E] more often; that is, nasalization lowered perceived
vowel height. When the co-articulatory source was presented, their perception of vowel
height in the contextual nasal condition was not different than that of the oral condition,
which shows that listeners’ were able to factor out the effects of nasalization.
9 The
article reads “continental Spanish”. I am making the assumption that this means
Iberian Spanish. No other information is provided about the speakers.
10
Hyman (2013) distinguishes phonologization from Jakobson’s Phonologisierung (Jakob-
son, 1931[1972]), which is best translated as phonemicization and refers to an already
phonological property changing from allophonic to phonemic. As mentioned elsewhere, the
focus of this dissertation will be on the development of allophonic nasalization. However, in
order to simplify, I will use of the term phonologization to refer to both these processes.
11 Although
less common, in some cases phonologization of nasality has taken place in the
context of carryover nasalization. Interestingly, in these cases, stage two (deletion of the
nasal consonant) usually does not take place (Sampson, 1999, pp. 25–6).
12 The
role of place of articulation by itself in the development of nasalization (that is, not
in the context of mergers of nasals), has also been examined. Foley (1975, pp. 199–200),
and Lightner (1973, p. 33) claim that deletion of the nasal consonant spreads progressively
from [VN] to other environments. Outside of the Romance-speaking area, M. Chen (1975,
p. 33) and Hombert (1986, p. 87) posit the opposite path: anterior nasals, such as [Vm],
lead to effacement of the nasal consonant before posterior nasals, [VN]. And Hajek (1997,
pp. 164-168) finds in his Northern Italian dialects data that though distinctive nasalization
175
appears to develop earlier in the context of /n/, place of articulation itself does not seem to
account for the patterns observed (p. 167).
13 Onset
nasal phonemes also exhibit variation in Spanish, but have attracted the attention
of much less research. Depalatalization of word-final nasal consonants (desdén ‘disdain’
vs. desdeñes ‘to disdain’) has been addressed from a formal perspective as a historical
morpho-phonological process (Harris, 1999; Lloret & Mascaró, 2007; Pensado, 1997). Other
investigations have examined weakening of the alveolar nasal in intervocalic position (Honorof,
2003; Shosted & Willgohs, 2006). More recently, the realization of the palatal nasal has
also garnered the attention of research (Fernández Planas, 2009; Kochetov & Colantoni,
2011a; Shosted, Hualde, & Scarpace, 2012), especially in comparison to the nasal-plus-glide
sequence (Bongiovanni, 2015a, 2015b, in revision).
14 In
Yucatecan Spanish, in addition to velar realizations, some speakers have been shown
to produce a word-final /m/ as a result of contact with Mayan (Michnowicz, 2008).
15 In
this and the following examples from Spanish and Latin where Spanish and Latin
words are glossed as the same, I will include a single translation for the Latin source between
quotation marks. Additionally, the symbol >is used to indicate historical development.
16 Though
not the focus of this dissertation, the development of the onset nasal consonants
is interesting and will be briefly commented on in this footnote. The three-way contrast
in Spanish nasals only obtains in the syllable onset. Modern Spanish bilabial and alveolar
nasals originated in the Latin singleton and geminate onset nasals, /m mm n nn/. The
palatal nasal, on the other hand, exhibited a more intricate evolution with three reported
sources. First the bilabial consonant was assimilated which led to /nn/ and, subsequently,
to a palatal nasal (Penny, 2002, p. 74-75). For example, autumn > otoño ‘autumn’. One
source is the sequence alveolar nasal plus glide, /n+j/. This transformation started when
Latin atonic /e/ and /i/ in hiatus became glides, such that a word like vinea ‘vineyard’
“
“
became vinia. Next, by a process of regressive assimilation, the nasal preceding the glide
was palatalized, and became a palatal nasal. As a result, vinia became viña.
176
The second source of the palatal nasal was the Latin sequence -gn-. In the development
from Latin to Spanish, syllable-final velar stops palatalized. The velar stop first became
a glide, and next, through a process of progressive assimilation fused with the nasal: the
result was a palatal nasal. Thus, ligna ‘timber’ first became lejna, which in turn paved the
way to leña /leña/. As a result, the sequence -gn- merged with the sequence /n+j/ into the
phoneme /ñ/.
The alveolar geminate, nn, was the third source of the palatal nasal in modern Spanish.
Through a lenition process that took place around the tenth century, the Latin alveolar
geminate became a palatal nasal, such that a word like annu turned into Spanish año ‘year’
/año/. However, the details as to how a lenition process led to /ñ/ are not clear in the
literature.
17 The
description presented here only accounts for the development of tonic vowels.
Atonic vowels did not undergo diphthongization. In the interest of maintaining focus, I am
only considering the contexts where nasality exhibits (or by comparison to other Romance
languages, fails to exhibit) a role in the development of the vowel system.
18 Sampson
(1999, p. 168) considers another case where nasality may have operated in the
evolution of the vowel system: words with initial conf- and inf- were were realized with a
deleted nasal consonant in popular speech. In fact, in medieval texts (see examples in 5),
there is evidence of the alternation between retained and elided nasal consonants. It was at
a later stage that the coda /n/ was restored in forms that were considered to contain prefixal
con- and in-, and deletion of the nasal consonant was evaluated as incorrect. Sampson’s
reasoning is that if the nasal consonant was restored it has to be because nasality was still
present in the signal. However, the extent to which the ealier deletion of /n/ may have lead
to nasal vowels is unknown and nasality has not left a discernible trace. Additionally, how
these nasal(ized) vowels may have had an impact in the development of the Spanish vowel
inventory is not clear in Sampson (1999).
(5)
a. confŭndere >cofonder / cohonder ‘to destroy’
177
b. infăntem >i(f)fante ‘royal child’
19 Ramsammy
(2011) also examines word-medial nasal plus nasal cluster (#-NN-#/),
nasals in word-medial pre-obstruent position (/#-NC-#/), nasal plus nasal clusters across
word boundary (/-N#N-/). In the interest of brevity, I will not consider the findings for
these contexts.
20 French
does not have a contrast between /i/-/ĩ/, but does between /e/-/ẽ/. Thus, the
latter contrast allows Martínez (2015) to test naïve vs. non-naïve perception of nasality for
this group of listeners.
21 The
research on linguistic attitudes of Argentine speakers presents evidence that at first
blush may seem as contradicting. On one hand, Argentine speakers have been reported
to favor non-accommodation in situations of dialect contact, which research has explained
as a result of the linguistic prestige of Argentine Spanish (Barrancos, 2008). At the same
time, another strand of research examining beliefs regarding their own linguistic variety has
found that Argentines regard their dialect as “contaminated”, when compared to Peninsular
Spanish. However, as Gallardo (1978, p. 92-3) points out these types of negative viewpoints
are artifacts of a colonizing linguistic norm that continues to influence the ideology in Latin
America.
22 It
was decided not to examine the role of frequency at present time as it falls beyond the
scope of this dissertation. It is well-known that token and type frequency have an effect on
linguistic variation, especially with regard to lenition processes (Bybee, 2003). The effect of
frequency is an important issue that deserves a thorough examination. Such a study could
investigate the role of lexical and phonotactic frequencies, and more importantly, frequency
in a conditioning environment (E. L. Brown & Raymond, 2012; Raymond & Brown, 2012).
Thus, an in-depth examination of frequency effects on nasalization patterns in Spanish would
be very relevant and constitutes a reasonable expansion of the research proposed in this
dissertation (see chapter 6 for a discussion of future research).
23 Due
to technological errors selecting the microphone input when setting up recording
178
sessions, there are two participants that do not have three repetitions. Speaker DRF20 is
missing the first repetition and speaker DRF15 only has the third repetition (both in SDS
sample).
24 Present
day nasometers are an adaptation of the TONAR system. Fletcher (1970)
conceived of the idea of two separate microphones and named it TONAR, The Oral Nasal
Acoustic Ratio. TONAR made an important contribution to speech pathologists since
prior to its existence clinicians had to assess nasality in patients based on impressionistic
judgments. One of the main purposes of TONAR was to provide clinicians with data that
they could easily interpret. In a number of subsequent studies, Fletcher established that
the TONAR measurements correlated well with pressure-flow measurements and listener
perception (e.g. Fletcher, 1976). Currently available nasometers (e.g. Kay Elemetrics,
lingWAVES, the Glottal Enterprises Nasometer) operate using the same principles as the
Fletecher’s TONAR. However, these devices have incorporated technological advances that
allow to record, display and calculate nasality with computers.
25 At
this juncture, it is important to make a note about the energy measurement. When
we describe an acoustic signal, we refer to three dimensions: time, frequency and amplitude.
Energy measurements analyzed in this dissertation relate to this last dimension, amplitude.
When we describe amplitude, we describe sound pressure over time. We have a zero-line,
which represents rest position. A deviation from this zero line is called displacement, and
is an indication of the intensity or power of the sound (i.e. the bigger the displacement,
the louder the noise). The distance between the zero-line and the maximal displacement is
amplitude. Figure 6.6 illustrates the amplitude terms described above.
179
max
1.0
pressure
0.5
displacement
amplitude
zero line
0.0
−0.5
−1.0
min
0
5
10
15
time
Figure 6.6: Oscilogram of a sine wave illustrating amplitude terms
We can measure the sound pressure in two ways. The ‘official’ unit to measure amplitude
is pascal, which as stated above is a measure of air pressure fluctuation, or sound pressure.
However, what people think is the relationship in loudness between two sounds (e.g. one
sound is louder than another sound) does not match actual sound pressure differences. This
is because human audition is not on a linear scale, but on a logarithmic scale. For this
reason, the other way to measure amplitude is as decibel (dB), which represents a unit
of loudness and it is better correlated with perceived loudness. This dissertation uses Pa
measures, not dB.
26 The
data presented here shows a skewed distribution. Before running the analysis, ratio
values were converted to rationalized arcsine transform units (RAU). This transformation
“stretches” out the upper and lower ends of the scale thereby allowing for valid comparisons
of differences across the entire range of the scale (Studebaker, 1985). However, this transformation did not yield a normally distributed dataset (see figure below). For this reason, it
was decided to proceed with the analysis with the raw data.
180
Figure 6.7: Box, jitter plot and density plot for consonant-to-vowel oral energy ratio (raw
and transformed on the RAU scale), per dialect. Diamonds represent the mean.
27 In
this sense, this study would fall into a “narrow” sense of laboratory approaches to
sound variation, following the operationalization by Colantoni (2011b, p. 10)
28 In
fact, the background questionnaire asked participants about the purpose of the study
at the end, and no participant was able to discover it.
181
182
Appendix A
Background questionnaire
Por favor, complete este breve cuestionario.
Pseudónimo del participante (a ser completado por la investigadora):
Edad:
Sexo:
M
F
Ocupación:
Lugar de residencia (por ejemplo, barrio):
Educación:
primaria en progreso
primaria completada
secundaria en progreso
secundaria completada
universidad en progreso
universidad completada
estudios en:
posgrado en progreso
posgrado completado
estudios en:
¿Cuál es su lengua materna?
¿De dónde son sus padres (o la/s persona/s que lo criaron) y cuál es su primera lengua?
Madre
Ciudad de origen:
Primera lengua:
Padre
Ciudad de origen:
Primera lengua:
¿Cuál es el trasfondo educativo de sus padres (o la/s persona/s que lo criaron)?
Madre
primaria en progreso
primaria completada
secundaria en progreso
secundaria completada
universidad en progreso
universidad completada
estudios en:
posgrado en progreso
posgrado completado
estudios en:
Padre
primaria en progreso
primaria completada
secundaria en progreso
secundaria completada
universidad en progreso
universidad completada
estudios en:
posgrado en progreso
posgrado completado
estudios en:
¿Habla otra lengua en su hogar?
1
183
NO
SÍ, ¿qué lengua(s)?
a. ¿En qué porcentaje considera que usa su primera y su segunda lengua en su hogar?
% uso de la primera lengua
% uso de la segunda lengua
b. ¿Cuál es su nivel de proficiencia de la segunda lengua?
Proficiencia: principiante
intermedio avanzado casi nativo hablante nativo
¿Habla alguna otra lengua además de su lengua materna y la lengua que habla en su hogar?
Lengua 1:
Proficiencia en lectura:
principiante intermedio avanzado casi-nativo
Proficiencia en oralidad: principiante intermedio avanzado casi-nativo
Lengua 2:
Proficiencia en lectura:
principiante intermedio avanzado casi-nativo
Proficiencia en oralidad: principiante intermedio avanzado casi-nativo
Lengua 3:
Proficiencia en lectura:
principiante intermedio avanzado casi-nativo
Proficiencia en oralidad: principiante intermedio avanzado casi-nativo
Por favor, indica ciudades y países en los que haya residido, y su edad cuando residía en cada ubicación.
Ciudad (y país)
¿Qué edad tenía?
Ejemplo: Puebla, México
17-22 años
¿Ha visitado otro país hispanohablante? ¿Cuál y por cuánto tiempo?
Ciudad y país
Duración
2
184
¿Tiene o ha tenido amigos o parientes que hablen una variedad (regional) de español diferente de la
suya? ¿De dónde son y con qué frecuencia se comunica con ellos?
Persona
Ejemplo: primo
Ciudad y país
Frecuencia de contacto
Bilbao, España
todos los días a menudo no frecuentemente
todos los días a menudo no frecuentemente
todos los días a menudo no frecuentemente
todos los días a menudo no frecuentemente
todos los días a menudo no frecuentemente
todos los días a menudo no frecuentemente
¿Tiene amigos o parientes que vivan en otro país hispanohablante? ¿Qué país?
¿Cuáles son algunas de las diferencias entre la variedad de español que usted habla y la de otras
variedades (regionales) de español?
En su opinión, ¿de qué se trata el estudio?
Muchas de las frases que leyó son inusuales. ¿Había alguna palabra o frase que le pareciera por demás
3
185
extraña o inusual?
¿Algún comentario?
4
186
Please complete this brief questionnaire.
Participant pseudonym (to be completed by researcher):
Age:
M
Gender:
F
Profession:
Place of residence (for example, neighborhood):
Education:
elementary school in progress
elementary school completed
high school in progress
high school completed
college in progress
college completed
studies in:
graduate school in progress
graduate school completed
studies in:
What is your first language?
Where are your parents (or caretakers) from and what is their first language?
Mother
City of origin:
First language:
Father
City of origin:
First language:
What is the educational background of your parents (or caretakers)?
Mother
elementary school in progress
elementary school completed
high school in progress
high school completed
college in progress
college completed
studies in:
graduate school in progress
graduate school completed
studies in:
Father
elementary school in progress
elementary school completed
high school in progress
high school completed
college in progress
college completed
studies in:
graduate school in progress
studies in:
187
graduate school completed
Is another language spoken in your home?
NO
YES, what language(s)?
a. What percentage do you believe to use of your first and second languages in your home?
% use of first language
% use of second language
b. How fluent are you in the language?
Proficiency: beginner intermediate advanced native
Do you speak other languages other than your native language and the language spoken at home?
Language 1:
Reading proficiency: beginner
intermediate advanced near-native
Speaking proficiency: beginner
intermediate advanced near-native
Language 2:
Reading proficiency: beginner
intermediate advanced near-native
Speaking proficiency: beginner
intermediate advanced near-native
Language 3:
Reading proficiency: beginner
intermediate advanced near-native
Speaking proficiency: beginner
intermediate advanced near-native
Please indicate the cities and countries where you have lived, and your age while living in each location.
City (and country)
How old were you?
Example: Puebla, México
17-22 years old
Have you visited another Spanish-speaking country? Which, and for how long?
City and country
Duration
188
Do you have (or had) friends or relatives that speak a different dialect of Spanish than your own? Where
are they from and how often do you communicate with them?
Person
Example: cousin
City and country
Frequency of contact
everyday often not frequently
Bilbao, Spain
everyday often not frequently
everyday often not frequently
everyday often not frequently
everyday often not frequently
everyday often not frequently
Do you have friends or relatives that live in another Spanish-speaking country? Which country?
What are some the differences between your dialect of Spanish and other dialects of Spanish?
In your opinion, what was the study about?
Many of the phrases you read are unusual. Was there a word or phrase that was extremely strange?
189
Appendix B
Participant demographic information
Table A.1: Detailed demographic information of participants in the sample. Star indicates
excluded participant.
ID#
Gender
Age
Country
Other languages
DRF01
female
23
Dominican Republic
English
DRF02
female
21
Dominican Republic
English, French, Portuguese
DRF03
female
18
Dominican Republic
DRF04
female
19
Dominican Republic
English
DRF05*
female
19
Dominican Republic
English
DRF06∗
female
23
Dominican Republic
DRF07
female
18
Dominican Republic
English
DRF08
female
18
Dominican Republic
English, Portuguese
DRF09
female
20
Dominican Republic
DRF10
female
20
Dominican Republic
DRF11
female
20
Dominican Republic
DRF12
female
19
Dominican Republic
DRF13
female
19
Dominican Republic
DRF14∗
female
18
Dominican Republic
DRF15
female
23
Dominican Republic
DRF16
female
27
Dominican Republic
DRF17
female
18
Dominican Republic
DRF18
female
20
Dominican Republic
DRF19
female
21
Dominican Republic
DRF20
female
20
Dominican Republic
French
English, French, Mandarin
English
Continued on next page...
190
(continued from previous page)
ID#
Gender
Age
Country
Other languages
DRF21
female
20
Dominican Republic
English
DRM01
male
19
Dominican Republic
DRM02
male
18
Dominican Republic
DRM03
male
23
Dominican Republic
DRM04
male
21
Dominican Republic
English
DRM05∗
male
18
Dominican Republic
English
DRM06∗
male
18
Dominican Republic
French
DRM07
male
21
Dominican Republic
English, Japanese
DRM08
male
18
Dominican Republic
English, Portuguese
DRM09
male
18
Dominican Republic
English, Japanese
DRM10
male
21
Dominican Republic
German
BAF01
female
18
Argentina
English
BAF02
female
19
Argentina
English, Portuguese
BAF03
female
20
Argentina
English, French
BAF04
female
21
Argentina
BAF05
female
23
Argentina
English, Portuguese
BAF06∗
female
na
Argentina
na
BAF07
female
22
Argentina
English
BAF08
female
26
Argentina
English, Italian
BAF09
female
20
Argentina
French
BAF10
female
24
Argentina
English, French
BAF11∗
female
25
Argentina
English, Italian
Continued on next page...
191
(continued from previous page)
ID#
Gender
Age
Country
Other languages
BAF12
female
20
Argentina
English, Italian, Slovenian
BAF13
female
na
Argentina
na
BAF14
female
22
Argentina
Portuguese, English
BAF15
female
22
Argentina
English, Dutch, Italian
BAF16
female
25
Argentina
English
BAF17
female
27
Argentina
English
BAF18
female
28
Argentina
English, French, Hebrew
BAF19
female
24
Argentina
English, French
BAF20
female
21
Argentina
English
BAF21
female
na
Argentina
na
BAM01
male
20
Argentina
Guaraní
BAM02
male
24
Argentina
English, Welsh, Russian
BAM03
male
25
Argentina
English
BAM04
male
20
Argentina
English, Italian
BAM05
male
24
Argentina
English, Portuguese
BAM06
male
25
Argentina
English, French, German
BAM07
male
23
Argentina
English, French
BAM08
male
28
Argentina
English, Portuguese
192
193
Stimuli
CVN##
CVC##
Sequence
/o/
/a/
/e/
/o/
/a/
Y ahí esperan ‘And there they were’
‘And there they were’
Digo que tiraron
‘I say they threw’
Un joven capitán
‘A young captain’
Pasó de un tirón
‘It happened in a go’
Continued on next page...
‘May they visit’
Que nos visiten
La única sartén
‘The only skillet’
‘The mysterious cases’
‘The number twenty two’
‘The bag of potatoes’
‘The parent meeting’
Los misteriosos casos
La bolsa de papas
La reunión de los papás
El número veintidós
‘The colorful fish’
Los coloridos peces
Los grandes cafés
/e/
‘The big coffee houses’
Unstressed
Stressed
Vowel
Table C.1: Stimuli phrases. Target syllable is bolded. Stressed syllable is underlined. ## = Pre-pausal; #V = Pre-vocalic;
#[t] = Pre-coronal; #[k] = Pre-dorsal.
Appendix C
194
NV##
CV##
NVN##
Sequence
/e/
/o/
/a/
/e/
/o/
/a/
‘They say they cook’
Leen el canon
‘They read the canon’
‘A pretty magnet’
Un ácido limón
‘An acid lemon’
‘A majestic house’
Siempre lo peso
‘I always weight it’
El antiguo cine
‘S/he is always present’
No lo pesó
‘S/he didn’t weight it’
Aunque gané
Continued on next page...
‘The old movie theater’
Una majestuosa casa
Siempre está
‘I did win’
‘May s/he kiss him’
‘I didn’t kiss him’
Que lo bese
Dicen que cocinan
Un lindo imán
No lo besé
‘May the young people forgive’
Que los jóvenes perdonen
Pasó en un santiamén
/e/
‘It happened in a flash’
Unstressed
Stressed
Vowel
Table C.1: (continued from previous page)
195
CVN#V
CVC#V
Sequence
‘The mysterious fate’
Los peces azules
‘But s/he tidied up’
Los cafés azules
/a/
/e/
/o/
/a/
/e/
‘Blue houses’
Los casos astutos
‘The clever cases’
Cuando visiten Armenia
‘When visit Armenia’
Siempre esperan atentos
‘They always wait attentively’
‘The clever parents’
Los veintidós aliados
‘The twenty two allies’
La sartén aceitosa
‘The oily skillet’
Un capitán astuto
‘A clever captain’
Continued on next page...
Las casas azules
Los papás astutos
‘The blue fish’
El misterioso destino
Pero si ordenó
‘The blue coffee houses’
‘The great dinner’
‘I said (to him): win’
/o/
La gran cena
le dije ganá
/a/
Unstressed
Stressed
Vowel
Table C.1: (continued from previous page)
196
CV#V
NVN#V
Sequence
/o/
/a/
/e/
/o/
La casa alegre
‘The joyous house’
Siempre peso aúcar
‘I always weight sugar’
Está alegre
‘S/he is happy’
Pesó albóndigas
‘He weighted almonds’
Continued on next page...
‘May s/he kiss him on the alert’
‘I weighted sugar’
Que lo bese alerta
‘But they win happily’
‘The happy magnet’
Pesé azúcar
Pero ganan alegres
El imán alegre
/a/
‘The joyous canon’
‘Have them tidy up attentively’
‘A clever flash’
‘The blue lemon’
Que ordenen atentos
Un santiamén astuto
El canon alegre
‘When they throwed sugar
‘An attentive pull’
El limón azul
Cuando tiraron azúcar ’
Un tirón atento
/o/
/e/
Unstressed
Stressed
Vowel
Table C.1: (continued from previous page)
197
CVN#[t]
CVC#[t]
NV#V
Sequence
/e/
/o/
/a/
/e/
/o/
/a/
‘The clever dinner’
El destino agradable
‘The nice fate’
‘May you win allies’
Aunque ganó atento
‘Though s/he won on the alert’
‘The warm potatos’
Los casos terribles
‘The terrible cases’
Que visiten temprano
‘The stingy parents’
Los veintidós tomates
‘Twenty two tomatoes’
La sartén torcida
Continued on next page...
‘May they visit early’
Las papas templadas
Los papás tacños
‘The crooked skillet’
‘The crooked fish’
‘Two warm coffee houses’
Los peces torcidos
La cena alegre
Ganá alerta
Dos cafés templados
‘The clever cinema’
El cine astuto
Siempre frené alerta
/e/
‘I always stopped on the alert’
Unstressed
Stressed
Vowel
Table C.1: (continued from previous page)
198
CV#[t]
NVN#[t]
Sequence
Cuando tiraron tabaco
‘When they threw tobacco’
Que cocinen tomates
Un tirón terrible
‘A terrible pull’
Un santiamén tacaño
/a/
/e/
/o/
/a/
/e/
‘When they win in 3rd place’
El canon tacaño
‘The stingy canon’
Que lo bese temprano
‘May s/he kiss him early’
La casa tallada
‘The carved house’
‘The crooked magnet’
Un limón templado
‘A warm lemon’
Pesé tabaco
‘I weighted tobacco’
Está terrible
‘It is terrible’
Continued on next page...
Cuando ganan terceros
The imán torcido
‘Have them cook tomatoes’
‘They wait drinking’
‘A fearful heartthrob’
‘A cheap flash’
Esperan tomando
Un capitán tacaño
/a/
/o/
Unstressed
Stressed
Vowel
Table C.1: (continued from previous page)
199
CVC#[k]
NV#[t]
Sequence
/o/
/a/
/e/
/o/
Las papas calientes
‘The hot potatos’
Los casos cargados
‘The loaded cases’
Los papás casados
‘The married parents’
Las veintidós cadenas
‘Twenty two chains’
Continued on next page...
‘The fleshy fish’
‘Three hot coffees’
Los peces carnosos
‘The tender dinner’
‘(You) win talent’
Ocho cafés calientes
La cena tierna
Ganá talento
/a/
‘The cheap fate’
‘The dark theater’
‘But I guessed everything’
‘When s/he won in third place’
El cine tenebroso
Pero adiviné todo
El destino tacaño
‘I weight tomatoes’
‘it crossed crooked’
Cuando ganó tercero
Peso tomates
Pasó torcido
/o/
/e/
Unstressed
Stressed
Vowel
Table C.1: (continued from previous page)
200
CV#[k]
NVN#[k]
CVN#[k]
Sequence
/e/
/o/
/a/
/e/
/o/
/a/
‘They were waiting in silence’
Cuando tiraron cadenas
‘When they threw chains’
‘A married captain’
Un tirón cargado
‘A loaded pull’
‘They always win hitting’
El canon cautivo
‘The captive canon’
Que le pese callado
‘A fleshy magnet’
el limón caliente
‘The hot lemon’
No pesé cadenas
Continued on next page...
‘May he weigth (it) in silence’
Siempre ganan cantando
The imán carnoso
‘I did weight chains’
‘Have them cook singing’
‘A hot flash’
Que cocinen cantando
Esperan callados
Un capitán casado
Un santiamén caliente
‘May they visit singing’
Que visiten cantando
La sartén caliente
/e/
‘The hot skillet’
Unstressed
Stressed
Vowel
Table C.1: (continued from previous page)
201
NV#[k]
Sequence
Peso catorce manzanas
‘I weight fourteen apples’
El cine canalla
Si pesó cajitas grises
‘If s/he weighted grey little boxes’
Pero adiviné contento
/o/
/a/
/e/
La cena caliente
‘The hot dinner’
El destino canalla
‘The rotten fate’
Ganá cajitas
‘(You) win little boxes’
Si ganó callado
‘But he won in silence’
‘The rotten theatre’
‘The captive house’
‘It is not hot’
‘But I guessed happily’
La casa cautiva
No está caliente
/a/
/o/
Unstressed
Stressed
Vowel
Table C.1: (continued from previous page)
Appendix D
Data Management
I. Prepare the raw data.
(a) Annotate the .wav file with a TextGrid that consists of one only tier. Keep
these files in the participant folder.
(b) In this one tier, target phrases are marked with an interval. These intervals
are left blank, as they will be automatically labeled with label_from_text
_file.praat (see step 3) and extracted into their own .wav file with
save_labeled_intervals_to_wav_files.praat (see step 4). Intervals that are not going to be extracted into their own .wav file are labeled
as “xxx". This way, when label_from_text_file.praat is ran, they
are ignored and left as is.
• Remember that the first and last five phrases do not contain target
tokens. They should be labeled “xxx".
II. Prepare the labels.
(a) For each stimuli file, save the .ppt file to .rtf in PowerPoint.
(b) Open the .rtf file in LibreOffice. Do not open files in Word. Save the
.rtf files as .txt files. When prompted for file format, choose “Text - Choose
Encoding (.txt)” (see Figure D.1 below). This option will allow file to be
encoded as UTF-8 and end lines with LF-only. When prompted, follow
Figure D.2 below.
202
Figure D.1: Save text file
Figure D.2: ASCII Filter Options
(c) Close the .txt file.
(d) Open the .csv file with the coding for the labels in LibreOffice. Do not
open files in Excel. When you open the file, make sure that only the tab
option is checked. When prompted, follow the figure below.
203
Figure D.3: Importing text file into Open Office spreadsheet
(e) Open the .txt file LibreOffice (the one you closed in (c)).
(f) For the first line in each text file (remember it is going to be a different line for each participant and each repetition), copy the text from
labelc oding3.cvs and paste it to the text file.
(g) Open the Mac terminal to run silvina_diss3.py. For this purpose,
run the following command, specifying each argument:
python script.py originalfile infofile newfile
For my dissertation, script.py= silvina_diss3.py; originalfile=
label_coding3.csv; infofile= the file of the participant (e.g. BAF01B.txt);
newfile= the output file (e.g. BAF01B_full.txt)
• The path of each file (script, file with labels, participant file and output
file) can be specified. Otherwise, everything has to be located in the
204
working directory. In order to specify paths, do as follows (with the
appropriate path line; this one is just an example):
python /home/Desktop/script /home/Desktop/originalfile
/home/Desktop/
infofile /home/Desktop/newfile
• Save label all output files in the same folder. This will facilitate
the next step. The path is as follows: /Users/MyFanyMacbook/
Documents/Indiana/Dissertation/
Data Management/labels
III. Label the raw data.
(a) Run label_from_text_file.praat. Use the output file to label the
TextGrid (e.g. BAF01B_full.txt). All output files are available at:
...
/Documents/Indiana/Dissertation/Data Management/labels
(b) Open the sound and textgrid in Praat and check that the file has been
labeled appropiately at multiple locations.
(c) Make necessary adjustments and run the scripts as many times as needed.
IV. Extract smaller .wav files.
(a) Open the sound file as LongSound file and its corresponding TextGrid.
Select them in the Objects window.
(b) Run labelf romt extf ile.praat and save files in the following folder:
/Users/MyFanyMacbook/Documents/Indiana/Dissertation/Data
For each repetition, there is a folder: “A", “B" and “C", respectively. Save
the files in the corresponding folder and check that the filename matches
205
the speaker’s production. This step is important. Once you are certain
there are no mistakes, move the files to the general “Data" folder.
(c) Make sure that “Exclude intervals labeled as xxx" is checked.
(d) IMPORTANT: Add speaker, sex, dialect and repetition information as
a suffix. That is, if the sound file is “BAF01A", then add as suffix
“_BAF01_F_BA_A".
• The name of the file includes the whole coding for each token: speaker,
sex, dialect group, repetition, target word (with no diacritics), target
vowel, target sequence, stress, and following segment.
V. Create Textgrid.
(a) Run text_grid_maker_plus.praat.
(b) This script will open a TextGrid with four interval tiers that are already
labeled (target, prepost, exclusion and comments).
• Tier 1 (interval): target consonant
• Tier 2 (interval): left boundary = onset of the vowel preceding the
target consonat; right boundary = offset of the following segment.
• Tier 3 (interval) exclusion criteria
• Tier 4 (interval) code (e.g. for exclusion criteria)
• Tier 5 (interval): comments
VI. Annotate.
(a) Annotate sound files following the criteria in chapter 4.
206
References
Alba, O. (2000). Nuevos aspectos del español en Santo Domingo. Santo Domingo:
Librería La Trinitaria.
Alba, O. (2004). Cómo hablamos los dominicanos: un enfoque sociolingüístico. Santo
Domingo: Grupo León Jimenes.
Alba, O. (2009a). Hipercorrección en los programas de noticias en la televisión de la
República Dominicana.
Alba, O. (2009b). La identidad lingüística de los dominicanos. Santo Domingo:
Librería La Trinitaria.
Alba, O. (2016). Una mirada panorámica al español antillano. Santo Domingo:
Editora Nacional.
Albalá, M. J. (1992). Análisis y síntesis de las consonantes nasales. Revista de filología
española, 72 (1), 37.
Al-Bamerni, A. (1983). Oral, velic and laryngeal coarticulation across languages
(Unpublished doctoral dissertation). University of Oxford.
Alfaraz, G. G. (2015). Variation of overt and null subject pronouns in the Spanish of Santo Domingo. In A. M. Carvalho, R. Orozco, & N. L. Shin (Eds.),
Subject pronoun expression in Spanish: A cross-dialectal perspective (pp. 3–16).
Washington, DC: Georgetown University Press.
Baković, E. (2000). Nasal place neutralization in Spanish. University of Pennsylvania
Working Papers in Linguistics, 7 (1), 2.
207
Barrancos, A. (2008). Linguistic accommodation by Argentinean immigrants in Spain:
The case of the pronoun vos and other features. Birkbeck Studies in Applied
Linguistics (BISAL), 3 , 27–51.
Beddor, P. (1993). The perception of nasal vowels. In M. K. Huffman & R. A. Krakow
(Eds.), Nasals, nasalization, and the velum (pp. 171–196). Academic Press.
Beddor, P. (2009). A coarticulatory path to sound change. Language, 85 (4), 785–821.
Beddor, P., & Gottfried, T. L. (1995). Methodological issues in cross-language speech
perception research with adults. In W. Strange (Ed.), Speech perception and
linguistic experience: Issues in cross-language research (pp. 207–232). Timonium,
MD [United States]: York Press.
Beddor, P., & Krakow, R. (1999). Perception of coarticulatory nasalization by speakers
of English and Thai: Evidence for partial compensation. The Journal of the
Acoustical Society of America, 106 (5), 2868–2887.
Beddor, P., Krakow, R., & Goldstein, L. (1986). Perceptual constraints and phonological change: a study of nasal vowel height. Phonology, 3 (01), 197–217.
Bell-Berti, F. (1976). An electromyographic study of velopharyngeal function in
speech. Journal of Speech, Language, and Hearing Research, 19 (2), 225–240.
Bell-Berti, F. (1993). Understanding velic motor control: studies of segmental context.
In M. K. Huffman & R. A. Krakow (Eds.), Nasals, nasalization, and the velum
(pp. 63–85). Academic Press.
Bell-Berti, F., & Hirose, H. (1973). Patterns of palatoglossus activity and their
implications for speech organization. The Journal of the Acoustical Society of
America, 54 (1), 320–320.
Bergen, J. J. (1976). The explored and unexplored facets of questions such as "¿Qué
tú tines?". Hispania, 93–99.
Boersma, P., & Weenink, D.
(2016).
208
Praat: doing phonetics by computer
(version 6.0.16).
[Computer program].
(Retrieved 5 April 2016 from
http://www.praat.org/)
Bongiovanni, S. (2015a). Are/n+ j/and/ñ/merged in Buenos Aires Spanish? An
initial acoustic analysis. In E. W. Willis, P. M. Butragueño, & E. H. Zendejas
(Eds.), Selected proceedings of the 6th conference on laboratory approaches to
romance phonology (pp. 17–29). Somerville, MA [United States]: Cascadilla
Proceedings.
Bongiovanni, S. (2015b). Neutralización del contraste entre/ñ/y/nj/en el español de
Buenos Aires: Un estudio de percepción. Signo y Seña| Revista del Instituto de
Lingüística(27), 11–46.
Bongiovanni, S. (in revision). An acoustical analysis of the merger of / ñ/ and / nj /
in Buenos Aires Spanish. Journal of the International Phonetic Association.
Borzone de Manrique, A. M. (1980). Manual de fonética acústica. Buenos Aires:
Hachette.
Braver, A. (2012). Textgrid maker plus. (Retrieved 29 September 2016 from
http://www.aaronbraver.com/resources/)
Brown, C. A. (1998). The role of the L1 grammar in the L2 acquisition of segmental
structure. Second Language Research, 14 (2), 136–193.
Brown, E. K. (2009). A usage-based account of syllable-and word-final /s/ reduction
in four dialects of Spanish (Unpublished doctoral dissertation). University of
New Mexico.
Brown, E. L., & Raymond, W. D. (2012). How discourse context shapes the lexicon:
Explaining the distribution of Spanish f-/h words. Diachronica, 29 (2), 139–161.
Büdenbender, E.-M. S. (2010). Comparing dominican linguistic (in)security in
the Dominican Republic and in the diaspora. In C. Borgonovo, M. E. nol
Echevarría, & P. Prévost (Eds.), 12th hispanic linguistics symposium (pp. 148–
209
159). Somerville, MA: Cascadilla Proceedings Project.
Bullock, B. E., & Toribio, A. J. (2014). Dominican Spanish. In M. Di Paolo &
A. K. Spears (Eds.), Languages and dialects in the us: Focus on diversity and
linguistics (pp. 151–162). Routledge.
Bullock, B. E., Toribio, J., & Amengual, M. (2014). The status of s in Dominican
Spanish. Lingua, 143 , 20–35.
Bybee, J. (2000). Lexicalization of sound change and alternating environments. In
M. B. Broe & J. B. Pierrehumbert (Eds.), Papers in laboratory phonology v:
Acquisition and the lexicon (pp. pp. 250–68). Cambridge: Cambridge University
Press.
Bybee, J. (2003). Phonology and language use (Vol. 94). Cambridge University Press.
Campos-Astorkiza, R. (2012). The phonemes of Spanish. In J. I. Hualde, A. Olarrea, &
E. O’Rourke (Eds.), The handbook of Hispanic linguistics (p. 89-110). Chichester,
West Sussex [England]: John Wiley & Sons.
Campos-Astorkiza, R. (forthcoming). Laboratory approaches to Spanish phonology.
In F. Martínez-Gil & S. Colina (Eds.), The Routledge Handbook of Spanish
Phonology. Taylor & Francis.
Canfield, D. L. (1981). Spanish pronunciation in the Americas. Chicago [United
States]: University of Chicago Press.
Cárdenas, D. N. (1960). Introducción a una comparación fonológica del español y del
inglés. Washington, DC: Center for Applied Linguistics of the Modern Language
Association of America.
Carricaburo, N. (2013). El voseo argentino: un fenómeno con abordajes múltiples. In
L. Colantoni & C. R. Louro (Eds.), Perspectivas teóricas y experimentales sobre
el español de la Argentina (pp. 133–142). Madrid/Frankfurt: Iberoamericana
Vervuert.
210
Cedergren, H., & Sankoff, D. (1975). Nasals: A sociolinguistic study of change
in progress. In C. Ferguson, L. Hyman, & J. Ohala (Eds.), Papers from a
symposium on nasals and nasalization (pp. 67–80). Stanford, CA: Stanford
University.
Chang, C. (2008). Variation in palatal production in Buenos Aires Spanish. In
M. Westmoreland & J. A. Thomas (Eds.), Selected proceedings of the 4th workshop
on spanish sociolinguistics. Somerville, MA: Cascadilla Proceedings Project.
Chen, M. (1973). Cross-dialectal comparison: a case study and some theoretical
considerations. Journal of Chinese Linguistics, 38–63.
Chen, M. (1975). An areal study of nasalization in Chinese. Journal of Chinese
Linguistics, 16–59.
Chen, M. Y. (1995). Acoustic parameters of nasalized vowels in hearing-impaired
and normal-hearing speakers. The Journal of the Acoustical Society of America,
98 (5), 2443–2453.
Chen, M. Y. (1997). Acoustic correlates of English and French nasalized vowels. The
Journal of the Acoustical Society of America, 102 (4), 2360–2370.
Clark, J., Yallop, C., & Fletcher, J. (2007). An introduction to phonetics and phonology.
Oxford: Blackwell.
Cline, W. J. (1972). The phonology of sixteenth century American Spanish. Ann
Arbor, MI: University Microfilms.
Clumeck, H. (1976). Patterns of soft palate movements in six languages. Journal of
Phonetics, 4 (4), 337–351.
Cohn, A. C. (1993). Nasalisation in English: phonology or phonetics. Phonology,
10 (01), 43–81.
Colantoni, L. (2011a). Broad-focus declaratives in Argentine Spanish contact and
non-contact varieties. Intonational phrasing at the interfaces: cross-linguistic
211
and bilingual studies in Romance and Germanic. Amsterdam: John Benjamins,
183–212.
Colantoni, L. (2011b). Laboratory approaches to sound variation and change. In
M. Díaz-Campos (Ed.), The handbook of hispanic sociolinguistics. John Wiley
& Sons.
Colantoni, L., & Gurlekian, J. (2004). Convergence and intonation: historical
evidence from Buenos Aires spanish. Bilingualism: Language and cognition,
7 (2), 107–119.
Colantoni, L., & Kochetov, A. (2012). Nasal variability and speech style: An
EPG study of word-final nasals in two Spanish dialects. Italian Journal of
Linguistics/Rivista di Linguistica, 24 , 11–42.
Colantoni, L., & Marinescu, I. (2010). The scope of stop weakening in Argentine
Spanish. In M. Ortega-Llebaria (Ed.), Selected proceedings of the 4th conference
on laboratory approaches to spanish phonology (pp. 100–114). Somerville, MA:
Cascadilla Proceedings Project.
Cole, J., Hualde, J. I., & Iskarous, K. (1999). Effects of prosodic and segmental
context on/g/-lenition in Spanish. In O. Fujimura, B. Joseph, & B. Palek (Eds.),
Proceedings of the fourth international linguistics and phonetics conference
(Vol. 2, pp. 575–589).
Colina, S. (2009). Spanish phonology: A syllabic perspective. Georgetown University
Press.
Conde, O. (2011). Lunfardo: un estudio sobre el habla popular de los argentinos.
Aguilar.
Corp., I. (2013). IBM SPSS statistics for windows, version 22.0. [Computer program].
Armonk, NY: IBM Corp.
Dalcher, C. V. (2008). Consonant weakening in Florentine Italian: A cross-disciplinary
212
approach to gradient and variable sound change. Language variation and change,
20 (2), 275–316.
Dang, J., Honda, K., & Suzuki, H. (1994). Morphological and acoustical analysis of
the nasal and the paranasal cavities. The Journal of the Acoustical Society of
America, 96 (4), 2088–2100.
Delvaux, V., Demolin, D., Harmegnies, B., & Soquet, A. (2008). The aerodynamics
of nasalization in French. Journal of Phonetics, 36 (4), 578–606.
Diakoumakou, E. (2004). Coarticulatory vowel nasalization in Modern Greek. (Unpublished doctoral dissertation). University of Michigan.
D’Introno, F., del Teso Martín, E., & Weston, R. (1995). Fonética y fonología actual
del español. Madrid: Catedra.
D’Introno, F., & Sosa, J. M. (1988). Elisió de nasal o nasalizaciõ de vocal eN caraqueño.
In R. Hammond & M. Resnick (Eds.), Studies in Caribbean Spanish dialectology
(pp. 24–34). Washington, DC [United States]: Georgetown University Press.
Dow, M. (2014). Contrast and markedness among nasal(ized) vowels: A phoneticphonological study of French and Vimeu Picard (Unpublished doctoral dissertation). Indiana University.
Eddington, D. (2011). What are the contextual phonetic variants of /B D G/ in
colloquial Spanish? Probus, 23 (1), 1–19.
Fails, W. (2011). O grau de nasalização das vogais oronasais no português paulistano
e no espanhol mexicano: Um estudo experimental comparativo. Hispania, 94 (3),
443–461.
Fant, G. (1960). Acoustic theory of speech production: with calculations based on x-ray
studies of Russian articulations (Vol. 2). Gravenhage: Mouton.
Fernández Planas, A. M. (2009). Características linguopalatales de la nasal palatalizada
en español. Estudios de fonética experimental, 18 , 162–174.
213
Fletcher, S. G. (1970). Theory and instrumentation for quantitative measurement of
nasality. The Cleft Palate Journal, 7 , 601–609.
Fletcher, S. G. (1976). Nasalance vs. listener judgements of nasality. Cleft Palate J ,
13 , 31–44.
Foley, J. (1975). Nasalization as universal phonological process. In C. A. Ferguson,
L. M. Hyman, & J. J. Ohala (Eds.), Násalfest: Papers from a symposium on
nasals and nasalization (pp. 197–212). Language Universals Project, Stanford
University.
Fontanella de Weinberg, M. B. (1966). Comparación de dos entonaciones regionales
argentinas. Thesaurus, 1 (1), 17–29.
Fontanella de Weinberg, M. B. (1968). El voseo en Buenos Aires, un problema
histórico lingüístico. Cuadernos del Sur(8-9), 174–179.
Fontanella de Weinberg, M. B. (1973). El rehilamiento bonaerense a fines del siglo
XVIII. Thesaurus. Boletín del Instituto Caro y Cuervo, 18 , 1–8.
Fontanella de Weinberg, M. B. (1978). Un cambio lingüístico en marcha: las palatales
del español bonaerense. Orbis, 27 , 215–247.
Fontanella de Weinberg, M. B. (1979). La oposición cantes/cantés en el español de
Buenos Aires. Thesaurus, 34 , 72–83.
Fontanella de Weinberg, M. B. (1980). Three intonational systems of Argentinean
Spanish. The melody of language, 115–126.
Fritzell, B. (1969). The velopharyngeal muscles in speech. an electromyographiccineradiographic study. Acta otolaryngologica, Suppl–250.
Fujimura, O. (1962). Analysis of nasal consonants. The Journal of the Acoustical
Society of America, 34 (12), 1865–1875.
Gallardo, A. (1978). Hacia una teoría del idioma estándar. Revista de lingüistica
teórica y aplicada, 16 , 85–119.
214
García, M., & Rodríguez, M. (1997). Estudio acústico de las consonantes nasales del
español. Estudios de fonética experimental, 9 , 37–64.
Gick, B., Wilson, I., & Derrick, D. (2012). Articulatory phonetics. John Wiley &
Sons.
Gobello, J. (1989). El lunfardo. Académia Porteña del lunfardo.
Gobello, J. (1996). Aproximación al lunfardo. Ediciones de la Universidad Católica
Argentina,.
Goodin-Mayeda, C. E. (2011). Perceptual compensation for acoustic effects of nasal
coupling by Spanish and Portuguese listeners. In S. M. Alvolrd (Ed.), Selected
proceedings of the 5th conference on laboratory approaches to romance phonology
(pp. 75–83). Somerville, MA: Cascadilla Proceedings.
Goodin-Mayeda, C. E. (2016). Nasals and nasalization in Spanish and Portuguese:
Perception, phonetics and phonology (Vol. 9). John Benjamins Publishing
Company.
Haché de Yunén, A. (1981). La/n/ final de sílaba en el español de Santiago de los
Caballeros. In O. Alba (Ed.), El español del caribe (pp. 143–154). Santiago de
los Caballeros: Universidad Católica Madre y Maestra.
Hajek, J. (1997). Universals of sound change in nasalization. Oxford [England]:
Wiley.
Harris, J. W. (1984a). Autosegmental phonology, lexical phonology, and Spanish
nasals. In M. Aronoff, R. Oehrie, & B. W. Stevens (Eds.), Language, sound,
structure: Studies in phonology presented to Morris Halle by his teacher and
students (pp. 67–82). Cambridge, MA: The MIT Press.
Harris, J. W. (1984b). Theories of phonological representation and nasal consonants
in Spanish. In P. Baldi (Ed.), Papers from the XIIth Linguistic Symposium on
Romance Languages (pp. 153–168). Amsterdam, Netherlands.
215
Harris, J. W. (1999). Nasal depalatalization no, morphological wellformedness sí;
the structure of Spanish word classes. MIT Working Papers in Linguistics, 33 ,
47–82.
Harris, J. W., & Kaisse, E. M. (1999). Palatal vowels, glides and obstruents in
Argentinian Spanish. Phonology, 16 (2), 117–190.
Heap, D. (1990). Questions with preverbal subject in Dominican Spanish: Morphological variation and syntactic restructuring. In annual meeting of the atlantic
provinces linguistic association (Vol. 14, pp. 59–67).
Henderson, J. B. (1984). Velopharyngeal function in oral and nasal vowels: A crosslanguage study (Unpublished doctoral dissertation). University of Connecticut.
Henríquez Ureña, P. (1940). El español en Santo Domingo. Buenos Aires: Biblioteca
de Dialectología Hispanoamericana.
Herrera, E. (2002). La asimilación de las nasales en español: un estudio instrumental.
Nueva revista de filología hispánica, 50 (1), 1–14.
Hombert, J.-M. (1986). The development of nasalized vowels in the Teke language
group (Bantu). In K. Bogers & H. van der Hulst (Eds.), The phonological
representation of suprasegmentals (pp. 359–379). Dordrecht: Foris.
Hombert, J.-M. (1987). Phonetic conditioning for the development of nasalization in
Teke. Pholia, 2 , 85–94.
Honorof, D. N. (1999). Articulatory gestures and Spanish nasal assimilation (Unpublished doctoral dissertation). Yale University.
Honorof, D. N. (2003). Articulatory evidence for nasal de-occlusivization in Castilian.
In M.-J. Solé, D. Recasens, & J. Romero (Eds.), Proceedings of the 15th international congress of phonetic sciences (pp. 1759–1763). Universitat Autònoma de
Barcelona.
Hooper, J. B. (1981). The empirical determination of phonological representations.
216
Advances in Psychology, 7 , 347–357.
House, A. S. (1957). Analog studies of nasal consonants. The Journal of Speech and
Hearing Disorders, 22 (2), 190–204.
House, A. S., & Stevens, K. N. (1956). Analog studies of the nasalization of vowels.
Journal of Speech and Hearing Disorders, 21 (2), 218–232.
Howell, D. C. (2012). Statistical methods for psychology. Cengage Learning.
Hualde, J. I. (1989a). Delinking processes in Romance. In C. Kirschner & J. DeCesaris
(Eds.), Studies in Romance linguistics (pp. 177–193). Amsterdam, Netherlands.
Hualde, J. I. (1989b). Procesos consonánticos y estructuras geométricas en español.
Lingüística(1), 7–44.
Hualde, J. I. (2005). The sounds of Spanish. Cambridge [England]: Cambridge
University Press.
Hualde, J. I. (2014). Los sonidos del español. Cambridge [England]: Cambridge
University Press.
Hualde, J. I., Simonet, M., & Nadeu, M. (2011). Consonant lenition and phonological
recategorization. Laboratory Phonology, 2 (2), 301–329.
Huffman, M. K. (1990). The role of f1 amplitude in producing nasal percepts. The
Journal of the Acoustical Society of America, 88 (S1), S54–S54.
Huffman, M. K., & Krakow, R. A. (Eds.). (1993). Nasals, nasalization, and the velum
(Vol. 5). San Diego: Academic Press.
Hyman, L. M. (2013). Enlarging the scope of phonologization. In A. C. L. Yu (Ed.),
Origins of sound change: Approaches to phonologization (pp. 3–28). Oxford
University Press.
Jakobson, R. (1931[1972]). Principles of historical phonology. In A. R. Keiler (Ed.), A
reader in historical and comparative linguistics (pp. 121–138). New York [United
States]: Holt, Rinehart & Winston.
217
Jiménez Sabater, M. A. (1975). Más datos sobre el español de la república dominicana.
Santo Domingo: Ediciones Intec.
Johnson, K. (2012). Acoustic and auditory phonetics. Wiley-Blackwell.
Johnson, M. (2016). Pragmatic variation in voseo and tuteo negative commands in
Argentinian Spanish. In M. I. Moyna & S. Rivera-Mills (Eds.), Forms of address
in the Spanish of the Americas (Vol. 10, p. 127). John Benjamins Publishing
Company.
Kaisse, E. (2001). The long fall: An intonational melody of Argentinian Spanish.
In J. Herschensohn, E. Malle, & K. Zagona (Eds.), Features and interfaces in
romance (pp. 37–57). Amsterdam: John Benjamins.
Kingston, J. (2008). Lenition. In L. Colantoni & J. Steele (Eds.), Selected proceedings
of the 3rd conference on laboratory approaches to spanish phonology (pp. 1–31).
Somerville, MA.
Kochetov, A., & Colantoni, L. (2011a). Coronal place contrasts in Argentine and
Cuban Spanish: An electropalatographic study. Journal of the International
Phonetic Association, 41 (03), 313–342.
Kochetov, A., & Colantoni, L. (2011b). Spanish nasal assimilation revisited: A
cross-dialect electropalatographic study. Laboratory Phonology, 2 (2), 487–523.
Krakow, R. A. (1989). The articulatory organization of syllables: A kinematic analysis
of labial and velar gestures (Unpublished doctoral dissertation). Yale University.
Krakow, R. A. (1993). Nonsegmental influences on velum movement patterns: Syllables,
sentences, stress, and speaking rate. In M. K. Huffman & R. A. Krakow (Eds.),
Nasals, nasalization, and the velum (pp. 87–116). Academic Press.
Krakow, R. A. (1999). Physiological organization of syllables: a review. Journal of
Phonetics, 27 (1), 23–54.
Krakow, R. A., & Huffman, M. K. (1993). Understanding velic motor control:
218
studies of segmental context. In M. K. Huffman & R. A. Krakow (Eds.), Nasals,
nasalization, and the velum (pp. 3–59). Academic Press.
Kuehn, D. P., & Moon, J. B. (1998). Velopharyngeal closure force and levator
veli palatini activation levels in varying phonetic contexts. Journal of Speech,
Language, and Hearing Research, 41 (1), 51–62.
Ladefoged, P. (2005). A course in phonetics. Wadsworth Publishing.
Lavoie, L. (2001). Consonant strength: Phonological patterns and phonetic manifestations. New York, NY: Routledge.
Lederer, J. S. (2003). The diachronic coronal-velar nasal relationship. In M.-J. Solé,
D. Recasens, & J. Romero (Eds.), 15th international congress of phonetic sciences
(p. 2801-2804). Universitat Autònoma de Barcelona.
Lennes, M. (2002a). Label intervals in an intervaltier from text file. (Retrieved 25
April 2015 from https://lennes.github.io/spect/)
Lennes, M. (2002b). Save intervals to small wav sound files. (Retrieved 29 September
2016 from https://lennes.github.io/spect/)
Liberman, A. M., Delattre, P. C., Cooper, F. S., & Gerstman, L. J. (1954). The role of
consonant-vowel transitions in the perception of the stop and nasal consonants.
Psychological Monographs: General and Applied, 68 (8), 1.
Lightner, T. M. (1973). Remarks on universals in phonology. In M. Gross (Ed.),
The formal analysis of natural languages. proceedings of the first international
conference (pp. 13–50). The Hague: Mouton.
Lindblom, B. (1990). Explaining phonetic variation: A sketch of the H&H theory. In
W. Hardcastle & A. Marchal (Eds.), Speech production and speech modelling
(pp. 403–439). Dordrecht: Kluwer Academic Publishers.
Lipski, J. M. (1994). Latin American Spanish. London [England]; New York [United
States]: Longman.
219
Lipski, J. M. (2011). Socio-phonological variation in Latin American Spanish. In
M. Díaz-Campos (Ed.), The handbook of Hispanic Sociolinguistics (pp. 72–97).
John Wiley & Sons.
Lloret, M.-R., & Mascaró, J. (2007). Depalatalization in Spanish revised. In F. Martinez
& S. Colina (Eds.), Optimality-theoretic studies in Spanish phonology (pp. 74–98).
Amsterdam/Philadelphia: John Benjamins.
López Morales, H. (1980). Velarización de “n” en el español de Puerto Rico. LEA:
Lingüística española actual, 2 (2), 203–217.
Lorenzino, G. A. (2016). El lunfardo en la evolución del español argentino. Literatura
y lingüística(34), 335–356.
Lubker, J. F., Fritzell, B., & Lindqvist, J. (1970). Velopharyngeal function: An
electromyographic study. Quarterly Progress Status Report-Speech Transmission
Lab (Stockholm), 4 , 9–20.
Maeda, S. (1993). Acoustics of vowel nasalization and articulatory shifts in French
nasal vowels. In M. K. Huffman & R. A. Krakow (Eds.), Nasals, nasalization,
and the velum (pp. 147–170). San Diego: Academic Press.
Malécot, A. (1956). Acoustic cues for nasal consonants: An experimental study
involving a tape-splicing technique. Language, 274–284.
Malmberg, B. (1965). Estudios de fonética hispánica (Vol. 1). Consejo Superior de
Investigaciones Científicas, Institut Miguel de Cervantes.
Martinet, A. (1952). Function, structure, and sound change. Word, 8 (1), 1–32.
Martínez, R. (2015). La perception naïve non native des voyelles nasales du portugais
(Unpublished master’s thesis). Université de Montréal.
Martínez-Celdrán, E. (2008). Some chimeras of traditional Spanish phonetics. In
L. Colantoni & J. Steele (Eds.), Selected proceedings of the 3rd conference on
laboratory approaches to Spanish phonology (pp. 32–46). Somerville, MA [United
220
States]: Cascadilla Proceedings.
Martínez-Sanz, C. (2011). Null and overt subjects in a variable system: The case of
Dominican Spanish (Unpublished doctoral dissertation). University of Ottawa
(Canada).
Menéndez Pidal, R. (1962). Manual de gramática histórica española (11th ed.).
Madrid: Espasa-Calpe.
Michnowicz, J. (2008). Final nasal variation in Merida, Yucatan. Spanish in Context,
5 (2), 278–303.
Narayan, C. R. (2008). The acoustic–perceptual salience of nasal place contrasts.
Journal of Phonetics, 36 (1), 191–217.
Navarro Tomás, T. (1980[1918]). Manual de pronunciación española. [Consejo Superior
de Investigaciones Científicas].
Núñez Cedeño, R. (1980). La fonología moderna y el español de Santo Domingo.
Santo Domingo: Editorial Taller.
Oficina Nacional de Estadística. (2015). Expansión urbana de las ciudades capitales
de rd: 1988-2010. Santo Domingo. (Retrieved 25 January 2016)
Ohala, J. J. (1971). Monitoring soft palate movements in speech. The Journal of the
Acoustical Society of America, 50 (1A), 140–140.
Ohala, J. J. (1981). The listener as a source of sound change. In C. S. Masek,
R. A. Hendrick, & . M. F. Miller (Eds.), Papers from the parasession on language
and behavior (pp. 178–203). Chicago [United States]: Chicago Linguistic Society.
Ohala, J. J. (1996). Speech perception is hearing sounds, not tongues. The Journal
of the Acoustical Society of America, 99 (3), 1718–1725.
Ohala, J. J. (2012). The listener as a source of sound change: An update. In M. J. Solé
& D. Recasens (Eds.), The initiation of sound change. perception, production and
social factors (pp. 21–36). Amsterdam [The Netherlands]/Philadelphia [United
221
States]: John Benjamins.
Ohala, J. J., & Ohala, M. (1993). The phonectics of nasal phonology: theorems and
data. In M. K. Huffman & R. A. Krakow (Eds.), Nasals, nasalization, and the
velum (pp. 225–249). Academic Press.
Otheguy, R., Zentella, A. C., & Livert, D. (2007). Language and dialect contact in
Spanish in New York: Toward the formation of a speech community. Language,
770–802.
Penny, R. J. (2002). A history of the Spanish language. Cambridge University Press.
Pensado, C. (1997). On the Spanish depalatalization of/ñ/ and /L/in rhymes. Fernando
Martınez-Gil and Alfonso Morales-Front (eds.), Issues in the phonology and
morphology of the major Iberian languages, 595–618.
Piñeros, C.-E. (2006). The phonology of nasal consonants in five Spanish dialects.
In F. Martínez-Gil & S. Colina (Eds.), Optimality-theoretic studies in Spanish
phonology (pp. 146–171).
Piñeros, C.-E. (2014). El inventario fonémico nasal del español: un estudio comparativo.
Revista Española de Lingüística, 41 (1), 73–111.
Pruthi, T., & Espy-Wilson, C. Y. (2007). Acoustic parameters for the automatic
detection of vowel nasalization. In Proceedings of interspeech 2007. 8th annual
conference of the international speech communication association (pp. 1925–
1928).
Quilis, A. (1965). Description phonétique du parler madriléne actuel. Phonetica,
12 (1), 19–24.
Quilis, A., & Fernández, J. A. (1973). Curso de fonética y fonología españolas: para estudiantes angloamericanos. Madrid [Spain]: Consejo Superior de Investigaciones
Científicas.
Ramsammy, M. (2011). The realisation of coda nasals in Spanish (Unpublished
222
doctoral dissertation). University of Manchester.
Ramsammy, M. (2013). Word-final nasal velarisation in Spanish. Journal of Linguistics,
49 (01), 215–255.
Raphael, L. J., Borden, G. J., & Harris, K. S. (2011). Speech science primer:
Physiology, acoustics, and perception of speech (6th edition ed.). Baltimore, MD:
Lippincott Williams & Wilkins.
Raymond, W. D., & Brown, E. L. (2012). Are effects of word frequency effects of
context of use? an analysis of initial fricative reduction in Spanish. In S. T. Gries
& D. Divjak (Eds.), Frequency effects in language (pp. 35–52). Berlin [Germany]:
De Gruyter Mouton.
Real Academia Española. (2011). Nueva gramática de la lengua española: Fonética y
fonología. Barcelona: Espasa Libros.
Recasens, D. (1983). Place cues for nasal consonants with special reference to Catalan.
The Journal of the Acoustical Society of America, 73 (4), 1346–1353.
Resnick, M. C. (1975). Phonological variants and dialect identification in Latin
American Spanish (Vol. 201). Berlin [Germany]: de Gruyter Mouton.
Robe, S. L. (1960). The Spanish of rural Panama: major dialectal features. Berkeley,
CA: University of California Press.
Rochet, A. P., & Rochet, B. L. (1991). The effect of vowel height on patterns
of assimilation nasality in French and English. In Proceedings of the xiith
international congress of phonetic sciences (Vol. 3, pp. 54–57).
Rohena-Madrazo, M. (2013). Variación y cambio de sonoridad de la fricativa postalveolar del español de Buenos Aires. In L. Colantoni & C. R. Louro (Eds.),
Perspectivas teóricas y experimentales sobre el español de la Argentina (pp.
37–57). Madrid/Frankfurt: Iberoamericana Vervuert.
Rohena-Madrazo, M. (2015). Diagnosing the completion of a sound change: Phonetic
223
and phonological evidence for /S/ in Buenos Aires Spanish. Language Variation
and Change, 27 (3), 287–317.
Ruhlen, M. (1973). Nasal vowels. Working Papers on Language Universals, 12 , 1–36.
Salvador, G. (1957). El habla de Cúllar-Baza. contribución al estudio de la frontera
del andaluz. Revista de filología española.
Sampson, R. (1999). Nasal vowel evolution in Romance. Oxford [England]; New York
[United States]: Oxford University Press.
Scarborough, R. (2013). Neighborhood-conditioned patterns in phonetic detail:
Relating coarticulation and hyperarticulation. Journal of Phonetics, 41 (6),
491–508.
Scarborough, R., Styler, W., & Zellou, G. (2011). Nasal coarticulation in lexical
perception: The role of neighborhood-conditioned variation. In W.-S. Lee &
E. Zee (Eds.), Proceedings of the 17th international congress of the phonetic
sciences (icphs xvii) (pp. 1750–1753). Hong Kong: University of Hong Kong.
Schourup, L. (1972). Characteristics of vowel nasalization. Research on Language &
Social Interaction, 5 (4), 530–548.
Schourup, L. (1973). A cross-language study of vowel nasalization. Ohio State
University Working Papers in Linguistics, 15 , 190–221.
Shosted, R., Hualde, J. I., & Scarpace, D. (2012). Palatal complexity revisited: An
electropalatographic analysis of /ñ/ in Brazilian Portuguese with comparison to
Peninsular Spanish. Language and Speech, 55 (4), 477–502.
Shosted, R., & Willgohs, B. (2006). Nasals unplugged: The aerodynamics of nasal
de-occlusivization in Spanish. In M. Díaz-Campos (Ed.), Selected proceedings of
the 2nd conference on laboratory approaches to Spanish phonetics and phonology
(pp. 14–21). Somerville, MA [United States]: Cascadilla Proceedings.
Siracusa, M. I. (1972). Morfología verbal del voseo en el habla culta de Buenos Aires.
224
Filología, 16 , 201–213.
Solé, M.-J. (1992). Phonetic and phonological processes: The case of nasalization.
Language and Speech, 35 (1-2), 29–43.
Solé, M.-J. (1994). New ways of analyzing sound change: Speech rate effects. Belgian
Journal of Linguistics, 9 (1), 21–44.
Solé, M.-J. (1995). Spatio-temporal patterns of velopharyngeal action in phonetic and
phonological nasalization. Language and Speech, 38 (1), 1–23.
Strange, W. (1995). Speech perception and linguistic experience: Issues in crosslanguage research. In W. Strange (Ed.), Speech perception and linguistic experience: Issues in cross-language research (pp. 3–45). Timonium, MD: York
Press.
Studebaker, G. A. (1985). A rationalized arcsine transform. Journal of Speech,
Language, and Hearing Research, 28 (3), 455–462.
Terrell, T. (1975). La nasal implosiva y final en el español de Cuba. Anuario de
Letras, 13 .
Terrell, T. (1978). La aspiración y elisión de /s/ en el español porteño. Anuario de
Letras. Lingüística y Filología, 16 , 41–66.
Terrell, T. (1982). Relexificación en el español dominicano: Implicaciones para la
educación. In O. Alba (Ed.), El español del caribe. ponencias del vi simposio de
dialectología (pp. 301–318). Santiago: Universidad Católica Madre y Maestra.
Terrell, T. (1986). La desaparición de /s/ posnuclear a nivel léxico en el habla
dominicana. In R. N. nez Cedeño, I. P. Urdaneta, & J. M. Guitart (Eds.),
Estudios sobre la fonología del español del Caribe (pp. 117–134). Caracas
[Venezuela]: La Casa de Bello.
Thomas, E. (2010). Sociophonetics: An introduction. Palgrave Macmillan.
Toribio, A. J. (2000). Nosotros somos dominicanos: Language and self-definition
225
among Dominicans. In A. Roca (Ed.), Research on Spanish in the United States:
Linguistic issues and challenges (pp. 252–270). Somerville, MA: Cascadilla
Press.
UNESCO. (2016). Colonial city of Santo Domingo UNESCO World Heritage Centre.
(add url, Accessed: Nov. 29 2016)
United Nations. (2015). World population prospects: The 2015 revision, key findings and advance tables. United Nations Publications. (Working Paper No.
ESA/P/WP.241)
Vaissiére, J. (1988). Prediction of velum movement from phonological specifications.
Phonetica, 45 (2-4), 122–139.
Vaquero, M. (1996). Antillas. In M. Alvar (Ed.), Manual de dialectología hispanica:
El español de América (pp. 51–67). Barcelona [Spain]: Ariel.
Vidal de Battini, B. E. (1964). El español de la Argentina. Buenos Aires: Ministerio
de Educación.
Wolf, C., & Jiménez, E. (1979). El ensordecimiento del yeísmo porteño, un cambio
fonológico en marcha. Estudios lingüísticos y dialectológicos: Temas hispánicos,
115–135.
Zamora, V. (1985[1965]). Dialectología española. Madrid: Gredos.
Zellou, G., & Tamminga, M. (2014). Nasal coarticulation changes over time in
Philadelphia English. Journal of Phonetics, 47 , 18–35.
Zsiga, E. C. (2013). The sounds of language: An introduction to phonetics and
phonology. Chichester, West Sussex, U.K: Wiley-Blackwell.
226
Silvina Bongiovanni
Indiana University
Department of Linguistics
Department of Spanish and Portuguese
Email: scbongio@indiana.edu
Education
2018, dual Ph.D. in Linguistics and in Hispanic Linguistics, Indiana University.
Thesis: “Production of anticipatory vowel nasalization and word-final nasals
in two dialects of Spanish”
Committee: Kenneth de Jong (co-chair), Erik Willis (co-chair), Julie Auger,
Manuel Díaz Campos.
2012, M.A. in Hispanic Linguistics, Indiana University.
2008, Licenciatura in Letras, Universidad de Buenos Aires.
Teaching Appointments
2010–Present, Associate Instructor, Department of Spanish and Portuguese,
Indiana University.
2009–2010, Lecturer in Spanish, Council of International Exchange Education
(CIEE)-Facultad Latinoamericana de Ciencias Sociales (FLACSO) Buenos
Aires.
2008–2010, Lecturer in Spanish, Laboratorio de Idiomas, Facultad de Filosofia y
Letras, Universidad de Buenos Aires.
2008, Associate Instructor in Linguistics, Carrera de Letras, Facultad de Filosofia
y Letras, Universidad de Buenos Aires.
2006–2007, Spanish Language Scholar, Reed College.
Publications
Peer-reviewed articles
Long, A., Solon, M., Bongiovanni, S. (2018). Context of learning and second
language development of Spanish vowels. Studies in Hispanic and Lusophone
Linguistics, 11(1), 59–87.
Bongiovanni, S., Long, A., Solon, M., & Willis, E. (2015). The effect of shortterm study abroad on second language Spanish phonetic development. Studies
in Hispanic and Lusophone Linguistics, 8(2), 243–283.
Bongiovanni, S. (2015a). Neutralización del contraste entre /ñ/ y /nj/ en el
español de Buenos Aires: Un estudio de percepción. Signo y Seña. Revista
del Instituto de Lingüística, 27, 11–46.
Bongiovanni, S. (2015b). Are /n+j/ and /ñ/ merged in Buenos Aires Spanish?
An initial acoustic analysis. In E. Willis, P. M. Butrageño & E. Herrera
Zendejas (Eds.), Selected Proceedings of the 6th Conference on Laboratory
Approaches to Romance Phonology (pp. 17–29). Somerville, MA: Cascadilla
Proceedings.
Peer-reviewed working papers research
Bongiovanni, S. (2016). An exploratory study of nasal decomposition and
mid-vowel gliding in Argentine Spanish. IULC Working Papers, 16.
Bongiovanni, S. (2014a). “No se preocupe la señora marquesa”: A study of
gender bias in example sentences in the RAE grammar textbook. IULC
Working Papers, 14(1).
Bongiovanni, S. (2014b) “¿Tomas [pepsi], [peksi] o [pesi]?”: A variationist
sociolinguistic analysis of Spanish syllable coda stops. IULC Working Papers,
14(2).
Works submitted
Bongiovanni, S. (in revision). Acoustical analysis of the merger of /ñ/ y /nj/
in Buenos Aires Spanish. Journal of the International Phonetic Association.
Works in progress
Bongiovanni, Silvina Variation and frequency effects: Assymetrical velarization
of Spanish word-internal syllable coda-stops.
Bongiovanni, Silvina and Avizia Long. Production and perception of stress
and vowel reduction in L2 Spanish.
Bongiovanni, Silvina and Samson Lotven. Nasality and voicing contrasts in
Spanish.
Avizia Long and Bongiovanni, Silvina. Pronunciation instruction, task complexity and production of L2 Spanish stress and vowel reduction.
Published pedagogical materials
Bongiovanni, S., Krusul, L., & Lacanna, G. (2010a). Taller de Léxico. Buenos
Aires: CIEE-Flacso.
Bongiovanni, S., Krusul, L., & Lacanna, G. (2010b). Taller de Consolidación
Gramatical. Buenos Aires: CIEE-Flacso.
Bongiovanni, S., Coucido, D., Scutiero, V., & Suárez, V. (2010c). Taller de
Desempeño Oral. Buenos Aires: CIEE-Flacso.
Leibovich, E., Marisimian, S., Fiszman, L., Bongiovanni, S., Ayala, C., Moyano,
A., Fantin, L., Muñoz Lascano, P., Margolis, F., & Cucci, M. (2010d). Lengua
6. Prácticas del Lenguaje. Buenos Aires: Tinta Fresca.
Leibovich, E., Marisimian, S., Fiszman, L., Bongiovanni, S., Ayala, C., Moyano,
A., Fantin, L., Muñoz Lascano, P., Margolis, F., & Cucci, M. (2010e). Lengua
5. Prácticas del Lenguaje. Buenos Aires: Tinta Fresca.
Leibovich, E., Marisimian, S., Fiszman, L., Bongiovanni, S., Ayala, C., Moyano,
A., Fantin, L., Muñoz Lascano, P., Margolis, F., & Cucci, M. (2010f). Lengua
4. Prácticas del Lenguaje. Buenos Aires: Tinta Fresca.
Published reviews
Bongiovanni, S. (2017). Review of Goodin-Mayeda, E. Nasals and nasalization
in Spanish and Portuguese. Perception, phonetics and phonology. Linguist
List.
Bongiovani, S. (2009). Review of I. Kuguel, A. Adelstein, & G. Resnik. 1300
neologismos en la prensa argentina. Débate Terminológico, 5.
Presentations
Refereed conference presentations
Bongiovanni, S. (2018, February). “Angeles y diablitos”: Pronunciation of
voiced stops and approximants.” 10th anniversary of Current Approaches to
Spanish and Portuguese Second Language Phonology. Indiana University,
Bloomington, IN.
Bongiovanni, S. & Weirich, P. (2017, September). “The PIN/PEN merger and
the time-course of nasality.” 22nd Mid-Continental Phonetics and Phonology
Conference. The Ohio State University, Colombus, OH.
Bongiovanni, S. (2016, January). “Variation and frequency effects: Asymmetry
in the velarization of Spanish word-internal coda stops.” Talk given at the
90th Annual Meeting of the Linguistic Society of America. Washington DC.
Bongiovanni, S. (2015, September). “Neutralization of /ñ/ and /n+j/ in Buenos
Aires Spanish: A production study.” Hispanic Linguistics Symposium. University of Illinois at Urbana-Champaign. Urbana-Champaign, IL.
Bongiovanni, S. (2015, September). “An Acoustic Characterization of the /ñ//n+j/ Contrast in Buenos Aires Spanish.” 20th Mid-Continental Phonetics
and Phonology Conference. Indiana University, Bloomington, IN.
Bongiovanni, S. (2014, November). “Perception of incomplete neutralization of
/ñ/ and /n+j/ in Buenos Aires Spanish.” 2014 Hispanic Linguistic Symposium.
Purdue University, West Lafayette, IN.
Bongiovanni, S. (2014, April). “ ‘No se preocupe la Sra. Marquesa’: A study
of gender bias in example sentences in the RAE grammar textbook.” Ohio
State University Congress on Hispanic and Lusophone Linguistics. Ohio State
University, Columbus, OH.
Bongiovanni, S. (2014, April). “Word-internal syllable-coda stops in the Spanish
of Mérida, Venezuela: A sociolinguistic analysis.” Poster presented at the
7th Workshop in Spanish Sociolinguistics. University of Wiscosin-Madison.
Madison, WI.
Long, A., Solon, M. & Bongiovanni, S. (2014, March). “Context of learning and
second-language development of Spanish vowels.” 4th Current Approaches to
Spanish and Portuguese Second Language Phonology, Georgetown University.
Bongiovanni, S., Long, A., & Solon, M. (2013, October). “The effect of shortterm study abroad on the acquisition of Spanish phonology.” 2013 Hispanic
Linguistic Symposium. University of Ottawa, Ottawa, Canada.
Bongiovanni, S. (2013, October). “ ‘¿Tomas [pepsi], [peksi] o [pesi]?’: A sociolinguistic analysis of Spanish syllable coda stops.” 41st New Ways of Analyzing
Variation. Indiana University. Bloomington, IN.
Bongiovanni, S. (October, 2013). “A comparison of nasal contrast in Buenos
Aires.” 6th Laboratory Approaches to Romance Phonology. Colegio de
México. Ciudad de México D.F, México.
Invited talks
“Dialectal differences as a lens into ohonological vairaiton and language change.”
Michigan State University. February 26th 2018.
“Variación linguística y fonología.” Pontificia Universidad Católica Madre y
Maestra – Campus Santo Tomás de Aquino (Santo Domingo, República
Dominicana). July 12th 2016.
“An introduction to Yerba Mate.” University of Delaware. Guest Talk in Contemporary Latin America I. January 5th 2014.
“An introduction to ‘Argentine-ness.’ ” University of Delaware. Guest Talk in
Contemporary Latin America I. January 7th 2014.
Departmental talks
“Variation and frequency effects: Asymmetry in the velarization of Spanish wordinternal syllable coda stops.” Department of Linguistics Alumni Weekend,
October 9th, 2015.
“Individual difference factors and L2 Spanish phonetic development at home and
abroad” (with Avizia Long and Megan Solon). Department of Spanish and
Portuguese Diálogos conference, February 12th 2014.
“Nasal Contrast in Buenos Aires Spanish: A Perception Study.” November 2013.
Department of Spanish and Portuguese Spring Brown Bag series, November
2013.
“The effect of short-term study abroad on the acquisition of Spanish phonology.”
Department of Spanish and Portuguese Spring Brown Bag series (with Avizia
Y. Long and Megan Solon), October 2013.
“A preliminary analysis of /ñ/ and /n+j/ merger in Buenos Aires Spanish.”
Department of Spanish and Portuguese Spring Brown Bag series, February
2013.
“A comparison of nasal contrast in Buenos Aires.” Department of Spanish and
Portuguese Spring Brown Bag series, October 2012.
Research Experience
Fieldwork
Collected recordings in Santo Domingo, Dominican Republic, June-July 2016.
Collected recordings in Buenos Aires, Argentina, June-July 2016.
Collected recordings and perception data in Buenos Aires, Argentina, May-August
2013.
Grant-funded research support
2016–2017, Dissertation Completion Fellowship, College of Arts and Sciences,
Indiana University, $25,000.
2016, Grant-In-Aid, Graduate School, Indiana University, $1,000.
2016, Householder Research Grant, Department of Linguistics, Indiana University,
$400.
2013, Tinker Field Research Grant, Center for Latin American and Caribbean
Studies, Indiana University, $1,200.
Research assistance
Prof. Kimberly Geeslin, Summer 2017, assistance with editing and indexing for
The Handbook in Spanish Linguistics.
Prof. Laura Gurzynski-Weiss, Summer 2017, assistance with Answer Key for
Introducción y aplicaciones contextualizadas a la lingüística hispánica.
Prof. Laura Gurzynski-Weiss, Fall 2014, generated research on various topics at
researcher’s request.
Prof. Kimberly Geeslin and Prof. Avizia Yim Long, Spring 2013, assistance with
editing for Sociolinguistics and SLA.
Prof. Guiomar Ciapuscio and Prof. Andreína Adelstein (Universidad de Buenos
Aires), 2007-2009, assistance with annotation of corpus COTECA (Corpus
Textual del español científico de la Argentina.)
Grants and Awards
Research-Related
2015, Travel Award, College of Arts and Sciences, Indiana University, $400.
2015, Travel Award, Department of Spanish and Portuguese, Indiana University,
$200.
2014, Honorable Mention Fred W. Householder Research Paper Award, Department of Linguistics, Indiana University.
2014, Graduate Student Research Award, Department of Spanish and Portuguese,
Indiana University.
2012, Outstanding Student Poster. Laboratory Approaches to Romance Phonology (LARP) 6 conference, Colegio de México.
2012, Travel Award, Department of Spanish and Portuguese, Indiana University,
$400.
2011, Summer Graduate Fellowship, Department of Spanish and Portuguese,
Indiana University, $3,500.
Teaching-Related
2016, Recognition of Outstanding Associate Instructor with Distinction, Department of Spanish and Portuguese, Indiana University.
2015, Recognition of Outstanding Associate Instructor with High Distinction,
Department of Spanish and Portuguese, Indiana University.
Teaching Experience
(instructor of record indicated with a star symbol)
Indiana University
Courses taught
Introduction to Spanish Linguistics∗ , Fall 2013, Spring 2014, Summer 2015, Fall
2015.
Conversation and Diction∗ , Spring 2016, Spring 2017.
Conversation and Composition∗ , Fall 2016.
Spanish Grammar in Context∗ , Spring 2013.
Intermediate Spanish II∗ , Fall 2012, Summer 2014, Spring 2015.
Intermediate Spanish I∗ , Spring 2011, Fall 2011, Spring 2012.
Accelerated First Year Spanish∗ , Fall 2010.
Training workshops
Praat Scripting Workshop, Spring 2017.
IU Honors Program in Foreign Languages
Communication∗ , Ciudad Real (Spain), Summer 2011, Summer 2012.
CIEE-Flacso Buenos Aires
Lexical and Phonetics Workshop∗ , Spring 2009, Fall 2009, Spring 2010.
Oral Production and Comprehension Workshop∗ , Fall 2009, Spring 2010.
Grammar Workshop∗ , Spring 2009, Fall 2010.
Advanced Oral Production and Comprehension Workshop∗ , Spring 2011, Fall
2011, Spring 2012.
Universidad de Buenos Aires, Carrera de Letras
Gramática∗ , Fall 2007.
Universidad de Buenos Aires, Laboratorio de Idiomas
Beginner Spanish.∗
Intermediate Spanish.∗
Advanced Spanish.∗
Reed College
Conversation, Fall 2006, Spring 2007.
Service
Editorial
Reviewing activity, Article reviewer for Studies in Hispanic and Lusophone Linguistics, 2018.
Co-Editor, Proceedings of the 48th Anual Conference in African Linguistics,
Indiana University, 2017–2018.
Reviewing activity, Abstract reviewer for Current Approaches to Spanish and
Portuguese Second Language Phonology (CASPSLaP) 2018, 9th International
Workshop on Spanish Sociolinguistics, 2017.
Reviewing activity, Chapter reviewer for Cambridge University Press, edited
volume, 2017.
Co-Editor, IULC Working Papers in Linguistics, 2015–2016.
Associate Editor, IULC Working Papers in Linguistics, 2014–2015.
Conference Organization
Member, Organizing Committee, The 10th anniversary of Current Approaches
to Spanish and Portuguese Second Language Phonology, Indiana University,
2018.
Member, Organizing Committee, Diálogos 11 Graduate Student Conference,
Indiana University, 2014.
Member, Organizing Committee, Diálogos 10 Graduate Student Conference,
Indiana University, 2013.
Conference Volunteer
MidPhon 20, Indiana University (September 2015)
Diálogos 12 Graduate Student Conference, Indiana University, (February
2015)
Pragmatics and Language Learning Conference, Bloomington, IN (April
2014)
New Ways of Analyzing Variation 41, Bloomington, IN (October 2012)
Departmental
Co-Founder and Co-Chair, Hispanic and Lusophone Linguistics Society, 2013–
2015.
Member, Graduate Student Association Committee, 2012–2014.
Assembly Representative, Graduate and Professional Student Organization, Indiana University, 2012–2013.
Community
Volunteer, La Escuelita, Monroe County Library, Bloomington IN, 2016.
Miscellaneous
Computing skills
Praat, R, SPSS, LATEX(comfortable)
Languages
Spanish, Native speaker
English, Near native
Italian, Intermediate
French, Beginner
Документ
Категория
Без категории
Просмотров
1
Размер файла
6 808 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа