Velar Palatalization: Catalan, Spanish and Bilingualism
Item Type text; Electronic Dissertation
Authors Ramírez Martínez, Marta
Publisher The University of Arizona.
Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
Download date 01/10/2021 16:46:12
Link to Item http://hdl.handle.net/10150/624564
VELAR PALATALIZATION:
CATALAN, SPANISH AND BILINGUALISM
by
Marta Ramírez Martínez
______Copyright © Marta Ramírez Martínez 2017
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF SPANISH AND PORTUGUESE
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
WITH A MAJOR IN SPANISH
In the Graduate College
THE UNIVERSITY OF ARIZONA
2017 2 THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE
As members of the Dissertation Committee, we certify that we have read the dissertation prepared by Marta Ramírez Martínez entitled Velar Palatalization: Catalan, Spanish and Bilingualism and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Philosophy.
______Miguel Simonet Date: 6 December 2016
______Ana M. Carvalho Date: 6 December 2016
______Sonia Colina Date: 6 December 2016
______Andrew Wedel Date: 6 December 2016
Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.
______Dissertation Director: Miguel Simonet Date: 6 December 2016
3 STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgement of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the copyright holder.
SIGNED: Marta Ramírez Martínez
4
ACKNOWLEDGEMENTS
Primer de tot, gràcies als meus participants, que van haver de sofrir la interminable lectura de frases en calorosos mesos d’estiu. Sense la seva generositat no hi hauria tesi doctoral. També els hi he d’agraïr infinitament a totes aquelles persones que de manera desinteressada em van facilitar el procès a Palma, a Artà i a Manacor. Gràcies, gràcies, gràcies.
Gràcies infinites al meu professor, director de tesi i mentor, el professor Miquel Simonet, per totes les hores que m’ha dedicat (i paciència que m’ha tingut) durant els darrers 7 anys. Aquest projecte mai no hauria estat així sense la seva meticulositat a l’hora d’escriure i de fer recerca. Gràcies, Miquel.
I would also like to thank my other committee members, Professor Sonia Colina, Professor Andy Wedel and Professor Ana Carvalho, for the suggestions and comments you provided to make this work a better version of it. Especialmente, gracias, Ana, por tu ayuda y apoyo durante estos años, así como por las oportunidades que me has brindado en diferentes proyectos y en el departamento.
Gracias también a mis compañeros desde el principio del viaje, Ryan, Kern, Casillas (y su conocimiento de estadística), Laura y Tasha, y a mis compañeras más recientes, Andrea y Miriam, por hacer el largo camino que es grad school mucho más llevadero con vuestro apoyo y amistad. Gracias a Ericka, Kariana, Karyme y Carlitos por darme una familia tan lejos de la mía. Gracias también a mis amigas de siempre, por seguir aguantándome a través de los años y de la distancia: Tam, Ori, Marga, Amelia, Isa, Vicky, Noemi, Carol, Elena, Rebe, Rocío, Mar y Kryss.
Finalment, gràcies especials als Bombers de Palma, a l’Escola Municipal de Mallorquí, al Bar Ca’n Lliro, a l’Institució Pública M. Alcover, a l’Ajuntament de Manacor i a la Farmàcia Llull.
5
DEDICATION
Per sa meva illeta estimada, Sa Roqueta, i pes meus illencs preferits, es meus pares.
6
TABLE OF CONTENTS
LIST OF FIGURES……………………………………………………….……………..10
LIST OF TABLES ………………………………………………………………………11
ABSTRACT .…………………………………………………………………………….14
CHAPTER 1. INTRODUCTION .……………………………………………….……...15
CHAPTER 2. LITERATURE REVIEW .……………………………………..………...20
2.1. Majorca: Linguistic, historical and sociopolitical context ………………….20
2.2. Velar palatalization as a diachronic process………………………………...24
2.3. Velar palatalization in the history the Romance languages………………....31
2.4. The case of Majorcan Catalan ……………………………………………....38
2.5. Phonologization ……………………………………………….……………43
2.6. Consequences of consecutive bilingualism ………………………………...54
2.7. The present dissertation ………….………….………….………….……….60
CHAPTER 3. METHOD………….………….………….………….………….……..…63
3.1. Participants………….………….………….………….………….……….…63
3.2. Materials………….………….………….………….………….……………67
3.2.1. Catalan materials………….………….………….………….…….67
3.2.2. Spanish materials………….………….………….………….…….70
3.3. Procedure………….………….………….………….………….…………...73
3.4. Acoustic analysis………….………….………….………….………….…...74
3.5. Statistical analysis………….………….………….………….………….…..80
7
TABLE OF CONTENTS – Continued
CHAPTER 4. CATALAN………….…………………………………….……………...83
4.1. Introduction………….……………….……………….…….………….……83
4.2. Results.………….……………….……………….……………….…………85
4.2.1. Analysis of /k/ in Catalan………….……………………...…….…85
4.2.1.1. Center of Gravity………….……………….……………86
4.2.1.2. Relative Intensity………….……………….…………….88
4.2.1.3. Variance………….……………….……………….…….91
4.2.1.4. Summary of results for /k/ in Catalan…………………...93
4.2.2. Analysis of /a/ in Catalan.………….………….……………….….95
4.2.2.1. F1 measure………….……………….……………….….96
4.2.2.2. F2 measure………….……………….……………….….98
4.2.2.3. Summary of results for /a/ in Catalan…………...... …..100
4.3. Summary of Catalan findings…………….………………….…………….101
CHAPTER 5. SPANISH…………….……………………………………….…………107
5.1. Introduction…………….………………….………………….……………107
5.2. Results…………….………………….………………….…………………110
5.2.1. Analysis of /k/ in Spanish………………………...……………...110
5.2.1.1 Center of Gravity…………….………………….……...111
5.2.1.2. Relative Intensity…………….………………….……...113
5.2.1.3. Variance…………….………………….………………115
5.2.1.4. Summary of results for /k/ in Spanish………………..…..117 8
TABLE OF CONTENTS – Continued
5.2.2. Comparing Catalan and Spanish /k/……………………………...120
5.2.2.1. Manacor…………………….………………………….121
5.2.2.1.1. Center of Gravity…………….………………122
5.2.2.1.2. Relative Intensity…………….……………….124
5.2.2.1.3. Variance…………….………………….…….126
5.2.2.1.4. Summary: Manacor Catalan and Spanish /k/..129
5.2.2.2. Artà ….……………………………….………………..130
5.2.2.2.1. Center of Gravity…………….………………131
5.2.2.2.2. Relative Intensity…………….………………133
5.2.2.2.3. Variance…………….………………….…….135
5.2.2.2.4. Summary: Artà Catalan and Spanish /k/…….137
5.3. Summary of Spanish findings…………….………………….…………….139
CHAPTER 6. DISCUSSION AND CONCLUSION…………….…………………….142
6.1. Summary of aims and experiment …………….………………….……….142
6.2. Summary of results…………….………………….……………………….144
6.2.1. Catalan data…………….………………….……………………..144
6.2.2. Spanish data (and Catalan-Spanish comparison) ………………..145
6.3. Discussion…………….………………….……………………….………..146
6.3.1. Velar palatalization in different Majorcan Catalan areas………..146
6.3.2. Velar palatalization: Coarticulation or phonologization? ……….149
9
TABLE OF CONTENTS – Continued
6.3.3. Two phonological changes: Can we establish the order of the
influence?...... 152
6.3.4. Non-dominant speech of Catalan-Spanish bilinguals: Evidence of
phonetic transfer? …………………………..……..……………156
6.4. Future directions…………………………………………………………...164
6.5. Concluding remarks………………………………………………………..167
APPENDIX 1…………….………………….………………….………………………171
APPENDIX 2…………….………………….………………….………………………173
APPENDIX 3…………….………………….………………….………………………174
APPENDIX 4…………….………………….………………….………………………176
APPENDIX 5…………….………………….………………….………………………178
APPENDIX 6…………….………………….………………….………………………180
APPENDIX 7…………….………………….………………….………………………181
APPENDIX 8…………….………………….………………….………………………182
REFERENCES…………….………………….………………….…………………….183
10
LIST OF FIGURES
1. Center of Gravity for group and vowel context for Catalan ………………………87
2. Relative Intensity for group and vowel context for Catalan ………………….…...90
3. Variance for group and vowel context for Catalan ………………………………..92
4. Normalized F1 and F2 values for /a/ ………………………………………………96
5. Center of Gravity for group and vowel context for Spanish ……………………..112
6. Relative Intensity for group and vowel context for Spanish ……………………..114
7. Variance values for group and vowel context for Spanish ………………………116
8. Center of Gravity for language and vowel context for Manacor ………………...123
9. Relative Intensity for language and vowel context for Manacor ………………...126
10. Variance for language and vowel context for Manacor …………………………128
11. Center of Gravity for language and vowel context for Artà ……………………..132
12. Relative Intensity for language and vowel context for Artà ……………………..134
13. Variance for language and vowel context for Artà……………………………….136
11
LIST OF TABLES
1. Languages that present examples of velar softening………………………………27
2. Languages that present examples of velar palatalization………………………….30
3. Romance First Palatalization examples……………………………………………32
4. Romance Second Palatalization examples…………………………………………32
5. F1 and F2 values for the low vowel /a/ in four different dialects of Catalan……...42
6. Different hiatus resolution strategies applied to different Basque dialects………...46
7. /o/-lowering only before /r/, no /o/-lowering before other contexts……………….48
8. /o/-lowering before coronal obstruents in neighboring villages to Schaffhausen….49
9. Tone lowering after a voiced obstruent, sonorant, glottal stop…………………….51
10. Tone lowering after an implosive………………………………………………….52
11. Processes that show phonologization in some languages………………………….53
12. Social and linguistic information per participant from Manacor……….………….65
13. Social and linguistic information per participant from Artà.………………………66
14. Sample Catalan stimuli for /k/, for /p/, and filler words…………………………...69
15. Sample Spanish stimuli for /k/ in each vowel condition and of filler words………72
16. Results for Center Gravity for group and vowel context for Catalan……………...87
17. Results for Relative Intensity for group and vowel context for Catalan…………...90
18. Results for Variance for group and vowel context for Catalan …………………...92
19. Crossdialectal difference in Catalan /k/ per vowel context………………………..94
20. Results for F1 for group and consonant context…………………………………...98
21. Results for F2 for group and consonant context…………………………………...99 12
LIST OF TABLES – Continued
22. Crossdialectal difference in Catalan /a/ per consonant context…………………..100
23. Results for Center of Gravity for vowel context for Spanish ……………………111
24. Results for Relative Intensity for group and vowel context for Spanish ………...113
25. Results for Variance for group and vowel context for Spanish…………………..116
26. Crossdialectal difference in Spanish per vowel context and acoustic measure…..117
27. Results for Center of Gravity for language and vowel context for Manacor……..123
28. Results for Relative Intensity for language and vowel context for Manacor…….125
29. Results for Variance for language and vowel context for Manacor……………...128
30. Crosslinguistic difference per vowel context and acoustic measure for Manacor..129
31. Results for Center of Gravity for language and vowel context for Artà………….132
32. Results for Relative Intensity for language and vowel context for Artà …………134
33. Results for Variance for language and vowel context for Artà…………………..135
34. Artà crosslinguistic difference per vowel context per acoustic measure…………138
(a) Mean Center of Gravity values per group and vowel context for Catalan……….180
(b) Mean Relative Intensity values per group and vowel context for Catalan……….180
(c) Mean Variance values per group and vowel context for Catalan ………………..180
(d) Mean values for normalized F1 and F2 data per group and consonant context….180
(e) Mean Center of Gravity values per group and vowel context for Spanish ………181
(f) Mean Relative Intensity values per group and vowel context for Spanish ………181
(g) Mean Variance values per group and vowel context for Spanish………………...181
(h) Mean Center of Gravity values per language and vowel context for Manacor…..182 13
LIST OF TABLES - Continued
(i) Mean Relative Intensity values per language and vowel context for Manacor…..182
(j) Mean Variance values per language and vowel context for Manacor.…………...182
(k) Mean Center of Gravity values per language and vowel context for Artà……….182
(l) Mean Relative Intensity values per language and vowel context for Artà……….182
(m) Mean Variance values per language and vowel context for Artà………………...182
14
ABSTRACT
The present investigation examines the process of velar palatalization, a feature of Catalan, as seen in the Catalan and in the Spanish of the bilingual speech community of Majorca, Spain. Velar palatalization involves a change in a velar consonant’s place of articulation from velar to palatal; that is, /k, g/ acquire a secondary palatal articulation or acquire a completely new place of articulation. Velar palatalization usually occurs before /i, e, ɛ/ due to coarticulation. Some languages, however, also present this feature before /a, ə/ and word-finally. This is the case of certain dialects of Majorcan Catalan. Traditional descriptions have observed (a) the presence of velar palatalization before front vowels in all dialects of Majorcan Catalan (non-palatalizing area), and (b) the presence of velar palatalization also before /a, ə/, and word-finally only in certain areas of Majorca (palatalizing area). The aims of this dissertation are threefold. The first aim is to provide acoustic data for /k/ in the Catalan spoken in the traditionally palatalizing area, taking as an example the dialect of Manacor, a town of 43,000 inhabitants on the southeastern area of Majorca; and in the Catalan spoken in the non-palatalizing area, taking as an example the dialect of Artà, a town of 7,400 inhabitants on the northeastern coast of Majorca. Providing acoustic data for this contrast is relevant because it has only been documented through descriptive observations. Secondly, my dissertation analyzes vowel /a/ in the Catalan spoken in the two areas. It has been suggested that velar palatalization before /a/ can occur in languages in which /a/ is especially fronted (e.g. French). A comparison of /a/ production from both areas can provide clues regarding the relationship between the process of /k/ palatalization and /a/ fronting (i.e., if /a/ is equally fronted for both areas but there is a palatalization distinction, this could be interpreted as evidence that /a/-fronting triggered /k/-palatalization for the palatalizing area). The third aim of this dissertation is related to societal bilingualism. In situations of language contact, it is not unusual for cross-linguistic transfer to occur; that is, it is common for a bilingual’s language A to affect the perception and production of this bilingual’s language B speech sounds. In particular, this dissertation examines whether velar palatalization, for the individuals that prove to manifest it in their Catalan, is transferred from their first to their second language. The results suggest, first, that there exists, in fact, a difference in the distribution of the process of velar palatalization between the two communities and, importantly, that the process of velar palatalization has been phonologized for the speakers of the palatalizing area. Secondly, the findings suggest that the processes of velar palatalization and /a/-fronting might have stemmed from a relationship of mutual influence in its inception. Finally, there is no evidence of phonological transfer of the process of velar palatalization from dominant to non- dominant speech. The implications of these findings to theories of phonologization as well as of consecutive bilingualism are discussed.
15
CHAPTER 1
INTRODUCTION
Velar palatalization is a historical phonological process observed in many languages that belong to different language families, such as as Indo-European, Salish,
Mayan, Sino-Tibetan, Niger-Congo, among many others (Bhat, 1974; Bloomfield, 1933;
Buckley, 2003; Chen, 1973; Halle, 2005; Hock, 1986; Guion, 1996; Ladefoged &
Maddieson, 1996). This process involves a partial or complete change in a velar consonant’s place of articulation from velar to (alveo)palatal. A complete change would imply that the velar stop /k, g/ changes its place of articulation to a palatal position completely. The subsequent change, termed velar softening, consists in a change in manner rather than place of articulation, so that the palatalized velar stop becomes an affricate (e.g. [tʃ, ts, dʒ, dz]), and then, in some cases, a fricative (e.g. [ʃ, s, ʒ, z]). A partial change would entail the acquisition of a secondary palatal articulation, e.g. [kj, gj]
(Bhat, 1974). This dissertation is concerned with a partial velar palatalization process that allegedly characterizes Majorcan Catalan.
The fronting of /k/ usually occurs before front vowels /i, e, ɛ/ due to coarticulatory forces (common contexts for palatalization). Some languages, however, also present this feature in front of the low vowel /a/, in front of /ə/ and in word-final positions, which are far less common triggers for this process (uncommon contexts for palatalization). This is the case of certain dialects of Majorcan Catalan. Majorcan Catalan is a language that has traditionally been spoken on the Spanish island of Majorca. Traditional descriptive accounts have reported the presence of /k/-fronting in the common contexts (before front 16 vowels) in all dialects of Majorcan Catalan. Velar palatalization in the uncommon contexts (before /a/, /ə/, and word-finally) has been observed only in certain areas of
Majorca (Alcover, 1920; Barnils, 1933; Bernat i Baltrons, 2001; Bibiloni, 1983; Moll,
1952; Moll, 1980; Recasens, 1991; Veny, 1978). Notice, however, that the areas in which uncommon palatalization is observed do not form a logical or discreet area of palatalization. There is no clear geographical division between the palatalizing area (in which palatalization happens both in common and uncommon contexts) and the non- palatalizing area (in which palatalization only occurs in common contexts). For the purposes of the present study, however, we will treat the sum of the palatalizing areas as one area (palatalizing area) and the sum of the non-palatalizing areas as another area
(non-palatalizing area).
The present study analyzes the speech of people from two different areas of
Majorca, one of which belongs to the traditionally palatalizing area (Manacor), and the other one belongs to the traditionally non-palatalizing area (Artà). The first aim of the study is to provide acoustic data for the distribution of /k/ in the Catalan spoken in the two established areas of the island, a distribution that has only been reported through impressionistic observations so far (Alcover, 1920; Barnils, 1933; Bernat i Baltrons,
2001; Bibiloni, 1983; Moll, 1952; Moll, 1980; Recasens, 1991; Veny, 1978). Production of /k/ is analyzed in front of different vowel contexts in order to potentially find crossdialectal differences in velar palatalization. One of the ways in which these differences might be present is in terms of the presence or absence of velar palatalization of /k/ in different vowel contexts, i.e., some vowel contexts might trigger velar palatalization for one of the two areas but not for the other one, some vowel contexts 17 might trigger velar palatalization for both areas, some vowel contexts might not trigger velar palatalization in either area. Results such as those would help confirm the presence of a phenomenon that has only been reported in traditional descriptive accounts. Another type of crossdialectal difference might be found in the degree of palatalization of /k/ for a specific vowel context, i.e., the size of /k/-fronting for a specific vowel context might be larger in one of the two geographical areas. For instance, vowel context /i/ is always expected to trigger /k/-fronting due to coarticulatory forces. However, it could be the case that the two geographical areas present differences in the degree of fronting in this vowel context. A more fronted /k/ before /i/ in one of the two areas would suggest that this phenomenon goes beyond palatal coarticulation and has become a phonologized process.
Thus, the present dissertation not only evaluates crossdialectal differences in terms of velar palatalization being applied in more contexts, but it also compares differences based on the size of /k/-fronting in order to see if the process has been phonologized.
Apart from analyzing the realizations of the velar stop consonant /k/, the present study analyzes the low vowel in the Catalan spoken in the two areas. It has been suggested that velar palatalization before /a/ can occur in languages in which such low vowel is especially fronted (e.g. French) (Buckley, 2003). If /a/ is equally fronted for both areas but there is a palatalization distinction, this could be interpreted as evidence that /a/- fronting triggered /k/-palatalization for the palatalizing area. If the palatalizing area presents /k/-palatalization and /a/-fronting but the non-palatalizing area does not present either of the two processes, this would indicate a relationship between both processes, without determining which process caused which. 18
The final aim of this paper is related to the bilingual nature of the island of
Majorca. Majorca is a Spanish island in which Majorcan Catalan shares co-officiality with the national language, Spanish. In situations of language contact, it is not unusual for crosslinguistic transfer to occur (Thomason & Kaufman, 1988; Van Coetsem, 1988;
Winford, 2005). Studies about these bilinguals have shown that in spite of having an early and extensive exposure to a second language, performance of bilinguals in their non-dominant language differs from that of bilinguals with that language as their dominant language (Amengual, 2011, 2016; Simonet, 2010, 2011a, 2011b). The present study seeks to analyze the productions of bilinguals that are Catalan-dominant to find possible patterns of crosslinguistic transfer from Catalan, their dominant language, onto
Spanish, their non-dominant language. In particular, this study examines whether velar palatalization, for the individuals that prove to manifest it in their Catalan, is transferred from their first to their second language.
This dissertation is structured as follows. Chapter (2) overviews the literature relevant to the present study. The chapter is divided different subsections. First, the linguistic, historical and sociopolitical context of Majorca is discussed. Second, a typological, crosslinguistic summary of the process of velar palatalization is conducted, which is a very common process in languages from different linguistic families. Third, the phenomenon of phonologization is described, and its importance for our understanding of the process of velar palatalization is discussed. Then, an overview of previous studies analyzing the effects of consecutive bilingualism on production is conducted, with a focus on the literature on Catalan-Spanish language contact. The chapter ends with a summary of the present study and research questions. Chapter (3) 19 reviews information about the methodology used for the present study, such as participants, materials, procedure, acoustic analysis and statistical analysis. Chapter (4) provides the findings for the Catalan data, both for /k/-production (organized by each of the three acoustic measures: Center of Gravity, Relative Intensity, Variance) and for /a/- production (organized by the two acoustic measures: F1 and F2). Chapter (5) summarizes the findings for the Spanish data only regarding /k/-production first (organized by each of the three acoustic measures: Center of Gravity, Relative Intensity, Variance) and followed by the findings for the within-speaker comparison between Catalan and Spanish productions (divided by area first and then by acoustic measure within each area).
Chapter (6) contains an overall discussion of the findings of the present study, discussing their implications for theories of phonologization, velar palatalization and crosslinguistic transfer in bilinguals.
20
CHAPTER 2
LITERATURE REVIEW
2.1. Majorca: Linguistic, historical and sociopolitical context
The presence of Catalan in Majorca dates back to the 13th century, with the arrival of the King of Aragon, James I the Conqueror, and his conquest of the island from the Arabs in 1229. People from Roselló, Girona and Barcelona, all of these Catalan- speaking regions of the Old Kingdom of Aragon, colonized the island, which explains why the Balearic inhabitants started speaking Catalan in its eastern variety (Blas Arroyo,
2007). Since most surviving Arabs were expelled from Majorca or forced to slavery after the colonization of the island by the Catalans, the linguistic impact of the language of previous inhabitants of Majorca onto Catalan was minimal (Moll, 1980, p. 32). As
Miralles i Montserrat (1989) discusses, during the three centuries following the colonization, Catalan was the main language of the territory, with minor linguistic contact with Latin (the language of high culture) and other minority languages, such as Arabic,
Hebrew, Occitan, or Aragonese (as cited in Amengual, 2013).
Historians maintain that Spanish was not spoken on the island until the late 14th century, when the Crown of Aragon, a union of territories from Eastern Spain and the
Mediterranean, was unified with Castile, which comprised most Western and Central
Spain. This unification, which was due to the dynastic union between Ferdinand V and
Isabella of Castile (the Catholic Monarchs), brought the centralization of political power in Castile, which was mainly Spanish-speaking, and the subsequent penetration of
Spanish into traditionally Catalan-speaking territories (Amengual, 2013). However, it 21 was not until the 18th century, with the enactment of the Nueva Planta Decrees that
Majorca started to experience a diglossic situation: Spanish was used in political, administrative, cultural and ecclesiastical domains, while Catalan was used only in the private domains. This linguistic situation was reinforced during the military regime of
Francisco Franco (1939-1975) who imposed language policies based on the repression of autochthonous languages, and during the touristic boom of the 1960s that brought tourism and immigration from monolingual territories of Spain to the islands. It was not until the 1980s that the situation started to change, when the Linguistic Normalization
Law granted co-official status with Spanish to the autochthonous languages that were traditionally spoken in those territories. Since then, different linguistic strategies and policies have tried to re-establish the use of such languages in society (Pieras, 1999).
Today, both languages are used in all realms of society, from government, administration, education and culture to informal and private domains, which makes it a difficult task to determine which of the two languages is the majority language (Simonet, 2011b).
According to some, the situation in these territories can, then, be characterized as a situation of stable societal bilingualism.
The demographic changes in the recent history of Majorca have led to postulate a clear division between a pre-touristic Majorca (traditional Majorca) and a post-touristic
Majorca (urban Majorca) (Radatz, 2007). Before the arrival of thousands of families from the Spanish mainland to the island of Majorca due to the touristic boom and the arrival of immigrant waves in the last decades, the linguistic landscape of the island could be seen as the sum of multiple isolated subdialects of Majorcan Catalan. Before the social and demographic changes, rural areas were somewhat isolated from the capital city, Palma, 22 and from one another. The island, however, can now be seen as a more uniform dialectal unit. Towns are better connected and islanders balance life and work between the rural areas and the capital city. The city and the towns are no longer two different and separate worlds (Melià, 1998). It is not unusual for people from the rural areas to commute to
Palma for work and return to the rural areas at night. It is not unusual either for people from Palma to have a weekend country house in a rural area. There is, therefore, bidirectional influence between rural and urban Majorca and both can currently be considered a single geographical and linguistic entity. This does not mean that all idiosyncratic features of the different areas of the island are lost, but there is a general tendency towards regional dialect levelling that reduces the differences between areas
(Kerswill, 2003).
In spite of such ‘presumed’ dialect levelling, Catalan in Majorca presents multiple subdialectal traits that have been maintained until today (Moll, 1960). For instance, the
Catalan spoken in the town of Sóller, one small Majorcan village situated on the
Northwestern coast of the island, presents a pattern of unstressed vowel reduction that differs from vowel reduction patterns found in the rest of Majorcan dialects (and that is similar to those present in other Eastern Catalan dialects spoken in the mainland)
(Llompart & Simonet, 2015). In unstressed position, whereas most Majorcan Catalan dialects (and other Western Catalan varieties) merge the three back vowels into two phonetic categories (/ɔ/ and /o/ merge into [o]; /u/ remains [u]), in Sóller Catalan, all three back vowels are merged into a single phonetic category, [u]. For instance, dona
[ˈdɔnə] ‘woman,’ poma [ˈpomə] ‘apple,’ cuna [ˈkunə] ‘cradle’ contain /ɔ, o, u/ that merge 23 into [u] when it appears in unstressed position, thus giving doneta [duˈnətə] ‘young/little woman,’ pometa [puˈmətə] ‘little apple,’ cuneta [kuˈnətə] ‘little cradle.’
Another idiosyncratic feature of only certain subdialects of Majorcan Catalan is found in the towns of Lloseta, Binissalem, Alaró, Sineu, Felanitx, Maria de la Salut, and
Sant Joan, all of which are villages found in the interior of the island (except for Felanitx, which is found on the Southeastern coast of the island). Majorcan Catalan is known for its use of /ə/ in stressed positions, as in pera [ˈpəɾə] ‘pear’. In these towns, however, stressed /ə/ has shifted to /ɛ/, as observed in other Eastern varieties of Catalan
(Amengual, 2016; Moll, 1968), as seen in pera [ˈpɛɾə] ‘pear’. Also in the town of Sineu there is another unique feature that differs from the rest of the dialects on the island
(Caimari Frau, 1983). When appearing after a palatal(ized) consonant, vowel [ə] (cavall
[kjəˈvaʎ] ‘horse’) becomes [e] (cavall [ˈkjeˈvaʎ] ‘horse’). Similarly, stressed [a] (casa
[ˈkjazə] ‘house’) becomes [ɛ] (casa [ˈkjɛzə] ‘house’). The town of Felanitx presents a similar change from the one just mentioned for Sineu, but it occurs in all consonantal contexts. In this town, vowel [a] (pasta [ˈpastə] ‘house’) becomes [ɛ] (pasta [ˈpɛstə]
‘house’), but vowel [ɛ] also changes (ferro [ˈfɛro] ‘house’), and it becomes [e] (ferro
[ˈfero] ‘house’) (Moll, 1960).
Another dialectal peculiarity is found in the Northern town of Pollença. In
Majorcan Catalan, differently from the rest of Catalan dialects, there is a generalized use of the article salat, which consists of a special set of forms of the definite article (mainly used in spoken language): es [ˈəs], sa [ˈsə], es [ˈəs], ses [ˈsəs], s’ [s] (Wheeler, Yates, &
Dols, 1999). Differently from this Majorcan rule, the village of Pollença displays the use of the regular definite articles present in all other dialects of Catalan: el [ˈəɫ] or even [ˈəu] 24
‘the’ (masculine singular), la [ˈɫə] ‘the’ (feminine singular), els [ˈəɫs] ‘the’ (masculine plural), les [ˈɫəs] ‘the’ (feminine plural), l’ [ɫ] ‘the’ (masculine/feminine singular when preceding a word that starts with a vowel) (Wheeler, Yates, & Dols, 1999). Note that all these features are subdialectal traits only found in certain areas of Majorca and do not follow general Majorcan Catalan phonological rules. This old dialectal variation is mostly maintained in Majorca nowadays.
The present dissertation analyzes several features of the Catalan variety spoken in
Majorca. Specifically, it will analyze the phonetic phenomenon of velar palatalization, which has been attested only in certain areas of the Majorcan territory (Recasens, 1991).
Additionally, it will analyze the transfer of this phonetic feature onto the Spanish spoken on the island. The following section describes and reviews the process of velar palatalization.
2.2. Velar palatalization as a diachronic process
Velar palatalization is one of the most common sound changes, and it has been attested in many languages from different language families (Bhat, 1974; Bloomfield,
1933; Buckley, 2003; Chen, 1973; Halle, 2005; Hock, 1986; Guion, 1996; Ladefoged &
Maddieson, 1996; among many others). Velar palatalization involves a partial or complete change in a velar consonant’s place of articulation from velar to (alveo)palatal.
This means that the change can result in two different outputs. A first output entails the velar stop consonant changing its place of articulation completely to a palatal position, a process known as complete velar palatalization. This process is sometimes followed by a change in manner of articulation, commonly known as velar softening, which entails the 25 stops evolving into alveopalatal affricates (e.g. /tʃ, ts, dʒ, dz/) or fricatives (e.g. /ʃ, s, ʒ, z/). In the second possible output, rather than a complete change in place of articulation, the velar stop consonant acquires a secondary palatal articulation (e.g. /kj, gj/) (Bhat,
1974). The first change implies a complete change in place of articulation first, followed by a change in manner of articulation (velar softening). The second change implies a fronting of the velar consonant, which is produced as a fronted post-palato velar stop consonant (Recasens, 2014). Velar palatalization could, therefore, be seen as the different stages through which a velar stop is fronted and changed in manner of articulation: /k/ >
/kj/ > /c/ > /ts, tʃ/ > /s, ʃ/; /g/ > /kj/ > /ɟ/ > /dz, dʒ/ > /z, ʒ/. The most common result in the world’s languages is an alveopalatal affricate, especially for voiceless stops (Guion,
1996, 1998; Halle, 2005). Not all languages, however, show the change to be exactly at that stage. Some languages maintain the original velar stop at an earlier stage of the change, e.g. produced as a post-palato fronted velar /kj, gj/, or as an intermediate alveopalatal stop /c, ɟ/. Some other languages have the original velar stop at a later stage of the change, e.g. changed into a fricative /s, ʃ/.
In the majority of cases, these changes are triggered by the presence of a front vowel /i, e, ɛ/ or a palatal glide /j/ after the stop consonant (Bhat, 1974). However, there is another context that can, to a much lesser extent and in a smaller amount of languages, trigger palatalization of a velar stop consonant. This is the case of the low vowel /a/.
Some languages, such as French, Picard, Sutselvan Romansh or Fassan Ladin show examples of velar consonants that have been palatalized before the low vowel /a/
(Recasens, 2014). Relevant to this dissertation, this seems to be the case of certain varieties of Majorcan Catalan, where the velar stop consonants are allegedly fronted into 26 an intermediate alveopalatal stop consonant [c, ɟ] before front vowels, low vowel /a/ and in word-final position.
As mentioned above, Guion (1996) explains that the first process, the one rendering a complete change in place and manner of articulation, is the most common process to have affected velar stops. Many of the world’s languages show examples of velar softening. See Table 1 below for a summary of languages that have undergone velar softening.
As explained in Guion (1996), Slavic offers an example of different processes of palatalization of velars resulting in a coronal: the First and Second Slavic Palatalizations.
These changes happened in Proto-Slavic (PS) and are shared by all Slavic languages. The
First Palatalization changed [k] to [tʃ] (e.g. PRE-PROTO-SLAVIC [PPS] wilk-e > OLD
CHURCH SLAVIC [OCS] vlǐtʃe ‘wolf’), [g] to [ʒ] (e.g. PPS mog-e > OCS moʒe ‘was able’) and [x] to [ʃ] (e.g. PPS dowx-e > OCS duʃe ‘ghost’). The Second Palatalization changed [k] to [ts] (e.g. PPS vlǐki > OCS vlǐtsi ‘wolves’), [g] to [dz] (PPS gɛlo > OCS dzɛlo ‘strongly’).
In most Slavic languages, additionally, [dz] evolved into [zj]. East Slavic languages, such as Russian, Ukrainian or Belorussian, and South Slavic languages such as Old Church
Slavic, Bulgarian, Macedonian, Serbo-Croat, Slovene further developed [x] onto [s] (e.g.
Late PS duxi > OCS dusi). West Slavic languages, such as Czech, Slovak, Upper and
Lower Serbian, Polish, Cassubian, developed [x] into [ʃ].
27
Table 1. Languages that present examples of velar softening 1 : language, change, example, and translation
LANGUAGE CHANGE EXAMPLE TRANSLATION PRE-PROTO-SLAVIC (1st) /k/ > /tʃ/ wilk-e > vlǐtʃe ‘wolf’ /g/ > /ʒ/ mog-e > moʒe ‘was able’ /x/ > /ʃ/ dowx-e > duʃe ‘ghost’
PRE-PROTO-SLAVIC (2nd) /k/ > /ts/ vlǐki > vlǐtsi ‘wolves’ /g/ > /dz/ gɛlo > dzɛlo ‘strongly’
INDO-IRANIAN /k/ > /tʃ/ ke > tʃa ‘and’ /g/ > /dʒ/ gi:wo-s > dʒi:vas ‘alive’
COWLITZ SALISH /k’/ > /tʃ’/ k’ilk > tʃ’ilk ‘window’ /k/ > /tʃ/ kitaq- > tʃæ:q- ‘argue’ /x/ > /ʃ/ túlxils- > túlʃils- ‘hint’
JOMVU (Bantu) /k/ > /tʃ/ -kéngédé > tʃendʒele ‘bell’ /g/ > /dʒ/ -ingid- > -ndʒi-a ‘enter’
SHONA (Bantu) /k/ > /ts/ -kíndo > mu–tsíndo ‘footstep’ /g/ > /z̤/ -gína > z̤íná ‘name’ OLD ENGLISH /k̟/ > /tʃ/ k̟irik̟e > tʃɹ̩ tʃ ‘church’ /ɣ̟ / > /j/ ɣ̟ eard > jɑrd ‘yard’ /k̟k̟/ > /tʃ/ wæk̟k̟ing(e) > wɑtʃɪŋ ‘watching’ /ɣ̟ ɣ̟ / > /dʒ/ kyg̟ g̟ el > kʌdʒəl ‘cudgel’ MAM (Mayan) /k/ > /tʃ/ ke:x > tʃe:j ‘horse, deer’
OLD CHINESE /k/ > /tɕ/ kje > tɕje ‘branch’ /kh/ > /tɕh/ khjet > tɕhjet ‘to trail, drag’ /g/ > /dʑ, ʑ/ gjip > dʑjip/ʑjip ‘ten’ /x/ > /ɕ/ xjiw > ɕjuw ‘catch’
1 All examples (including phonetic transcriptions) have been extracted from Guion (1996). 28
The Indo-Iranian branch of Indo-European also offers an example of palatalization of velars before front vowels. Velars ([k], [g]) were changed to alveopalatal affricates ([tʃ], [dʒ], respectively) before a front vowel (Hoffmann, 1982; Mayrhofer,
1965), e.g. PRE-PROTO-INDO-IRANIAN gi:wo-s, ke > SANSKRIT (ancient Indic language) dʒi:vas, tʃa, AVESTAN (ancient Iranian language) dʒi:va-, -tʃa ‘alive’, ‘and’, respectively.
The Salishan languages present a split between languages that maintained velar stops and those that palatalized their velars onto alveopalatals. Specifically, Cowlitz
Salish presents both the velars and the alveopalatal sounds, overlapping in distribution.
Velars were palatalized before [i]: PROTO-SALISH k’ilk, kitaq-, tútxils > COWLITZ SALISH tʃ’ilk, tʃæ:q-, túlʃils- ‘window’, ‘argue’, ‘hint’, respectively.
Bantu languages offer two examples of velar palatalization. A first palatalization originated from a merger with a palatal glide [j], which was analogically extended to velar palatalizations before front vowels. This is illustrated by Standard Swahili, which shows prefix /ki-/ produced as [ki] when followed by a consonant (e.g. ki-jito ‘brook’) but as [tʃ] when followed by a vowel (e.g. tʃ-ama ‘society’). The vowel /i/ was glided historically and then merged with the preceding velar to produce [tʃ]. A second velar palatalization can be observed in some Bantu languages, in which velar palatalization has been extended to other environments through analogy, e.g. Mwiini, Kanyok and some dialects of Nyambo palatalize velars only in prefixes before front vowels; Bemba, Cewa,
Tonga and Shona palatalize velars only in morpheme initial position; Cifundi and Jomvu palatalize velars only in morpheme internal environments; and Yao, Tumbuka, Mwera and Nyankore-Kiga palatalize all velars before front vowels independently from morphological environment. For example, Jomvu palatalizes velars morpheme-internally 29 before front vowels, as seen in: PROTO-BANTU –kéngédé > JOMVU tʃendʒele ‘bell’.
Another example comes from Shona, in which [g] became [z̤] (coronal murmured fricative) (e.g. PROTO-BANTU –gína > SHONA z̤í n á , ‘name’) and [k] became [ts] before [i]
(PROTO-BANTU -kíndo > SHONA mu – tsíndo ‘audible footstep’).
English also shows velar palatalization. Old English fronted [ɣ̟ ] and [k̟ ] evolved to
[j] and [tʃ], respectively, in Middle English word initially before a front vowel (æ, ǣ, e, ē, i, ī) (e.g. OLD ENGLISH [OE] k̟ irik̟ e > MODERN ENGLISH [MoE] tʃɹ̩tʃ), word internally between two front vowels or between a front vowel and a syllabic consonant, and word finally after front vowels. Additionally, geminate velars (voiced and voiceless) were also palatalized before a front vowel: fronted [k̟k̟] became [tʃ] (e.g. OE wæk̟k̟ing(e) > MoE wɑtʃɪŋ ‘watching’) and fronted [g̟ g̟ ] became [dʒ] (e.g. OE kyg̟ g̟ el > MoE kʌdʒəl ‘cudgel’).
Most dialects of Mam, a Mayan language, change [k] to different coronal affricates before front vowels [i] and [e]: Ostumcalco and Ixtahuacán change it to a laminal alveopalatal affricate [tʃ] (e.g. tʃe:j); Todos Santos changes it to an apical post alveolar affricate [tɻ̝̊] (e.g. tɻ̝̊ e:j). Only one dialect of Mam, Tacaná, changes it to palatal stop [c] (e.g. ce:j ‘horse, deer’).
The final example Guion (1996) mentions is the case of Chinese. Old Chinese
h (OC) presented velars ([k, k , g, x]) that were later changed into alveopalatal allophonic variants of velars ([tɕ, tɕh, dʑ/ʑ, ɕ]) before the combination of a palatal glide and a front
h h vowel in Middle Chinese (MC) (e.g. OC kje, k jet, gjip, xjiw > tɕje, tɕ jet, dʑjip/ʑjip, ɕjuw
‘branch’, ‘to trail, drag’, ‘ten’, ‘catch’, respectively.
After reviewing languages which show the development of a completely palatalized sound, Guion (1996) continues to explain that a second common output 30
(though not as common as the first one) of velar palatalization is a palatalized velar /kj/.
She explains that this outcome is usually the result of a palatalization of velars that is often extended to the palatalization of stops in other places of articulation, as seen in the cases of Russian and Irish Gaelic exemplified (Table 2) and discussed below.
Table 2. Languages that present examples of velar palatalization: language (first column), change (second column), example (third column) and translation (fourth column)
LANGUAGE CHANGE EXAMPLE TRANSLATION
j j RUSSIAN /k/ > /k / kit > k it ‘whale’
/g/ > /gj/ genɪ >gjenjɪ(j) ‘genius’
/x/ > /xj/ xitrɨ(j) > xjitrɨ(j) ‘cunning’
j j 2 IRISH GAELIC /k/ > /k / k a:n ‘head’
The case of Russian starts after the above-mentioned first palatalization in Slavic languages. By the 11th century AD, the labial, alveolar and velar consonants (these last ones a little later) had adopted palatalized variants before front vowels and [j], thus leading to the presence of separate phonemes for the palatalized and non-palatalized varieties of the bilabial stops and nasal ([p, b, m] vs. [pj, bj, mj]), dental stops and nasals
([t, d, n] vs [tj, dj, nj]), labiodental and alveolar fricatives, ([f, v, s, z] vs. [fj, vj, sj, zj]) and the liquids ([l, r] vs. [lj, rj]) (Mathews, 1967). Modern Russian presents a phonemic split between non-palatalized velars occurring before non-front vowels (e.g. kak ‘how’), consonants (e.g. kuklə ‘doll’) and word finally (e.g. rok ‘fate’) and palatalized velars
2 Whereas all other examples are taken from Guion (1996), the example on Irish Gaelic is taken from Hickey (1985). 31 appearing before front vowels (e.g. kjit ‘whale’, kjɛm ‘whom’). Additionally, Modern
Russian presents velar palatalization in front of non-front vowels within loanwords
(Jones & Ward, 1969) (e.g. tkjot ‘weaves’).
Irish Gaelic also presents cases of secondary palatalization. Old Irish already presented a split between non-palatalized and palatalized consonants, similar to what happened in Russian: [p, b, t, d, k, g, m, n, ŋ, f, w, v, s, x, ɣ, r, l] vs [pj, bj, tj, dj, kj, gj, mj, nj, ŋj, fj, wj, vj, sj, xj, ɣj, rj, lj]. These palatalized phonemes derived from the loss of final vowels in Pre-Old Irish.
To sum up, the process of velar palatalization is one of the most common sound changes and it has been attested in many different languages. The process can lead to the development of a fully palatalized sound or to the development of a palatalized velar. In any case and as observed, the process is usually triggered by the presence of a front vowel or a glide after the velar consonant.
2.3. Velar palatalization in the history the Romance languages
Velar palatalization is a historical process that has also affected Romance languages. As summarized in Calabrese (1993), the Latin consonant system was changed when palatal consonants developed into the Romance consonant system. The history of
Common Romance (as Buckley 2003 names it) has been divided into two different stages of palatalization (Calabrese, 1993). The first stage was seen in Late Latin or Proto-
Romance (1st and 2nd c. AD) and it affected all consonants (independently of place of articulation) before glide /j/. This change affected all Romance languages. (See Table 3.) 32
The second stage of velar palatalization in Late Latin or Proto-Romance
(sometime between the 3rd and 5th centuries AD; Buckley, 2003; Calabrese, 1993) affected only velar stops /k, g/ before front vowels /i, e, ɛ/. (See Table 4.) This change affected all Romance languages, except for Sardinian and Dalmatian (Hall, 1976, p. 197).
Table 3. Romance First Palatalization examples: horizontally (Latin, Proto-Romance,
Spanish, Italian, Catalan, French, translation), vertically (examples for any consonant +
/j/)3
LAT. PR. ROM. SPAN. IT. CAT. FR. TRANS. diurnu diurnu ______dʒorno dʒorn dʒorn ‘day’
______basiare bezar baʃare bezar baiziær ‘kiss’ kastanea kastania kastaɲa kastaɲɲa kastaɲa kastaiɲe ‘nut’ palea palia paʒa paʎʎa paʎə paʎə ‘straw’
Table 4. Romance Second Palatalization examples for Latin, Proto-Romance, Spanish,
Italian, Catalan, French, translation4
LAT. PR. ROM. SPAN. IT. CAT. FR. TRANS. fakie- fakie /ia hats fatʃtʃa fats fatsə ‘face’ pantike- pan(ti)kia pantʃa pantʃa pantʃa pantʃə‘paunch’
3 These examples have been extracted from Hall (1976) 4 These examples have been extracted from Hall (1976) 33
Notice that this division between the two stages of palatalization in Common
Romance should not be confused with the two stages of palatalization usually described for French. Within the French sound change literature, the French First Velar
Palatalization usually refers to the second stage just mentioned above for Common
Romance (/k, g/ palatalizing before front vowels) (Buckley, 2003). The French Second
Velar Palatalization refers to a later sound change (late 5th century AD or early 4th century
AD) in Old French that also affected velar stop consonants but this time in front of the low fronted vowel /a/. For the purposes of the present dissertation, the French Second
Velar Palatalization will be named Third Velar Palatalization. The case of French will be discussed later.
In Spanish, velar stops became alveopalatal affricates when followed by the front vowels (/i/, /e/) or by a glide (/j/) (Penny, 1991). Regarding voiceless velar /k/, when this sound appeared before a front vowel or a glide, the result was an affricate /ts/ both in
5 word-initial position (e.g. PROTO-ROMANCE kera > OLD SPANISH tsera ‘wax’ ) and after consonant before front vowel (e.g. PROTO-ROMANCE dulke > OLD SPANISH dultse
‘sweet’). In intervocalic position, the velar voiceless changed into the affricate /dz/ (e.g.
PROTO-ROMANCE fakere > OLD SPANISH hadzer ‘to do’). Notice that these changes did not happen in front of a back vowel either in word-initial position (e.g. PROTO-ROMANCE kognate > OLD SPANISH kuɲada ‘sister-in-law’) or in intervocalic position, where /k/ became voiced /g/ (e.g. PROTO-ROMANCE sekûru > seguro ‘safe’). The case of the voiced velar /g/ is less clear. Whereas this sound changed to /dʒ/ before front vowels in most
Western Romance languages (Boyd-Bowman, 1954; Hall, 1976; Pope, 1934; Price,
5 All examples (and transcriptions) for Spanish have been extracted from Hall (1976). 34
1971), in Spanish, it gave rise to several different outputs. In initial position, /g/ changed into /dʒ/ (PROTO-ROMANCE gelare > OLD SPANISH dʒelar ‘to freeze’), /h/ (PROTO-
ROMANCE gelu > OLD SPANISH hielo ‘frost’), /j/ (PROTO-ROMANCE generu > OLD
SPANISH jerno ‘son-in-law’), Ø (PROTO-ROMANCE gingi̯ua > OLD SPANISH entsia ‘to freeze’). In intervocalic position, /g/ disappeared (PROTO-ROMANCE rêge > OLD SPANISH rrei ‘king’). In spite of all this variation, some lexical items with /g/ did, in fact, show instances of velar palatalization, as seen with the example in which /g/ becomes /dʒ/
(Hall, 1976, pp. 196-197).
Catalan went through a similar process of velar palatalization as Spanish did.
Velar stops in Catalan became affricated into /ts/ and /dʒ/ before front vowels (/e, i/).
Additionally, these Catalan affricates were further changed into fricatives /s, ʒ/ (e.g.
PROTO-ROMANCE caelu, gelu > OLD CATALAN tselu, tʒelu > MODERN CATALAN /s/el,
/ʒ/el ‘sky’, ‘ice’6). These changes did not take place in front of non-front vowels (e.g.
PROTO-ROMANCE coda, gula > MODERN CATALAN coa, gola ‘tail’, ‘gluttony’, respectively). In intervocalic position, voiceless velars became voiced before non-front vowels (e.g. PROTO-ROMANCE secundu > MODERN CATALAN segon ‘second’), and became /z/ (and were then dropped) before front vowels [PROTO-ROMANCE placere >
OLD CATALAN plaser > MODERN CATALAN plaer ‘pleasure’]. Regarding /g/ in intervocalic position, it disappeared before front vowels (e.g. PROTO-ROMANCE magistru
> MODERN CATALAN mestre ‘teacher’) and before non-front vowels in postonic position
(e.g. PROTO-ROMANCE tegula > MODERN CATALAN teula ‘tile’), but the velar was maintained in pretonic position when followed by a non-front vowel (e.g. PROTO-
6 The examples for Catalan have been extracted from Moll (1952). 35
ROMANCE legumen > MODERN CATALAN llegum ‘vegetable’). Importantly, note that both in Catalan and in Spanish, it is possible to find words containing word-initial /k/, which is explained through the introduction or reintroduction of such terms at a state in which velars had already gone through the sound change in which they were modified (Badia
Margarit, 1951; Fouché, 1924; Moll, 1952).
Finally, and relevant to the present dissertation, consider the case of French.
French Velar Palatalization took place in two stages, as mentioned above, traditionally known as the French First and Second Velar Palatalization (Buckley, 2003; Jacobs &
Berns, 2013) but referred to as Second and Third Velar Palatalization, respectively, here.
As mentioned before, the Second Velar Palatalization that affected most Romance languages took place in Late Latin or Proto-Romance sometime between the 3rd and 5th centuries AD. Velar stops were first palatalized (/kj /, /gj/) before front vowels /i/, /e/ and
/ɛ/. Once Latin developed into different Romance languages these sounds were affricated
(> /ts/, /dʒ/) in the same conditions (and later changed into a fricative /s/, /ʒ/ in Modern
j j French) (e.g. PROTO-ROMANCE k entu, g ente > GALLO-ROMANCE centu, ɟente OLD
7 FRENCH tsentu, dʒente > MODERN FRENCH sã, ʒã ‘cent’, ‘people’ ) (Jacobs & Berns,
2013). The same palatalization occurred when the velar appeared after another consonant
(e.g. GALLO-ROMANCE dulke, margin > OLD FRENCH dôlts, mardʒe ‘sweet’, ‘edge’, respectively). This change, however, did not happen word-initially before back vowels
(e.g. GALLO-ROMANCE kornûtu, gustare > OLD FRENCH kornuð, gôstar ‘horned’, ‘to taste’, respectively); or in intervocalic position, a position in which velar consonants were dropped both before front vowels (e.g. GALLO-ROMANCE fakere, rêge > OLD FRENCH
7 Examples on French have been extracted from Hall (1976) and from Jacobs & Berns (2013). 36
8 fairə, rêi ‘to do’, ‘king’, respectively) and before back vowels (e.g. GALLO-ROMANCE loku > OLD FRENCH lieu ‘place’).
After this Romance Second Velar Palatalization, a third stage of palatalization affected French (and French exclusively) around the late 5th century AD or the early 6th century AD (Buckley, 2003; Pope, 1934; Price, 1971). This change did not affect all of the Romance languages, but was limited to some of the Gallo-Romance dialects (e.g.
Picard, Sutselvan Romansh and Fassan Ladin). Specifically, the French Third Velar
Palatalization affected word-initial velars in front of the low vowel /a/ (e.g. LATE LATIN
j j karru, gamba > GALLO-ROMANCE k arru, g amba > carru, ɟamba > OLD FRENCH tʃar, dʒambə > MODERN FRENCH ʃaʁ, ʒamb ‘wagon, ‘leg’) and word-internal velars after consonant before front vowel /a/ (e.g. LATE LATIN iskala, largam > GALLO-ROMANCE
j j isk ala, larg a > iscala, larɟa > OLD FRENCH estʃiaelo, lardʒe > MODERN FRENCH eʃɛl, laʁʒ ‘ladder, ‘large’). As in the previous stage of palatalization, this change was not applied in intervocalic position, since velars seemed to disappear in this position independently of the following vowel (e.g. PROTO-ROMANCE amika, doga > OLD FRENCH amiə, dôvə ‘friend’, ‘stave’, respectively). As explained in Jacobs and Berns (2013), additionally, velar /k/ and /g/ followed by front vowels, which had evolved into a palatalized velar during the Second Velar Palatalization, were re-introduced in Gallo-
Romance through borrowings from other languages (from Germanic or Arabic origins).
These forms also went through the process of palatalization in the same way that velars before the low vowel did (they became /tʃ/ instead of /ts/) (e.g. GALLO-ROMANCE (e)skina
8 Price (1971) suggests that intervocalic velars followed the same pattern than word- initial velars /k/ > /ts/. After this, this affricate was voiced into /dz/ and then lenited into a palatal glide /j/ and then disappeared. 37
j (e)sk ina > escine > OLD FRENCH estʃine > MODERN FRENCH é[ʃ]ine, ‘corner’). Notice that the output for voiceless /k/ in the French Third Velar Palatalization, /tʃ/, did not merge with the result of the Second Velar Palatalization, /ts/. The same cannot be said about voiced /g/, which became /dʒ/ in both processes (Buckley, 2003).
The Third Velar Palatalization process has been the object of many discussions because /a/ is not a crosslinguistically common trigger of a palatalization change
(Buckley, 2003; Guion, 1998; Jakobs & Berns, 2013). Buckley (2003) proposes that this change initially originated due to the presence of allophonic variation of vowel /a/ between a fronted [æ] and a non-fronted [a]. Fronted [æ] is a far more common trigger of palatalization across languages than [a] is (e.g. Persian, Late Old English). The coarticulation between a velar stop consonant with a fronted [æ] could have triggered the phonetic alternation between [k] and [kj] (initially only before allophone [æ], not before allophone [a]). Since [æ] was just a variant of phoneme /a/, it seems possible that there was a phonological categorization that extended sound context [æ] to [a] within the mental construct of /a/, thus leading to a velar palatalization in front of all fronted and non-fronted instances of French /a/. We return to this below, as it is especially relevant for the purposes of the present dissertation.
To summarize the facts pertaining to the Romance languages, all Romance languages went through a palatalization process that triggered palatalization of any consonant before a glide /j/ (First Palatalization process). Also, most Romance languages experienced another palatalization process that affected only velar consonants before front vowels or glide (Second Palatalization process). Finally, French experienced another final palatalization process (French Third Velar Palatalization process) that 38 caused velars to also be palatalized before low vowel /a/. Relevant to the present dissertation is the fact that Majorcan Catalan has also been described to have a low vowel that is more fronted than in other Catalan varieties (Recasens, 2014). Therefore, it could be the case that the more fronted nature of such low vowel has triggered velar palatalization in certain varieties of Majorcan Catalan, as Buckley (2003) hypothesizes for Old French. This is discussed in the following section.
2.4. The case of Majorcan Catalan
The case of Majorcan Catalan is special in that a velar palatalization in front of
/a/, /ə/ and in word-final position has been attested both for the voiceless and for the voiced stops (Alcover, 1920; Barnils, 1933; Bernat i Baltrons, 2001; Bibiloni, 1983;
Moll, 1952; Moll, 1980; Recasens, 1991; Veny, 1978). As seen above, Catalan underwent the general palatalization processes affecting all Romance languages (First and Second palatalizations). Catalan, did not, however, undergo the Third Velar Palatalization process (entailing velar softening) that French underwent. The presence of a velar palatalization before /a/, /ə/ and in word-final position in Majorcan Catalan, however, suggests that Majorcan Catalan might be at a first or initial stage of what might have happened for French on the process towards velar softening. Both earlier impressionistic documentations of this process (Alcover, 1920; Barnils, 1933; Caimari Frau, 1983; Moll,
1960) and more recent palatographic and acoustic analyses (Recasens, 1991, 2011, 2013,
2014; Recasens & Espinosa, 2006, 2009, 2010) notice that this process observed in
j Majorcan Catalan does not result in a fronted post-palato velar stop consonant [k ]/[k+] or
j [g ]/[g+], but in a drastic change of place of articulation to an (alveo)palatal stop [c] or [ɟ]. 39
Whereas other dialects of Catalan, such as Eastern Catalan, contain a fronted post-palato
j velar stop allophone of /k/ and /g/ before front vowels and the palatal glide /j/ ([k ]/[k+],
j [g ]/[g+], respectively) (Recasens & Pallarès, 2001), Majorcan Catalan shows a much more fronted realization of velar stops in those same contexts, as well as before /a/
(Recasens, 2014; Recasens & Espinosa, 2006). As observed, therefore, Majorcan Catalan could be in a middle stage in the palatalization process that was observed in the French
Third Velar Palatalization process: /k/ > /kj/ > /c/ > /tʃ/ > /ʃ/. 9
Even if this feature has been reported for Majorcan Catalan, not all varieties of
Majorcan Catalan present the fronted allophone before front vowels, /a/, /ə/ and in word- final position. A geographical division of the island has been provided to account for the difference in the articulation of Catalan velar stop /k/ and /g/ (Bernat i Baltrons, 2001;
Bibiloni, 1983; Moll, 1980; Recasens, 1991; Veny, 1978). According to the traditional impressionistic descriptions, the palatalizing area (e.g. Palma, Valldemossa, Pollença,
Son Servera, Felanitx, Manacor, Calvià, Algaida, Santanyí) presents a palatalized articulation of this segment in front of /i/, /e/, /ɛ/, /a/ and /ə/ as well as in final position, and it presents a non-palatalized velar articulation of this segment, [k], in front of /o/, /ɔ/ and /u/. Notice, then, that the fronted allophone appears not only before front vowels /i/,
/e/, /ɛ/, as expected, but also appears in front of non-front vowels /a/ and /ə/ and in final position. This is not found in the non-palatalizing area (e.g. Bunyola, Esporles, Sa Pobla,
Alcúdia, Sineu, Artà, Santa Margalida, Sóller, Campos), where a fronted articulation of
9 Since the aim of the present dissertation is not to determine whether /k/ is produced as a fronted velar, [kj], or as a completely palatal consonant [c], but rather, whether it is produced as a fully velar or as a fronted version of /k/ (in either allophonic form, [kj] or [c]), for the purposes of the present dissertation, the term /k/-fronting or palatalized articulation will be used to refer to both possible realizations: /k/ > [kj]; /k/ > [c]. 40
/k/ only occurs before front vowels /i/, /e/ and /ɛ/. Note, however, that we lack information with respect to the size of the fronting gesture: Is it the case that palatalizing and non-palatalizing dialects manifest this process in comparable sizes in the common environments? That is, even if vowel contexts /i/, /e/ and /ɛ/ showed /k/-fronting due to coarticulatory effects in all dialects, it might be possible that each dialect presents a different degree of frontedness for this vowel context. A different size of the fronting gesture (i.e., one dialect fronting /k/ more than the other dialect for the same vowel context) would imply that for one of the dialects, the gesture has become exaggerated, interpreted as intended and eventually phonologized.
Note that the geographical division between towns that palatalize versus towns that do not palatalize does not conform a logical separation, but rather it is the result of the impressionistic observations provided in previous dialectological studies (Bernat i
Baltrons, 2001; Bibiloni, 1983; Moll, 1980; Recasens, 1991; Veny, 1978). This entails the absence of a clean geographical division of the island between palatalizing and non- palatalizing areas in reality. The present dissertation takes two villages as representatives of both the palatalizing and non-palatalizing area (one village per area).
The present dissertation analyzes the production of velar stop consonants in the speech of Majorcan Catalan speakers from the two traditionally established areas.
Generally, studies that have analyzed velar palatalization in Majorcan Catalan only provide acoustic and/or articulatory data about palatalizing speakers (Recasens &
Espinosa, 2006; Recasens, 2014). Additionally, it is yet to be seen whether the degree of palatalization of /k/ before front vowels is identical in the two areas. Therefore, to my knowledge, the actual difference between both areas regarding articulation of /k/ before 41 front and non-front vowels has yet to be formally analyzed. The present dissertation fills this gap by analyzing the speech of people born and raised in the town of Manacor
(belonging to the traditionally palatalizing area), and in the town of Artà (belonging to the non-palatalizing area). A comparison of the acoustic features of velar stop consonants between the two areas (Manacor vs. Artà) will be provided in order to see whether there is an actual difference in their production of velar stops.
Additionally, this dissertation will not only concentrate on the analysis of the speakers’ productions of velar stop consonants, but also on their production of the low vowel /a/. As seen in the previous section, it has been suggested that palatalization in front of /a/ happens in varieties where this low vowel has an especially fronted allophone
(Buckley, 2003). This might be the case of Majorcan Catalan, where the low vowel has been described to have a more fronted articulation than in other Catalan dialects
(Recasens, 1991, 2011, 2014; Recasens & Espinosa, 2006). As seen in Table 5, Majorcan
Catalan presents a much higher value for F2 (thus indicating a more fronted realization of
/a/) than Valencian, Eastern or Western Catalan varieties. Additionally, Majorcan Catalan presents a somewhat higher value for F1 than the Valencian dialect, and a much higher value than the other two varieties of Catalan, thus indicating a higher realization of /a/ (/a/ is especially closed) in Majorcan Catalan than in the other varieties.
Due to the apparently different nature of Majorcan Catalan /a/, it could be the case that, similar to what happened in French, a more fronted realization of the low vowel has triggered this change in certain varieties of Majorcan Catalan. The current dissertation provides an analysis of the front vowel /a/ as produced by people from the two geographical areas in two phonetic environments. 42
Table 5. F1 and F2 values for the low vowel /a/ in four different dialects of Catalan:
Majorcan Catalan, Valencian Catalan, Eastern Catalan and Western Catalan. F1 and F2 values are the average of vowel /a/ productions by 5 different speakers per dialect of
Catalan (Recasens, 2014; Recasens & Espinosa, 2006)
MAJORCAN VALENCIAN EASTERN WESTERN
F1 739 681 730 676
F2 1464 1419 1358 1415
Even though Recasens (2014) and Recasens and Espinosa (2006) provided F1 and
F2 data for the low vowel in Majorcan Catalan, the information provided is not enough to answer the research question addressed in this dissertation. Firstly, note that Recasens and Espinosa (2006), in their first experiment, do analyze /a/ in different consonant contexts, such as labial (/p, b, f/), dentoalveolar (/t, d, s/), palatal (/tʃ, dʒ, ʎ/), and alveolar contexts (/l, r/), but they do not provide data for /a/ in the specific context of /k/, which is crucial for this dissertation. This study reports a high degree of context-dependent variability for /a/, with an approximate F1 ranging from 710 Hz to 760 Hz and an approximate F2 ranging from 1750 Hz to 1350 Hz. Secondly, note too that the specific consonant contexts (i.e., the specific neighboring consonant sounds) are not specified clearly, can only be determined through the examples provided on a table, and are treated in blocks of sounds based on place of articulation. Also note that the results only provide the approximate values for /a/ in the different consonant contexts as displayed in a figure, rather than providing the exact mean for each consonant context. Lastly and importantly, there is no reference to the specific subdialect of Majorcan Catalan of the five 43 participants from Majorca. There is, therefore, no acoustic data from the two dialects of
Majorcan Catalan under study here. The fact that the two areas under study differ in whether /k/-palatalization applies before /a/ or not suggests the following question: Are there differences among the dialectal areas regarding the quality of /a/? If /a/ is produced similarly in both areas and /k/ is produced differently, this could indicate that /a/-fronting preceded /k/-palatalization. If, however, both /a/ and /k/ are produced differently in both areas, this would establish a relationship between the two processes, but it would not establish which process preceded which process. In this case, a third possibility regarding the relationship between the two processes would be considered: rather than one process triggering the other process or vice versa, it could be the case that both processes triggered each other simultaneously, leading to their phonologization. This dissertation, therefore, examines whether it is possible to establish the origin of the process of velar palatalization and the origin of the process of /a/-fronting, as well as the potential relationship between both processes.
2.5. Phonologization
The relationship between synchrony and diachrony is at the center of interest for phonologists who try to understand the intricacies of sound change. Specifically, understanding how phonology (i.e., phonological patterns) comes into being has received much attention. Several proposals have been made regarding the possible sources of phonological structure. Bybee (2001) proposes that phonological patterns arise from frequency distributions in the lexicon. Moulton (1967) and Garret and Blevins (2009) suggest that phonology can arise from morphophonemic analogical processes. 44
The most commonly accepted proposal regarding the establishment of phonological patterns, however, is the derivation of phonological structure from phonetic substance. For this, Hyman (1972, 1973) proposed the term phonologization.
Phonologization is understood as “the process or processes by which automatic phonetic patterns give rise to a language’s phonological patterns” (Garrett & Johnson, 2013)10.
Phonologization entails intrinsic phonetic variation that is the involuntary result of coarticulation and that can be predicted by universal phonetic principles becoming an extrinsic feature that is unpredictable. Phonologization, thus, entails an intrinsic feature becoming extrinsic. Hyman (1976) proposes three stages leading to the processes of phonologization and phonemicization. During Stage I, phonetic variation is intrinsic and unintended since it is due to coarticulatory forces. This means that this phonetic variation is produced as an obligatory and involuntary result of coarticulation instead of as a conscious decision to produce it in that way. During Stage II, this phonetic variation is exaggerated to a degree that cannot be attributed to universal phonetics only and it is used as a cue to identify a phonological element. This stage is known as phonologization.
Finally, Stage III entails that such variation becomes distinctive, phonemic. This stage is known as phonemicization. The current section concentrates on phonologization.
One common example to illustrate the process of phonologization is the well- known case of the English vowel length differentiation in stressed monosyllables (Bybee,
2001, pp. 43-44): vowels are longer before a voiced consonant (bed [bɛ:d], cab [kɑ:b]) than before a voiceless one (bet [bɛt], cap [kɑp]). According to Bybee (1977), vowel
10 This term is not to be confused with Jackobson’s (1931) concept of phonologisation, which makes reference to an already phonological property changing from allophonic to phonemic and which could be better understood as phonemicization. 45 length contrast is not usually recognized for English, since the vowel length difference is predictable from the voicing of the consonant. However, vowel length differences are used as perceptual cues to distinguish words ending in voiced or voiceless stops (Denes,
1955; Luce & Charles-Luce, 1985; Port & Dalby, 1982). Additionally, English vowel length differences have been shown to be larger than for many other languages, such as
French, Russian, Korean and Norwegian (Chen, 1970; Zimmerman & Sapon, 1958). As
Carrasco, Hualde, and Simonet (2012) put it, it seems that the universal phonetic force triggering the length difference in vowels depending on the context has been reanalyzed in English as a phonological difference. A phonetic pattern has been phonologized. In relation to the three stages proposed by Hyman (1976), a step further can be observed in which durational differences are then phonemicized by final devoicing, that is, the contrast is simply marked by vowel length (bed [bɛ:t] vs. bet [bɛt]) (Hyman, 2008). The final stage, therefore, can include the loss of the difference trigger.
The Basque language and its multiple dialects provide another clear case of phonologization of strategies for hiatus resolution (Erdozia, 2001; Hualde, 1990). In different dialects of Basque, /ɾ/ and /g/ (produced as [ɣ]) in intervocalic position have been historically deleted, thus creating sequences of two strong vowels that form a hiatus.
When the hiatus entails a mid vowel as the first constituent of the vowel sequence, this hiatus has been historically resolved through different strategies in the different dialects of Basque. For instance, in the Uharte dialect, in which this consonant deletion is still optional, vocalic sequences containing a hiatus (such as [ea] or [eo]) are maintained as such. In two other dialects, Etxarri and Arbizu, consonant deletion is mandatory. The result is each dialect using a different hiatus resolution strategy, as observed in Table 6. 46
Table 6. Different hiatus resolution strategies applied to different Basque dialects
(Hualde, Simonet, & Torreira, 2008; as taken from Erdozia, 2001; Hualde, 1990)
UHARTE ARBIZU ETXARRI TRANS.
eran ean aan jan ‘drink’
egon eon oon jon ‘stay’
egosi eosi oosi josi ‘boil’
In Arbizu, the solution is to assimilate the first mid vowel of the vocalic sequence into the second vowel of the sequence, thus giving place to sequences such as [ea] > [aa] or [eo] > [oo]. Differently, in Etxarri, the solution is to glide the first mid vowel of the vowel sequence, as in [ea] > [ja] or [eo] > [jo]. It seems that after deletion of intervocalic
/ɾ/ and /g/, different strategies for hiatus resolution surfaced and that sociolinguistic factors determined that each dialect solved hiatus through different strategies.
Another example is the process of vowel nasalization, which has been phonologized in certain languages, as explained in Carignan, Shosted, Shih, and Rong
(2011). Certain Romance languages, such as Portuguese or French, started with Latin vowel-nasal (VN) sequences that were eventually phonologized as nasal vowels (Ṽ). In such cases, the nasalization was exaggerated to a point that the nasal consonant of the VN sequence would disappear (the triggering element disappears) and the vowel quality changed (Sampson, 1999, p. 33): Latin /an/ became /ã/ in Modern Portuguese, and Latin
/in/ became /ɛ/̃ in modern French. Another language that also presents vowel nasalization is English (Bybee, 2001). Before a voiceless consonant, the phonetic form of a VN is a nasalized vowel with very little or no nasal consonant at all (can’t [kæ̃t], camp [kæ̃p]); before a voiced consonant, the phonetic form of a VN is a nasalized vowel followed by a 47 nasal consonant (hand [hæ̃nd]). It seems that English speakers treat the difference between Ṽ and VN as a contrastive difference. Bybee (2001) reports a study that shows that the presence or absence of the nasal consonant in words such as ample or amble is an important cue to correct word identification: when the nasal resonant was removed from a word such as amble, listeners were more likely to understand it as ample (Malécot,
1960). English speakers seem to have phonologized vowel nasalization.
Mielke (2004), in his approach to phonologization, makes a distinction between changes that affect multiple segments from the very beginning and changes that affect one segment at its origin and are then widespread to other segments. The first type of change would occur when a phonetic effect is spread before it becomes phonologized. In the case of vowel nasalization, all could have been affected at the same time if they are all nasalized and such nasalization is reinterpreted as contrastive. This would affect all vowels since they were all nasalized before nasalization became phonologized. Similarly, final devoicing in Russian, German or Turkish and postnasal voicing in Greek or many
Bantu languages provide examples of sound changes that affected multiple segments at the same time. In the first case, voiced obstruents or voiced consonants are devoiced already by the time phonetic voicing is reinterpreted as a phonological distinction. In the second case, voiceless obstruents are already voiced by the time it is phonologized.
The second type of change that Mielke (2004) mentions is the one that starts with a phonetic tendency that is initially phonologized only for one segment, and is thereafter spread analogically to other segments. For instance, in the case of vowel nasalization, lower vowels tend to be nasalized more deeply than higher vowels due to articulatory forces, i.e., tongue lowering facilitates velum lowering due to their connection via the 48 palatoglossus muscle (Johnson, 1997; Moll, 1962; Lubker, 1968, as cited in Mielke,
2004). Specifically, Chen (1973) explains that vowel nasalization in Old French is said to have started with /a/ around the 11th century and spread vowel by vowel in the 14th c until affecting all vowels /a, e, aj, ej, o, oj, i, u/ (as cited in Mielke, 2004). It is then easy for nasalization to start with one vowel and extend to the rest of vowels if it is the case that all vowels share the phonetic/acoustic feature (nasalization) even to a lesser degree. This last point is key: these changes begin at one segment and extend to others because others already present the feature that has been phonologized at least to a lesser degree, but robustly.
Another example of phonologization is /o/ lowering (/o/ > /ɔ/) in northeastern varieties of Swiss German. This process originally took place only before /r/ due to articulatory reasons, i.e., pre-rhotic position favors lowering the mid vowel. This is observed in examples in Table 7.
Table 7. /o/-lowering only before /r/ (left), no /o/-lowering before other contexts (right)11
/o/-LOWERING NO /o/-LOWERING
[bɔrə] ‘to bore’ [losə] ‘to listen’
[fɔrə] ‘fir tree’ [ross] ‘horse’
[hərn] ‘horn’ [xrottə] ‘toad’
This /o/ lowering started as allophonic variation resulting from articulatory forces, but was later reinterpreted as an intended gesture and then phonologized. This change started only for the /r/ context in the city of Schaffhausen, but it became spread to nasal
11 These examples have been extracted from Mielke (2004). 49 contexts in this dialect and to many other consonantal contexts in the varieties of 17 nearby villages. This entailed the spread of the rule (/o/ was lowered) not necessarily due to articulatory forces, but due to generalization. This rule generalization took place in various patterns depending on the dialect area (Keel, 1982; Janda & Joseph, 2001;
Mielke, 2004), as observed in (1) below.
(1) Canton of Schaffhausen generalizations: /o/ > /ɔ/
a. Schaffhausen: before /r/ and nasals (but not obstruents)
b. 13 other villages: before /r/, nasals, and coronal obstruents
c. 17 other villages: before /r/ and coronal obstruents
d. 5 other villages: before /r/, coronal obstruents, non-coronal
obstruents except /b/
Table 8. /o/-lowering before coronal obstruents in all neighboring villages to
Schaffhausen
CORONAL OBSTRUENTS
[lɔsə] ‘to listen’
[rɔss] ‘horse’
[xrɔttə] ‘toad’
Therefore, what started as a phonetically motivated process that affected /o/ in only one consonantal context (= / _r) was phonologized for that context and was then extended to other contexts. The generalization of this rule did not involve environments that would be expected to favor /o/-lowering due to articulatory-acoustic reasons. As 50 explained in Janda and Joseph (2003), it is only the allophonic beginnings of sound changes that are usually phonetically or acoustically motivated. These tendencies, inherent to the human vocal/auditory system, occur naturally due to perception or production forces, but they require that a (psycho)physical feature is exaggerated in some way so that it stands out and is reinterpreted as intended. As Janda (2000) suggests, sound changes do not just happen because of articulatory gestures or because of auditory or acoustic phonetic features, but rather they occur because one feature is exaggerated or is misperceived and is reinterpreted as the intended gesture or result.
A further example is the one observed in several tonal languages. Normally, voiced consonants behave as depressors, lowering the tone from H to L of vowels adjacent to such consonants. This phonetic effect is especially true for voiced obstruents, which are said to lower the F0 of the following vowel (Bradshaw, 1999; Hyman &
Schuh, 1974; Mielke, 2004, Odden, 2005). According to Hyman and Schuh (1974), there is a hierarchy of phonetic F0 lowering: breathy voice (more lowering) > voiced obstruents > sonorants > voiceless unaspirated > voiceless aspirated > implosives (less lowering). As Odden (2005) explains, this lowering effect is explained through several phonetic causes: (1) F0 is correlated with transglottal airflow (with higher airflow rates at voiceless consonant release than at voiced obstruent release), (2) F0 is correlated with larynx height (voiceless obstruents have a higher larynx position than voiceless obstruents), and (3) higher cricothyroid activity correlates with higher pitch and voicelessness. In any case, this effect can be observed, for instance, in syllables with an obstruent onset differing in voice start with a high tone (/pá/, /bá/). A voiced obstruent before a vowel has an intrinsic effect on F0, that is, F0 is automatically lowered due to 51 coarticulatory forces. This gives rise to a rising tone after voiced obstruent ([bǎ]). This is then exaggerated and it starts to be used as a cue to identify voice. A phonetic feature, then, is phonologized. The final stage of the process proposed by Hyman (1976) shows a devoicing of such obstruent ([pǎ]). The rising tone has been phonemicized (/pá/, /pǎ/).
In some languages, such as Nupe, Ngizim, Ewe or Kanazawa Japanese (Odden,
2002, 2005; Mielke, 2004) F0 lowering is observed in vowels adjacent to sonorant consonants. In these languages, it seems that what initially was a purely phonetic effect – the F0 lowering– has been reanalyzed as a phonological tone. Similarly, in Zina Kotoko, an underlying mid tone in the first syllable is phonologically lowered to low after voiced obstruents (Xa), sonorants (Xb) and glottal stops (Xc), which are contexts that condition
F0-lowering, as seen in the previous hierarchy.
Table 9. Tone lowering after a voiced obstruent (left), sonorant (center) glottal stop
(right)12
VOICED OBSTRUENT SONORANT GLOTTAL STOP dùnk-ə́m ‘throw’ jèj-ə́m ‘call’ ʔə̀kf-ə́m ‘approach’ vìt-ə́m ‘blow a fire’ làb-ə́m ‘tell’
This language, however, takes one step even further, similar to what occurred with Swiss German, since this phonologization has been generalized to other contexts that should not present this phonetic effect –but the opposite one–, such as after implosives. See Table 10 below:
12 Examples are from Odden (2002) and Mielke (2004). 52
Table 10. Tone lowering after an implosive
IMPLOSIVE
ɗə̀v-ə́m ‘put’
ɓàl-ə́m ‘dance’
Mielke (2004) suggests that the F0-lowering triggered by voiced obstruents was probably the phonetic beginning for these phonological patterns, as happens with some other consonant-tone interactions. Speakers, however, generalized this feature along the phonetic feature of voicing (instead of along the phonetic feature of F0-lowering) to include other sonorant and implosive segments. Therefore, again, what started as a merely phonetic consequence of some segment production was phonologized and then extended to other contexts.
These processes are just a few out of many other processes that entail phonologization in certain languages. Hyman (2008) provides a list of processes that may be phonetic in one language, but that have been phonologized in other languages. See
Table 11 in the next page.
As noticed in Table 11, the process relevant to this dissertation, velar palatalization, is also included on the list of processes that show phonologization in certain languages. As seen in the previous section, velar palatalization usually starts as a coarticulatory force that brings the point of articulation of the velar stop consonant forward in the mouth before front vowels. It is claimed that this process, therefore, starts as a universal coarticulatory effect (gestural blending) and is later exaggerated when it is reinterpreted as an intended feature. It is, at this point, an extrinsic and unexpected feature that became language- or dialect-specific. 53
Table 11. Processes that show phonologization, with their subsequent developments and loss of trigger
PROCESS DEVELOPMENTS (& LOSS OF TRIGGER) a. lengthening before voiced Cs /ab/ à [a:b] (> a:p) b. palatalization /ki/ à [kji ] (> tʃi, ʃi, tsi, si) c. high vowel frication /ku/ à[khu] (> kxu, kfu, pfu, fu) d. anticipatory nasalization /an/ à [ãn] (> ãN, ã:, ã) e. umlaut, metaphony /aCi/ à [æCi] (> ɛCi, ɛCə, ɛC) f. tonogenesis from coda /aʔ/ à [áʔ] (> á) g. tonogenesis from phonation /a̰ ʔ/ à [à̰ ʔ] (> à) h. tonal bifurcation from onset /bá/ à [bǎ] (> pǎ)
More specific to the present dissertation, it is expected that both palatalizing and non-palatalizing areas show evidence of /k/-fronting before front vowels. It is hypothesized, however, that the palatalizing area produces a much larger difference between /k/+/i/ and /k/+/o/ than the non-palatalizing area does. This means that /k/ is exaggeratedly more fronted before /i/ in the palatalizing area than it is in the non- palatalizing area. Whereas the /k/-fronting in the non-palatalizing area would be interpreted as the result of coarticulatory forces, the exaggerated /k/-fronting in the palatalizing area would be interpreted as a sign that what initially started as a coarticulatory gesture has been reinterpreted as an intended feature for this community of speakers. This difference between communities (especially with the exaggeration of /k/- fronting in the palatalizing area) would be evidence that this phonetic feature has been phonologized in the palatalizing area of Majorca. 54
2.6. Consequences of consecutive bilingualism
The first year of age is critical for the acquisition of native-like sounds (Bosch &
Sebastián-Gallés, 2003; Dupoux et al., 1997; Kuhl, 1994; Kuhl et al., 1992, 2006; Maye
Weiss, & Ashlin, 2008; Otake et al., 1993; Werker, 1994; Werker & Tees, 1984). Even if infants are able to discriminate native and non-native speech contrasts during the first months of their lives, their perceptual abilities become language-specific during the first year of life, that is, by the time they reach age one, they can only discriminate the phonetic inventory of their L1. The relevance of age of acquisition has proved to be critical both for foreign language acquisition and for consecutive bilingual acquisition. In the case of foreign language acquisition, it has been suggested that there is a direct relationship between age of acquisition of a language and non-native/foreign accent in the
L2 (Asher & García, 1969; Fathman, 1975; Seliger, Krashen, & Ladefoged, 1975; Suter,
1976; Oyama, 1976; Tahta, Wood, & Lowenthal, 1981; Piper & Cansin, 1988; Flege,
1988; Patkowski, 1990; Thompson, 1991; Flege & Fletcher, 1992; Flege et al., 1995;
Flege et al., 1999; Moyer, 1999).
Similarly, in the case of consecutive bilingual acquisition, it has been suggested that the longer the period between acquisition of an L1 and acquisition of an L2, the larger the consequences for the native-like acquisition of linguistic patterns of the L2.
This has been suggested both for perception (Bosch et al., 2000; Flege & MacKay, 2004;
Goto, 1971; Hojen & Flege, 2006; Pallier et al., 1997; Sebastián-Gallés & Bosch, 2005;
Sebastián-Gallés & Soto-Faraco, 1999; Werker & Tees, 1984) and for production (Flege,
Munro, & MacKay, 1995; Flege et al., 1999; Fowler, Sramko, Ostry, Rowland, & Hallé,
2008; Long, 1990; Oyama, 1976, Piske, Flege, MacKay, & Watt, 2002; Simonet, 2010, 55
2011a). The consequences of consecutive acquisition can entail a higher difficulty to produce native-like targets, many times leading to phonetic transfer from the L1 to the L2
(Flege, 1991; Long, 1990). As explained in Amengual (2013: 9), “This transfer may result in L2 phonology reflecting L1 categories, articulation gestures particular to L1 to the exclusion of those normally required in L2, and the application of L1-specific phonological processes”. According to these studies, even when acquisition of an L2 takes place during childhood (early consecutive bilingualism), the bilingual will face difficulties to acquire L2 phonological patterns as if it was their L1. Sequential acquisition of languages makes L1 phonological patterns to act as a filter that interferes with L2 acquisition (Flege, Schirru, & MacKay, 2003). Some studies suggest that these difficulties can be overcome and that bilinguals can acquire near-native (and notice the use of the term near-native instead of native) patterns (Amengual, 2011, 2016; Best &
Strange, 1992; Flege, 1995, 2007; Herrick, 2006). Other studies, however, have shown that, even with an extensive and early exposure to the L2, acquiring two languages consecutively entails unsurmountable consequences for the acquisition of L2 phonology
(Guion, Flege, & Loftin, 2000; Fowler et al., 2008; Mack, 1989; Pallier et al., 1997; Piske et al., 2002; Sebastián-Gallés & Soto-Faraco, 1999; Simonet, 2010, 2011a). In any case, even in those studies that have shown that an approximation to the patterns of the L2 can be reached, these bilinguals still present differences in the phonetic implementation of L2 sounds when compared to the native speakers of such L2.
In order to understand the intricacies of bilingual production patterns, multiple studies have analyzed the behavior of early Catalan-Spanish bilinguals. For instance, an experimental study by Simonet (2011b) analyzes sentence-final F0 contour in read aloud 56 declarative sentences in Spanish and Catalan as produced by Majorcan bilinguals that are dominant in Spanish or dominant in Catalan. The main results suggest that Catalan productions of young Spanish-dominant females show similar contours to Catalan- dominant ones, which suggests that intonation at the end of declarative sentences is in the process of changing towards a convergence between both languages, that is, Spanish intonation is moving towards Catalan intonation. Other results show evidence of transfer of L1 patterns on the L2, such as the case of young Spanish-dominant males who produce a Spanish-like intonation in their Catalan, or the case of Catalan-dominant females who produce Spanish with a Catalan-like intonation. These results altogether suggest that even after an early and extensive exposure to the L2, bilingual speakers do not produce L2 sounds in a native-like manner. Finally, many bilinguals maintained different distributions for intonation in each language, which suggests that they have developed a new phonetic category for their L2 intonation.
The effects of contact on these two languages are also evaluated in Simonet
(2010). This experimental study analyzes read aloud productions of alveolar laterals in both languages and by the two types of bilinguals to understand whether sequential bilinguals develop separate phonetic categories for L2 sounds that are similar to sounds in the L1. First, it is shown that both groups of bilinguals differ in their productions of laterals in both languages, which, again, suggests that an early exposure to the L2 does not guarantee the development of phonetic patterns in a native-like way. Additionally, laterals produced in the L2 are different from laterals in their L1, as well as from laterals produced by dominants in that L2, thus showing evidence for the formation of new phonetic categories. Finally, phonetic category assimilation is also observed in that L2 57 laterals are more similar to their L1 laterals than to the laterals produced by dominant speakers of that L2.
Ramírez and Simonet (accepted) analyzed the production patterns of Majorcan
Catalan-dominant and Spanish-dominant bilinguals with respect to the Catalan voiced alveopalatal fricative consonant, /ʒ/. Impressionistic observations had suggested that some Majorcan Catalan speakers produce /ʒ/ as [ʝ] instead of [ʒ]. Participants were asked to repeat a series of stimuli in a delayed repetition task in which the stimulus was sometimes produced with the ‘traditional’ variant, [ʒ], and some other times with the
‘innovative’ variant, [ʝ]. The results of the task showed systematic differences between the two groups of bilinguals that were due to the linguistic profile of the subjects: whereas Catalan-dominant bilinguals produced the sound as [ʒ], Spanish-dominant bilinguals’ productions were acoustically different from [ʒ]. Importantly, no effects of imitation of the auditory stimulus were observed, which indicated that the two groups were behaving differently in their production of /ʒ/ due to their linguistic profile rather than due to stimulus automatic imitation.
Other studies have analyzed the production of Catalan vowel contrasts. For instance, production of Catalan contrasts /e/-/ɛ/ and /o/-/ɔ/ has been analyzed for different areas of Catalan-Spanish contact (Amengual, 2011, 2016; Herrick, 2006, Lleó et al.,
2009; Simonet, 2011a; Recasens, 1991). Lleó et al. (2009) analyze, via a picture naming and a question answering task, the production of mid vowels and /ə/ in bilinguals from
Barcelona divided by age and place of residency (Spanish-speaking city district vs.
Catalan-speaking city district). The results show a loss of contrast between mid vowels and a production of /ə/ as [a], especially in the groups of speakers who are more exposed 58 to contact with Spanish since childhood, that is, the two youngest groups of participants in the Spanish-speaking neighborhood. Differently, the oldest participants show a higher retention of the vocalic patterns of Catalan.
Contrast between Catalan middle vowels has also been studied in Majorca
(Amengual, 2011, 2016; Simonet, 2011a, 2014). In Simonet (2011a), Catalan- and
Spanish-dominant speakers produced read aloud back middle vowels in Spanish and
Catalan. The productions of the middle vowel in Spanish were the same for both groups of bilinguals. The productions of the Catalan vowels showed some differences. Firstly,
Spanish-dominant speakers did not produce the vocalic contrast, but produced a single middle back vowel, which was a mixture of the two vowels of Catalan, showing again that an early and extensive exposure to the L2 does not guarantee the acquisition of new patterns in the L2. Additionally, this middle vowel of Spanish-dominant speakers in
Catalan was different from their mid back vowel in Spanish and different too from the two mid vowels of Catalan-dominant speakers in Catalan, showing a case of dissimilation of phonetic categories between Spanish /o/ and the mixture between Catalan /o/ and /ɔ/ of these bilinguals. These results contrast with the ones found in Simonet (2014), in which
Spanish-dominant speakers, even if they were showing interference with Spanish, produced the Catalan vowel contrast. It could be the case that the participants in the first study were more dominant in Spanish than Spanish-dominant in the second study, or it could be the case that the difference in the type of activity (reading vs. repetition) has an effect in the production of this Catalan contrast.
The results in Simonet (2014) are in consonance with the ones presented in
Amengual (2011, 2016). Amengual (2011, 2016) also analyzed these two Catalan vowel 59 contrasts as produced by bilinguals divided by their profile of linguistic dominance. In this study, neither Catalan- nor Spanish-dominant bilinguals merged the Catalan phonemes /ɛ/ and /ɔ/ with /e/ and /o/, respectively, when speaking Majorcan Catalan. This indicated that Spanish-dominant bilinguals had been able to maintain a contrast between the two phonemes in their non-dominant language (even though the acoustic distance between the members of each vowel pair was smaller for the Spanish-dominant speakers than for the Catalan-dominant speakers).
By looking at all these studies, there seems to be robust evidence for L1-to-L2 transfer in this bilingual speech community. As reported, some studies (Lleó et al., 2009;
Ramírez & Simonet, accepted; Simonet, 2010, 2011a, 2011b) suggest that the presence of crosslinguistic transfer is an inevitable consequence of acquiring a second language after the native language, even when such acquisition has taken place during childhood.
Contrastingly, some other studies (Amengual, 2011, 2016; Simonet, 2014) have demonstrated that it is possible to overcome such difficulties to the point of developing non-native phonemic contrasts. However, note that even these latter studies show effects of transfer: Spanish-dominant bilinguals do learn the phonological categories, but these categories are not phonetically identical to those of the Catalan-dominant ones.
Therefore, transfer from the L1 to the L2 (at least minor transfer effects, i.e., transfer of features) seems to be a robust finding of the Catalan-Spanish bilingual community. This phenomenon has almost always been analyzed from the perspective of Catalan (Spanish- to-Catalan transfer by Spanish-dominant subjects). The present dissertation explores the possibility of Catalan-to-Spanish transfer in Catalan-dominant subjects. Specifically, it analyzes the speech of Catalan-dominant speakers of Majorcan Catalan regarding their 60 productions of velar stop consonants both in Catalan and in Spanish. The potential effects of linguistic transfer when producing Spanish speech will be analyzed. The following section summarizes this dissertation and reviews the research questions and hypotheses.
2.7. The present dissertation
The aim of this dissertation is three-fold. First, it seeks to provide acoustic data for the distribution of /k/13 in the Catalan spoken in the two established areas of the island. By analyzing the productions of people belonging to the two different areas, it will be possible to see whether the process of velar palatalization has been phonologized in the palatalizing area (speakers from the town of Manacor) compared to the non- palatalizing area (speakers from the town of Artà). Notice that, to our knowledge, the division of these two areas has only been reported through impressionistic observations
(Bernat i Baltrons, 2001; Bibiloni, 1983; Moll, 1980; Recasens, 1991; Veny, 1978).
Those studies that have analyzed the phenomenon of velar palatalization in Majorca
(Recasens, 2014) only provide data for the palatalizing area, which makes it impossible to assess whether palatalization is indeed a phonologized feature of these dialects. The traditionally noticed difference between the two areas has yet to be analyzed acoustically.
Therefore, the first aim of this dissertation is to analyze whether there is in fact a distinction between the two areas, and to assess the size of this difference (if attested).
This entails analyzing what the distribution of palatalization is, not only geographically,
13 Notice that the present dissertation will only analyze the production of the voiceless velar stop /k/, and will not analyze the production of the voiced velar stop /g/. The reasons for this are merely practical: it is easier to segment a voiceless velar than a voiced velar due to the voicing period before the burst of the consonant in voiced stops and due to spirantization in many prosodic positions. 61 but also within each subdialect in terms of the phonetic contexts in which palatalization occurs.
Apart from analyzing the realizations of the velar stop consonant /k/, the present dissertation seeks to analyze the low vowel in the Catalan spoken in the two areas. If /a/ is equal for both areas but there is a palatalization difference, it could suggest that /a/ has triggered palatalization for the palatalizing area, as it has been suggested for a similar process in Old French (Buckley, 2003). If /a/ and /k/ are more fronted in the palatalizing area than in the non-palatalizing area, this would suggest a relationship between both processes, but the order of influence would not be determined. Again, to our knowledge, there is no study that compares the phonetic properties of vowel /a/ between both areas.
Note here that we do have acoustic data for Majorcan Catalan /a/, and that it indeed suggests that this vowel is more fronted in this dialect than in other dialects of Catalan.
However, we do not know whether there are differences between palatalizing and non- palatalizing Majorcan Catalan dialects, and whether /a/ is more fronted before velars than before other consonants.
The final aim of this dissertation is related to the bilingual nature of the island of
Majorca, in which both Spanish and Catalan are spoken. In situations of language contact, it is not unusual for crosslinguistic transfers to occur (Thomason & Kaufman,
1988; Van Coetsem, 1988; Winford, 2005). Many times, bilinguals transfer the phonological/phonetic patterns of one of their languages onto their other language.
Studies about these bilinguals have shown that in spite of having an early and extensive exposure to an L2, performance of bilinguals in their non-dominant language differs from that of bilinguals with that language as their dominant language (Simonet, 2010, 2011). 62
The present dissertation seeks to analyze the productions of bilinguals that are Catalan- dominant to find possible patterns of crosslinguistic transfer from Catalan, their dominant language, onto Spanish, their non-dominant language.
To sum up, the present dissertation seeks to (1) analyze whether there is, in fact, a palatalizing distinction in the Catalan of the two traditionally described areas of Majorca
(comparison between Manacor and Artà), and to assess its size, (2) analyze whether the production of the low vowel /a/ is fronted (and higher) for both areas or not (and as a function of its preceding consonant or not), and (3) investigate patterns of crosslinguistic transfer in non-dominant speech. All of the phonetic studies reported in this dissertation use speech production data.
63
CHAPTER 3
METHOD
3.1. Participants
A total of 40 adult Catalan-Spanish bilingual speakers born and raised on the island of Majorca were recruited for the present study. All 40 participants participated in two production tasks (one in Catalan and one in Spanish). Participant ages ranged between 18 and 55. They were divided by gender (21 male, 19 female) and by group (20 from Manacor, 20 from Artà). Due to historical and political reasons, it is not possible to find monolingual speakers of Catalan. Selected participants, therefore, were Catalan- dominant bilinguals. Only Catalan-dominant bilinguals (and not Spanish-dominant bilinguals) were selected in order to better portray the distribution of a phonological feature of the Catalan language (i.e., velar palatalization). Additionally, by only selecting this group of bilinguals, the effects of a non-dominant language on such Catalan feature are minimized. In fact, the only potential crosslinguistic influences that are of interest for the present study are those from Catalan to Spanish, i.e do Catalan-dominant bilinguals transfer their phonological features from Catalan (specifically, velar palatalization) onto their production in Spanish or do they maintain two separate distributions, one per language? In order to answer research questions for the present study, only Catalan- dominant bilinguals will be analyzed.
In order to determine the patterns of bilingual language dominance of all these participants and in order to ensure that only Catalan-dominant participants were selected, all participants were asked to complete a Bilingual Language Profile or BLP. This is a 64 linguistic questionnaire developed by Birdsong, Gertken and Amengual (2012), and it has been used to determine language dominance patterns in prior work conducted in the
Catalan-Spanish contact situation (Amengual, 2016, 2015; Simonet, 2014; among many others). Numerical self-reported answers to questions on four different modules
(language history, use, proficiency and attitudes) render four different scores, one per module. The sum of these scores from the four modules is an indication of language orientation. Each participant receives one index per language (Catalan, Spanish). The subtraction of these two indexes renders a dominance score per participant, which ranges from –218 to 218. Scores near –218 suggest dominance in one of the two languages
(Spanish, for the purposes of the present study); scores near 0 suggest balanced bilingualism; and scores near 218 suggest dominance in the other language (Catalan). All participants in this study received positive scores (Manacor: M = 86.70, SD = 21.78;
Artà: M = 71.69, SD = 27.06), indicating that they were Catalan-dominant bilinguals. For the participants’ social and linguistic information, see Tables 12 (Manacor) and 13 (Artà).
The linguistic variable that is relevant for this study is language dominance. All participants learned both languages during childhood. All participants were exposed to their dominant language (Catalan) at home at age 0 and were exposed to their non- dominant language (Spanish) when they entered the schooling system (at 6 years of age at the latest). All participants, therefore, came from Catalan-speaking households exclusively and were exposed to Spanish elsewhere. Their self-reports indicate more positive attitudes towards Catalan than Spanish, a higher past use of Catalan than
Spanish, a higher current use of Catalan than of Spanish, and a high proficiency in both languages. 65
Table 12. Social (gender “Gd”, age “Age”) and linguistic information (History, Use,
Proficiency, Attitudes and dominance level “Score”) for Catalan “Cat” and Spanish
“Spn” per participant from Manacor. Averages and standard deviations for BLP scores for each module, language and dominance level are provided at the bottom
GD AGE HISTORY USE PROFIC. ATTITUDES SCORE CAT SPN CAT SPN CAT SPN CAT SPN F 54 54.4 41.3 49.5 2.7 54.4 45.4 54.4 47.6 75.9 F 55 45.4 39.9 53.4 1.0 54.4 54.4 40.0 0 98.6 F 53 44.4 35.4 52.3 2.1 52.2 47.6 54.4 31.7 86.3 F 20 44.4 24.0 52.3 2.1 54.4 45.5 54.4 22.7 111.4 F 22 49.9 24.0 52.3 2.1 54.4 54.5 54.4 29.5 100.9 F 37 49.9 35.4 52.3 2.1 43.1 52.2 43.1 43.1 55.5 F 37 43.1 48.5 54.5 0 54.4 52.2 54.4 22.7 83.1 F 40 49.9 29.0 43.6 10. 54.4 49.9 54.4 18.1 95.3 F 42 50.8 29.9 47.9 6.5 54.4 45.4 43.1 18.1 96.3 F 24 46.7 23.1 54.5 0 54.4 47.6 54.4 4.5 134.8 M 34 45.8 32.2 54.5 0 54.5 54.5 43.1 29.5 81.7 M 55 45.4 40.8 41.4 13.0 45.4 47.6 54.4 36.6 48.7 M 29 47.6 39.0 51.2 3.2 52.2 49.9 54.4 31.7 81.5 M 34 48.1 42.2 45.7 8.7 52.2 54.4 47.6 29.5 58.8 M 44 53.1 38.5 49.5 4.9 54.4 52.2 54.4 13.6 102.3 M 41 49.0 46.3 53.4 1.0 52.2 47.6 54.4 34.0 80.0 M 42 51.3 36.7 52.8 0 54.4 52.2 54.4 24.9 99.1 M 41 49.9 30.8 54.5 0 54.4 47.6 54.4 15.8 118.9 M 51 49.9 46.7 48.5 5.9 52.2 47.6 43.1 34.0 59.3 M 53 46.7 35.4 51.7 2.7 49.9 47.6 43.1 18.1 87.6 M 40.4 48.3 36 50.8 3.4 52.6 49.8 50.9 25.3 87.8 sd (11.0) (3.0) (7.5) (13.9) (14.3) (3.1) (3.2) (5.4) (11.9) (21.7)
66
Table 13. Social (gender “Gd”, age “Age”) and linguistic information (History, Use,
Proficiency, Attitudes and dominance level “Score”) for Catalan “Cat” and Spanish
“Spn” per participant from Artà. Averages and standard deviations for BLP scores for each module, language and dominance level are provided at the bottom
GD AGE HISTORY USE PROFIC. ATTITUDES SCORE CAT SPN CAT SPN CAT SPN CAT SPN F 38 54.6 31.7 20.4 2.1 54.4 47.6 40.8 20.4 66.3 F 40 48.1 34.5 51.2 3.2 54.4 49.9 54.4 34.0 86.5 F 42 48.1 36.7 38.1 16.3 52.2 43.1 54.4 49.9 46.7 F 42 48.1 36.7 49.0 5.4 52.2 45.4 54.4 49.9 66.3 F 42 45.8 34.5 51.6 1.5 54.4 47.6 54.4 27.2 95.5 F 25 45.8 34.5 48.5 5.9 54.4 49.9 54.4 22.4 102.2 F 20 42.6 24.0 45.7 4.3 52.2 52.2 54.4 47.6 66.8 F 39 49.4 34.9 53.4 1.0 54.4 52.2 54.4 29.5 94.0 F 55 43.1 35.4 51.7 2.7 47.6 49.9 54.4 49.9 59.0 M 18 41.3 31.7 51.2 3.2 54.4 49.9 52.2 43.1 71.1 M 21 44.4 27.2 44.6 9.2 54.4 49.9 52.2 45.4 63.9 M 54 45.8 54.4 54.5 0 54.4 49.9 54.4 20.4 84.4 M 19 42.2 24.5 54.5 0 54.4 43.1 54.4 34.0 103.9 M 19 42.2 24.9 54.5 0 54.4 47.6 54.4 36.3 96.7 M 30 47.2 32.6 48.5 0.5 54.4 52.2 45.4 54.4 55.6 M 20 44.0 23.1 51.2 3.2 52.2 45.4 54.4 20.4 9.7 M 25 44.4 29.5 47.1 3.5 54.4 54.4 54.4 54.4 58.5 M 43 52.6 30.8 54.5 0 54.4 54.4 54.4 20.4 110.3 M 28 49.0 25.4 49.5 4.9 52.2 52.2 52.2 40.8 79.6 M 55 45.4 43.1 17.3 3.7 54.4 54.4 54.4 54.4 15.8 M 33.7 46.2 32.5 46.8 3.5 53.5 49.5 53 37.7 71.6 sd (12.7) (3.5) (7.3) (10.3) (3.8) (1.7) (3.4) (3.5) (12.8) (27.0)
67
3.2. Materials
All participants participated in two production experiments. The first production experiment entails reading materials in Catalan. The second production experiment entails reading materials in Spanish. All participants, therefore, read materials in Catalan and Spanish.
3.2.1. Catalan materials
Speech material for the production experiment in Catalan consists of a total of 60
Catalan words. In Catalan, phoneme /k/ corresponds to graphemes “c”; “q” and “k” in the following contexts:
(2) Phoneme /k/ to grapheme correspondences
/k/ - “c”
“c” + “a”: casa ‘house’
“c” + “o”: copa ‘glass’
“c” + “u”: cuna ‘cradle’
/k/ - “c”
“c” + “l”: clar ‘clear’
“c” + “r”: creu ‘cross’
/k/ - “q”
“qu” + “e/i”: què ‘what’, qui ‘who’
“q” + “ua/uo/üe/üi”: quan ‘when’, quota ‘quota’, qüestió ‘matter’,
aqüífer ‘aquifer’
“k”: karate ‘karate’ 68
In order to elicit the production of a word-initial voiceless velar stop /k/, the stimuli used for this experiment contain a word-initial grapheme “c”, “qu” or “k”. These stimuli are divided into four different conditions (15 stimuli per condition) depending on the vocalic sound that follows the word-initial consonant: /a/, /ə/, /o/, /i/. One front vowel
(/i/, e.g., quina /'ki.nə/ ‘which one’), one back vowel (/o/, e.g., copa /'ko.pə/ ‘glass’), the low vowel (/a/, e.g., cada /'ka.ðə/ ‘each’) and the neutral vowel (/ə/, e.g., camí /kə.'mi/
‘way’) were selected for the present study. These conditions allow for a study of /k/ in four vocalic contexts that, according to previous literature (Alcover, 1920; Barnils, 1933;
Bernat i Baltrons, 2001; Bibiloni, 1983; Moll, 1952; Moll, 1980; Recasens, 1991; Veny,
1978), lead to different realizations of /k/ in different dialects on Majorca. Specifically, vowel context /i/ allows for an analysis of /k/ in a context that, according to prior phonological descriptions of these dialects, leads to a fronted realization of /k/, even leading to a palatal place of articulation. The /o/ vowel context allows for an analysis of
/k/ in a context that has been reported to lead to a back and velar realization of /k/.
Finally, /a/ and /ə/ vowel contexts allow for a comparison of the production of /k/ in varieties of Catalan that have been reported to produce a fronted version of /k/ (even leading to a palatal place of articulation) in these two vowel conditions, and in other varieties that have not been reported to produce a fronted or palatalized version of /k/ in either of them (Alcover, 1920; Barnils, 1933; Bernat i Baltrons, 2001; Bibiloni, 1983;
Moll, 1952; Moll, 1980; Recasens, 1991; Veny, 1978). Specifically, this design allows for the comparison of the two dialects under study, Manacor and Artà, in their production of /k/ in different vowel contexts. It will allow to test whether there are, as reported in the literature, differences between the two dialects in their production of /k/ in the context of 69
/a/ and /ə/ (Manacor potentially presenting a fronted /k/ and Artà potentially presenting a non-fronted /k/), and similitudes between the two dialects in their production of /k/ in the context of /i/ and of /o/ as a fronted and a non-fronted /k/, respectively.
Regarding number of syllables, the majority of Catalan stimuli are bisyllabic.
However, due to the fact that Catalan did not have enough lexical items in the /i/-vowel condition, several three or four-syllable stimuli had to be selected for this condition.
Some stimuli were stressed on the first syllable and some other stimuli were stressed on the second syllable: all words in the /a/, /o/ and /i/ conditions were stressed on the first syllable (therefore, /a/, /o/ and /i/ were stressed); all words in the /ə/ condition were stressed in the second syllable (therefore, /ə/ was unstressed). Even though Majorcan
Catalan allows /ə/ to appear in stressed position, not enough disyllabic words with stressed word-initial /k/+/ə/ structure were found, which is why the first three vowel conditions (/a/, /o/ and /i/) appeared in stressed position and the fourth vowel condition
(/ə/) appeared in unstressed position. Below is a list of Catalan stimuli in their corresponding vowel conditions. (See Appendix 1 for a full list of Catalan stimuli).
Table 14. Sample Catalan stimuli for /k/ (+ /a/, /o/, /i/, /ə/), for /p/, and filler words
/k/ /k/ + /a/ cada ['ka.ðə] ‘each’ casa ['ka.zə] ‘house’
/k/ + /i/ quina ['ki.nə] ‘which’ quinze ['ki.nə] ‘fifteen’
/k/ + /o/ coca ['ko.kə] ‘cake’ cola ['ko.lə] ‘glue’
/k/ + /ə/ camí [kə.'mi] ‘way’ cavall [kə.'vaʎ] ‘horse’
/p/ /p/ + /a/ pare ['pa.ɾə] ‘father’ pasta ['pa.ɫə] ‘shovel’
Fillers bota ['bo.tə] ‘boot’ figa ['fi.ɣə] ‘figue’ 70
In order to make a comparison between /a/-production after /k/ and /a/-production in a different consonant context, a control group of items was included in the speech material list (n=5). These five lexical items were disyllabic words that started with a stressed /pa/ syllable (e.g., pasta ['pas.tə] ‘pasta’). (See Table 14 for some examples in the control condition and Appendix 2 for a full list of control items for /a/-production).
Finally, stimuli (as well as control items) were interspersed among many distractors or filler words in Catalan (n = 75). (See Appendix 4 for the Catalan distractor list). Distractors were used so that participants did not predict the purpose of the experiment or find out the design of the experiment (Harrington, 2010). This is important since participants have been found to change (reduce in duration and/or quality) predicted target stimuli in comparison to non-predicted target stimuli (Fowler & Housum, 1987;
Lieberman, 1963).
3.2.2. Spanish materials
Speech materials for the production experiment in Spanish consist of a total of 45 stimuli. In Spanish, as in Catalan, phoneme /k/ corresponds to graphemes “c”, “q” and
“k” in contexts observed in (3). In order to elicit the production of a word-initial voiceless velar stop /k/, the stimuli used for the present production experiment contain a word- initial grapheme “c”, “qu” or “k”. These stimuli are divided into three different groups
(15 stimuli per group) depending on the vocalic sound that follows the word-initial consonant: /a/, /o/, /i/. One front vowel (/i/, e.g., quiso /'ki.so/ ‘wanted’, 3rd person singular), one back vowel (/o/, e.g., cosa /'ko.sa/ ‘thing’), and the low vowel (/a/, e.g., cana /'ka.na/ ‘white hair’) were selected for the present study. 71
(3) Phoneme /k/ to grapheme correspondences
/k/ - “c”
“c” + “a”: cala ‘inlet’
“c” + “o”: contra ‘against’
“c” + “u”: cupo ‘space’
“c” + “l”: cloro ‘chlorine’
“c” + “r”: cruz ‘cross’
/k/ - “q”
“qu” + “e/i”: queso ‘cheese’, quiso ‘want’ (past, 3rd sg.)
/k/ - “k”
“k”: karate ‘karate’
Note that the experiment in Spanish contains one less vocalic condition than the experiment in Catalan. The reason for this is that Spanish, unlike Catalan, does not have schwa. The three vocalic conditions for Spanish coincide with the three vocalic conditions (minus schwa) in the Catalan experiment so as to be able to compare production of /k/ in the same vocalic conditions between the two languages. Just as for the Catalan data, these vowel contexts allow for an analysis of /k/ in conditions that have been reported to lead to different realizations of /k/ in Catalan. Specifically, context /i/ allows for an analysis of /k/ in a context that, according to the literature, leads to a more fronted production of /k/ both in Manacor and in Artà. Context /a/ allows for a comparison of /k/ between the two dialects in a context that has been reported to lead to a fronted /k/ in Manacor and to a non-fronted /k/ in Artà. Context /o/ allows for an analysis 72 of /k/ in a context that has been reported to lead to a non-fronted production of /k/. To sum up, these contexts allow for the analysis of potential crosslinguistic transfers from
Catalan to Spanish if these previously reported patterns for Catalan are observed in the productions in Spanish of /k/.
Regarding number of syllables, the majority of Spanish stimuli are bisyllabic.
However, due to the fact that not enough items were found for the /i/-vowel condition, several three-syllable stimuli had to be selected for this condition. All words are stressed on the first syllable (independently from the number of syllables they have). Below is a list of Spanish stimuli used in the present study in their corresponding vowel conditions
(See Appendix 3 for a full list of Spanish stimuli).
Table 15. Sample Spanish stimuli for /k/ in each vowel condition (/a/, /i/, /o/) and of filler words
/k/ /k/ + /a/ canon ['ka.non] ‘canon’ cable ['ka.βle] ‘cable’
/k/ + /i/ quinto ['kin̪ .to] ‘fifth quiste ['kis.te] ‘cyst’
/k/ + /o/ costo ['kos.to] ‘costo’ cosa ['ko.sa] ‘thing’
FILLERS Bebé [be. 'βe] ‘baby’ lote ['lo.te] ‘set’
Finally, stimuli were interspersed among many distractors or filler words (n = 75).
(See (15) above for sample filler words and see Appendix 5 for a full list of distractors in
Spanish). These distractors were also Spanish words.
73
3.3. Procedure
All participants conducted the two production tasks (one in Catalan, one in
Spanish) in one experimental session per person. The sequence of events taking place in each experimental session was the following: (1) they were asked to read and sign an informed consent form, (2) they conducted the two experimental tasks, (3) they filled out the BLP.
In order to control for language mode effects (Grosjean, 1999), half of the participants began performing the task in Catalan (were greeted and given instructions in
Catalan too) and then finished with the task in Spanish (and given instructions in
Spanish); and the other half began performing the task in Spanish and finished with the task in Catalan. Production of speech in language A entails the activation of language A in the mind of the speaker. Subsequent production of speech in language B activates language B. The activation of language B, however, does not entail the deactivation of language A. Therefore, the productions of language B could be affected by language A’s phonetic characteristics. In order to counterbalance the potential effects of language mode, half of the participants produced speech in Catalan first and finished in Spanish, and the other half of the participants produced speech in Spanish first and finished with
Catalan (Grosjean, 1999).
The stimuli and distractors for each language were randomized three different times in order to avoid position effects, that is, in order to even out potential differences based on the position of a stimulus within a list of stimuli. Three different lists were created for each language. Participants were balanced regarding the randomization list they were given for each language. They were instructed to read every word of each list 74 within the carrier phrase Sa paraula és _____ ‘The word is _____’ so as to make the production of the sentences more natural and to make sure each word appears in the same rhythmic position (Ladefoged, 2003). They were instructed to read each stimulus ‘as naturally as possible’ and at a ‘regular speed’. Participants read each item twice. The exact objective of the study (i.e., analysis of /k/ and /a/ productions) was not revealed to the participants until they had finished their participation. The Catalan /k/ data consists of a total of 4800 tokens: 20 participants × 2 groups (Manacor, Artà) × 4 vocalic conditions
(/a/, /ə/, /o/, /i/) × 15 target words × 2 repetitions. The Catalan /p/ data consists of a total of 400 tokens: 20 participants × 2 groups (Manacor, Artà) × 5 target words × 2 repetitions. The sum of the two types of stimuli renders a total of 5200 tokens for Catalan data. A total of 3600 tokens were collected from the Spanish production task: 20 participants × 2 groups (Manacor, Artà) × 3 vocalic conditions (/a/, /o/, /i/) × 15 target words × 2 repetitions.
Participant productions were recorded in a quiet room at the home of each of the participants in Majorca, Spain. Productions were recorded directly to disk using Praat
(Boersma & Weenink, 2011) and a head-worn dynamic microphone (Shure SM10A).
Speech was digitized at 44.1 sampling rate and 16-bit quantization. An analog-to-digital converter and microphone preamplifier were used (Sound Devices USBPre2).
3.4. Acoustic analysis
All target productions (n = 5200 Catalan tokens + 3600 Spanish tokens) were extracted for analysis. Fillers (n = 6000 Catalan fillers + 6000 Spanish fillers) were left out of the analysis. Due to recording errors such as laughter, coughing or other 75 disfluencies, or due to background noise, 20 Catalan tokens and 26 Spanish tokens had to be removed from subsequent analyses. This rendered a total of 5180 Catalan tokens and a total of 3574 Spanish tokens for analysis.
All target productions were hand-segmented in Praat. Synchronized waveform and spectrographic displays were used for identification and visual segmentation
(Hualde, Simonet & Nadeu, 2011). Each target production was marked for: (1) onset and offset of the burst of the stop consonant (both for Catalan /k/ and /p/, and for Spanish /k/),
(2) beginning and end of subsequent vowel, (3) length of target word. The beginning and end of each vowel after each target production was taken to occur at the first and last – respectively– pitch period in which F2 was visible and/or where an intensity similar to that of the vowel’s midpoint was present (Simonet, 2011a). Finally, the length of the target word was marked from the stop consonant burst onset of the first word segment to the end of the final speech segment in the word. All this annotation process was conducted via four levels of annotation or textgrids on Praat. The first annotation level was used to mark the onset of the consonant burst. The second level was used to mark the length of the vowel following the word-initial stop consonant. The third level was used to mark the length of the target word. Finally, the fourth level was used to annotate the name of the stimulus (the name of the produced stimulus). These temporal acoustic landmarks assigned to each target production were used in order to conduct several acoustic analyses. By using a Praat script, different acoustic properties were extracted from each target sound file.
An approach to the classification of voiceless stop consonants in terms of place of articulation analyzes the inherent acoustic properties of the consonant burst, looking at 76 different aspects of its noise spectrum (Forrest, Weismer, Milenkovic, & Dougall, 1988).
Burst frequency is the frequency spectrum of the release burst and it varies depending of the length of the vocal tract in front of the constriction: “the closure at the time of the release is still quite narrow, and thus the front and back cavities of the vocal tract are not acoustically coupled” (Johnson, 2011, p. 176). Velar stops are characterized by a lower burst frequency than other more fronted places of articulation because the portion of the vocal tract in front of the velar constriction is substantially longer than the portion of the vocal tract in front of a more fronted point of articulation, such as a palatal or a coronal consonant. The burst frequency for velar stops usually falls between 1500 and 2500 Hz
(Reetz & Jongman, 2009). Palatal stops, therefore, will show a higher burst frequency than velar stops.
For the present study, one spectrum was generated for each target production. It was generated extracting a variable Gaussian window that was dependent on the length of
VOT, that is, a section of the oscillogram going from the articulator release to the beginning of modal phonation or voicing was extracted. This section was then submitted to a Fourier transformation and the spectrum was calculated.
From the spectrum, two spectral moments were calculated, namely Mean and
Variance (Reetz & Jongman, 2009). These acoustic metrics were extracted using Praat’s built-in function, with a power of 2. The first spectral moment, mean, reflects the average energy concentration across a specific section of the spectrum (Cicres, 2011). The spectral Mean is also known as Center of Gravity and it has been used as an acoustic correlate of place of articulation in stops and fricatives: the more back the place of articulation is, the lower the Center of Gravity value it has; the more fronted the place of 77 articulation is, the higher the Center of Gravity value it has (Gordon, Barthmaier, &
Sands, 2002; Schmid, 2011). The present study used the Center of Gravity measure as the most reliable correlate of place of articulation of stop consonants.
The second spectral moment that was explored was Variance of the spectrum
(also known as Standard Deviation). Variance reflects deviation of frequencies in relation to the mean. Variance of the spectrum has been previously used to indicate place of articulation of different segments. For instance, Stoel-Gammon, Williams and Buder
(1994) used this measure, among others, to report differences between American English and Swedish /t/. They reported that coronal English stops (which have been described as alveolar) had smaller Variance values, which indicated a more compact burst spectrum than coronal Swedish stops (which have been described as dental). Similarly, Sundara
(2005) uses the Variance measure to find differences between Canadian English and
Canadian French coronal stops, which differ in place of articulation. Systematic differences across the two languages were found for this measure: English coronal bursts are more compact (have lower Variance values) than French coronal bursts. Sundara
(2006) finds similar results for English-French bilinguals, who show robust differences in
Variance values for their two languages. To sum up, these studies have used the Variance measure to examine place of articulation in coronal consonants. Even though there are no studies of velar stops using this measure to determine place of articulation, it should be amenable to be used as an index of place of articulation also in velar stops.
The last measure used for analysis of the stop consonant productions is Relative
Intensity, which was extracted from the intensity contour in Praat. Relative Intensity measures the difference between the minimum intensity during the production of a 78 consonant and the maximum intensity during the following vowel (Jongman, Blumstein, and Lahiri, 1985). Jongman et al. (1985) first introduced this measure in order to account for differences in place of articulation of voiceless coronal stops in Malayalam. They found that whereas alveolar stops presented a louder burst and relative burst amplitude ratios below 5, dental stops presented by a softer burst and relative burst amplitude ratios above 5. Thus, Jongman et al. (1985) were able to successfully use the Relative Intensity measure to demonstrate differences in place of articulation for coronal stops. In a later study, Stoel-Gammon et al. (1994) were also able to use this measure to distinguish between American English and Swedish coronal stops. English alveolar stops had louder bursts and lower Relative Intensity of the burst than Swedish dental stops. These studies suggest that differences in place of articulation for coronal stops could be observed by differences in Relative Intensity values. Even though these studies have used the Relative
Intensity measure to index place of articulation in coronal stops, this measure should also be able to index place of articulation in velar stop consonants. Note that even though the
Variance and Relative Intensity measures have been used in previous studies as indicators of place of articulation, these are more indirect and obscure than the Center of Gravity measure is; that is, the Center of Gravity measure is a much more reliable indicator of place of articulation than the other two measures. Therefore, the main measure for the current analyses will be Center of Gravity. For the purposes of the present dissertation, two conditions will be established as different if at least two out of the three acoustic measures (Center of Gravity supported by either the Relative Intensity or the Variance measure) determine there is a statistical difference. This entails that if only one of the three measures determines the presence of a statistical difference, but the other two 79 measures determine that there is no difference, the results will be interpreted as there being no difference.
To summarize, in order to examine potential context- or group-related differences in the production of velar stop consonants by the participants, two acoustic measures were extracted from the spectrum: Center of Gravity and Variance. Additionally, one acoustic measure was extracted from the intensity curve: Relative Intensity. These three acoustic measures will be used for the analysis of the present research questions of /k/ production.
Apart from the stop consonants, each vowel following the target word-initial velar stop was segmented using Praat. Following the procedure explained in Simonet (2011a,
2014), formant trajectories, especially the second formant (F2), as well as intensity displays were taken as indicators of vowel onsets and offsets. As explained above, the onset of the vowel was marked on the first pitch period in which F2 was visible and/or an intensity similar to that of the vowel’s midpoint was present. The offset was marked on the last pitch period. F1 and F2 were extracted from the vowel midpoint using a Praat built-in function. Formant values were extracted in Hertz and were normalized in a speaker-intrinsic procedure (Simonet, 2011a). Normalization was conducted in order to filter out idiosyncratic acoustic variation, such as physiological gender differences
(different size of the speaker’s vocal tract) (Casillas & Simonet, 2016). The Lobanov method was used (1971), which transforms Hz values into z-scores. This method calculated an overall mean and standard deviation value for F1 and F2 for each participant using all of the vowels available (/a, i, o, ə/). This created a by-subject anchor point that was an approximation to each participant’s vowel space. Then, the z-score 80 formula was applied to each target formant value, i.e., the overall mean from the actual
F1 values for all /a/ tokens (of each speaker) was extracted and subtracted from the corresponding by-subject anchor. This allowed that each target formant was relative to each participant’s vowel space (Casillas & Simonet, 2016). The same was done for F2.
This was done for each participant.
The degree of frontedness of the vowels was determined by the F2 value of the vowel: a vowel with a more forward place of articulation presents a higher F2 value and a vowel with a more backward place of articulation presents a lower F2 value. Even if F2 was used as the main acoustic metric to measure vowel differences between groups, F1 values were also extracted for analysis. F1 is inversely related to vowel height (the higher the vowel, the lower the F1). Potential differences between groups in vowel height were also explored.
In order to examine whether the low vowel is more fronted in Majorcan Catalan in general or whether it is only fronted when it is produced after /k/, it was necessary to compare /a/-productions after /k/ to /a/-productions after /p/ in the Catalan productions.
The same procedure for /a/ segmentation after /k/ was followed to segment /a/ after /p/.
3.5. Statistical analysis
For the analysis of /k/, three different blocks of analyses were conducted: (1)
Catalan /k/ productions only; (2) Spanish /k/ productions only; (3) comparison of Catalan and Spanish /k/ productions. For the first two blocks, Catalan and Spanish productions of
/k/ were analyzed independently. The dependent variables were three: (1) Center of
Gravity, Relative Intensity and Variance from a variable window. Each acoustic 81 measurement was explored separately. An identical procedure was used to explore all three measurements: In all cases, a mixed-effects generalized linear regression model was used. For all models, participants were random intercepts (Baayen et al., 2008), with random slopes for each vowel (Barr et al., 2013). Statistical significance of group, vowel, and group by vowel interactions were assessed using hierarchical partitioning of variance via nested model comparisons. Simultaneous Tests for General Linear Hypotheses analyzed pairwise comparisons using Tukey Contrasts with adjusted p-levels. The independent between-subject factor was group (Manacor, Artà) and the within-subject factor was vowel context (/a/, /ə/, /o/, /i/ for Catalan data; /a/, /o/, /i/ for Spanish data).
Pairwise comparisons were based on the findings of these statistical models.
For the third block of analyses, a within-subject comparison of Catalan and
Spanish productions of /k/ was conducted. The dependent variables were the same as for the previous analyses: Center of Gravity, Relative Intensity and Variance. A separate analysis was conducted for each acoustic measure. Data were analyzed using mixed- effects generalized linear regression models. There were two within-subject variables for this analysis: language (Catalan, Spanish) and vowel context (/a/, /i/, /o/). Notice that the only vowel contexts compared are /a/, /i/ and /o/ (and not /ə/) because these were the three vowel contexts that had productions in both languages (Spanish productions did not contain /ə/).
For the analysis of Catalan /a/, another set of analyses was conducted. Only
Catalan data were analyzed for this part of the analysis. The dependent variables were normalized F1 and F2 values, as explained above. These two acoustic measurements were explored separately. An identical procedure was used to explore the two dependent 82 variables: A mixed-effects generalized linear regression model was used. The independent between-subjects variable was group (Manacor, Artà) and the independent within-subjects factor was consonant context (/k/ or /p/). Again, follow-up tests were based on the findings of these statistical models. All statistical analyses for the present study were conducted with R (R Development Core Team, 2016) version 3.2.1.
83
CHAPTER 4
CATALAN
4.1. Introduction
This chapter analyzes the acoustic characteristics of the burst of /k/ in the Catalan spoken in two areas of the island of Majorca (Manacor, Artà). Therefore, this chapter only concentrates on production in Catalan. The aims of this chapter are twofold. The first aim is to analyze whether the process of velar palatalization has been phonologized in the traditionally palatalizing area (Manacor) compared to the traditionally non- palatalizing area (Artà). For that, the nature of the burst of /k/ productions is analyzed statistically to observe potential differences based on group. Additionally, differences within each subdialect in terms of phonetic contexts are analyzed. Specifically, the effect of vowel context (/k/ + /a/, /i/, /o/, /ə/) on /k/ production is analyzed in order to see whether certain contexts trigger velar palatalization more than others. Traditional accounts of the phenomenon (Alcover, 1920, among many others) suggest that the palatalizing area presents velar palatalization before /a/, /i/, /ə/ (apart from other contexts not analyzed in the present study), but not before /o/. The non-palatalizing area has been said to front /k/ only before /i/, due to consonant-to-vowel coarticulation. Note also that these traditional accounts only report a dialectal distinction in the phonological distribution of palatalized /k/, but do not make any references to potential differences between these two dialectal areas with regards to their degree of palatalization. Therefore, the present dissertation analyzes (1) whether the traditional contextual distribution
(previously only reported impressionistically) is observed in these data and (2) whether 84 there are differences in the degree of palatalization (irrespective of position or in the palatalizing contexts such as /i/). If /i/ triggers an even more palatalized version of /k/ in
Manacor than it does in Artà (where it should trigger fronting but to a smaller degree), it will suggest that the process of velar palatalization has been phonologized for the
Manacor group, that is, that what started as a coarticulatory gesture has been exaggerated and reinterpreted as an intended gesture. To sum up, the present dissertation examines whether the traditionally reported phonological distribution can actually be supported with acoustic data and whether there is not only a difference in the contextual distribution but also in the degree of palatalization.
The second aim of this chapter is to analyze the low vowel /a/ in the Catalan spoken in the two areas in order to potentially determine the relationship between the process of /a/-fronting and /k/-palatalization. Velar palatalization has been said to occur in dialects that present fronted versions of /a/, such as French (Buckley, 2003). The present dissertation analyzes whether this is the case of Majorcan Catalan in general or of one of its dialects only. Specifically, it examines the following possibilities: If /a/ is similarly fronted in both areas (i.e., /a/ is fronted in Majorcan Catalan in general) but there is a difference in the degree or distribution of palatalization (i.e., Manacor speakers differ from Artà speakers in terms of different contexts or degrees), it could suggest that
/a/ has triggered palatalization for the area that palatalizes /k/ (Manacor). On the other hand, if both /a/ and /k/ present crossdialectal differences (i.e., they are both more fronted in Manacor than in Artà), it could suggest that there is in fact a not casual relationship between the two processes, but the order of influence would not be determined: (a) /a/- 85 fronting might have triggered /k/-palatalization in its inception, (b) /k/-palatalization might have triggered /a/-fronting, or (c) both processes might have influenced each other.
The following subsection (4.2) provides the results for the data from Catalan production. First, the results for the analysis of /k/ are presented (4.2.1). Then, the results for the analysis of /a/ are presented (4.2.2). Finally, a summary of the findings for the
Catalan data is presented (4.2.3). A discussion of Catalan results is found in Chapter 6
(Discussion and Conclusions).
4.2. Results
4.2.1. Analysis of /k/ in Catalan
This section is concerned exclusively with an analysis of the Catalan /k/ production. The acoustic data were analyzed using mixed-effects generalized linear regression models to test the effect of group (Manacor, Artà) and vowel context (/a/, /i/,
/o/, /ə/) on different acoustic measures (Center of Gravity, Relative Intensity, Variance).
For all models, participants were random intercepts (Baayen et al., 2008), with random slopes for each vowel (Barr et al., 2013). Statistical significance of group, vowel context, and group by vowel context interactions were assessed using hierarchical partitioning of variance via nested model comparisons. Simultaneous Tests for General Linear
Hypotheses analyzed pairwise comparisons using Tukey Contrasts with adjusted p-levels.
The following subsections are divided according to each of the acoustic measures.
86
4.2.1.1. Center of Gravity
The acoustic Center of Gravity data were submitted to a mixed-effects generalized linear regression model, which tests main effects using a nested model comparison. The analysis had a (4) × 2 mixed design, with vowel context (/a/, /i/, /o/, /ə/) as the within-subjects factor, and with group (Manacor, Artà) as the between-subjects factor. The best model used for this analysis explained 77% of variance. The results yielded a main effect of vowel context (χ2(3) = 105.49, p < 0.001) and a group by vowel context interaction (χ2(3) = 58.70, p < 0.001). There was no main effect of group (χ2(1) =
0.88, p > 0.05). Table 16 shows the results for the mixed-effects generalized linear regression model test for the Center of Gravity measure. Figure 1 plots the Center of
Gravity values for group (Manacor, Artà) and vowel context (/a/, /i/, /o/, /ə/). Table (a), in
Appendix 6, shows mean and standard deviation values for the Center of Gravity measure.
In order to further explore the group by vowel context interaction, several pairwise comparisons were conducted, which compared group productions per each vowel context separately. For vowel contexts /a, ə, i/, Center of Gravity varied as a function of group. Specifically, for the subset of data containing vowel context /a/, the
Center of Gravity measure was 1033.14 units (± 140.72 se) higher in Manacor than in
Artà (t = 7.34, p < 0.001); for the /i/ vowel context, it was 526.0 units (± 177.8 se) higher in Manacor than in Artà (t = 2.96, p < 0.01); and for /ə/ vowel context, it was 1226.52 units (± 115.16 se) higher in Manacor than in Artà (t = 10.65, p < 0.001). Differently, for vowel context /o/, the Center of Gravity measure did not vary as a function of group (t =
0.11, p > 0.05). 87
Table 16. Results for the mixed-effects generalized linear regression model test for
Center of Gravity for group (Artà, Manacor) and vowel context (/a/, /i/, /o/, /ə/) for
Catalan productions. The baseline levels are Artà (for group) and /a/ (for vowel context)
Estimate SE Df t-value Probability
(Intercept) 1707.70 99.52 40.06 17.160 p < 0.001
Manacor 1032.41 140.72 40.04 7.337 p < 0.001
/i/ 1101.42 114.80 39.93 9.594 p < 0.001
/o/ −699.73 96.97 40.18 −7.216 p < 0.001
/ə/ −217.42 61.94 39.99 −3.510 p < 0.05
Manacor: /i/ −506.62 162.36 39.94 −3.120 p < 0.05
Manacor: /o/ −1027.81 137.10 40.12 −7.497 p < 0.001
Manacor: /ə/ 194.07 87.59 39.97 2.216 p < 0.05
Figure 1. Center of Gravity (Hz) for group (Artà, Manacor) and vowel context (/a/, /i/,
/o/, /ə/) for Catalan productions only
6000
4000 Group Art`a Manacor
Center of2000 gravity
0 /a/ /i/ /o/ /@/ Vowel
88
As seen in Figure 1, for both groups of speakers, the more fronted the vowel context is (/i/ > /a/ > /ə/), the higher its Center of Gravity values are; the less fronted the vowel context is (/o/), the lower the values are. Even if this effect is observed in both groups of speakers, it is more marked for the Manacor group, which shows even higher values for /k/ in the context of /i/, /a/ and /ə/ (in that order) than the group from Artà does.
Importantly, the Center of Gravity of /k/ does not differ as a function of dialect in the context of /o/. This suggests that this is not the case that /k/ is generally more palatalized in Manacor than it is in Artà: it is more palatalized, but only in the contexts that favor palatalization as they have been described for this dialect.
Summing up the results for the Center of Gravity measure, there were significant differences between groups for /k/ in the vowel contexts /a/, /i/ and /ə/. Differently, there were no differences for /k/ in the context /o/. The difference between groups was larger for vowel contexts /a/ and /ə/ (as expected), but vowel context /i/ also presented differences based on group. Finally, a pattern was observed for both groups: the more fronted vowel contexts (/i/ > /a/ > /ə/) showed higher Center of Gravity values for /k/ than the back vowel context did (/o/). This difference was especially salient for Manacor speakers, i.e., fronted vowel contexts triggered higher Center of Gravity values for /k/ than for the other group of speakers.
4.2.1.2. Relative Intensity
The acoustic Relative Intensity data were submitted to a mixed-effects generalized linear regression model. The analysis had a (4) × 2 mixed design, with vowel context (/a/, /i/, /o/, /ə/) as the within-subjects factor, and with group (Manacor, Artà) as 89 the between-subjects factor. The best model used for this analysis explained 55% of variance. The results yielded a main effect of group (χ2(1) = 11.88, p < 0.001), a main effect of vowel context (χ2(3) = 35.12, p < 0.001), and a group by vowel context interaction (χ2(3) = 9.58, p < 0.03). Table 17 shows the results for the linear mixed effects model test. Figure 2 plots the Relative Intensity values for group and vowel context.
Table (b), in Appendix 6, shows mean and standard deviation values.
In order to further explore the group by vowel context interaction, several follow- up tests were conducted, which compared group productions per each vowel context separately. For vowel contexts /a, ə, i/, Relative Intensity varied as a function of group.
Specifically, for the subset of data containing vowel context /a/, the Relative Intensity measure was 3.18 units (± 0.73 se) higher in Manacor than in Artà (t = 3.62, p < 0.001); for the /i/ vowel context, it was 3.61 units (± 0.73 se) higher in Manacor than in Artà (t =
4.98, p < 0.001); and for /ə/ vowel context, it was 3.27 units (± 0.74 se) higher in
Manacor than in Artà (t = 4.42, p < 0.001). Differently, for vowel context /o/, the
Relative Intensity measure did not vary as a function of group (t = 1.30, p > 0.05).
Summing up the results for the Relative Intensity measure, there were significant differences between groups for the quality of /k/ as produced in vowel contexts /a/, /i/ and
/ə/. Specifically, Manacor speakers produced a /k/ with higher Relative Intensity values in these three vowel contexts than Artà speakers did. Additionally, productions for /k/ in the
/o/ vowel context did not vary as a function of group: Manacor and Artà speakers produced /k/ with similar Relative Intensity values when it was produced in the context of
/o/.
90
Table 17. Catalan results for the mixed-effects generalized linear regression model test for Relative Intensity for group (Artà, Manacor) and vowel context (/a/, /i/, /o/, /ə/).
Baseline levels: Artà (group) and /a/ (context). Non-significant values marked as “n. s.”
Estimate SE Df t-value Probability
(Intercept) 10.05 0.62 40.02 16.18 p < 0.001
Manacor 3.18 0.87 40.01 3.61 p < 0.001
/i/ 0.15 0.53 39.94 0.29 n. s.
/o/ 0.58 0.58 40.09 0.99 n. s.
/ə/ −1.53 0.39 39.93 −3.93 p < 0.001
Manacor: /i/ 0.43 0.75 39.95 0.57 n. s.
Manacor: /o/ −2.18 0.82 40.05 −2.65 p < 0.05
Manacor: /ə/ 0.08 0.55 39.91 0.15 n. s.
Figure 2. Relative Intensity for group (Artà, Manacor) and vowel context (/a/, /i/, /o/, /ə/) for Catalan productions only
30
20
10 Group Art`a Manacor
0 Relative intensity
-10
/a/ /i/ /o/ /@/ Vowel
91
4.2.1.3. Variance
The acoustic Variance data were submitted to a mixed-effects generalized linear regression model. The analysis had a (4) × 2 mixed design, with vowel context (/a/, /i/,
/o/, /ə/) as the within-subjects factor, and with group (Manacor, Artà) as the between- subjects factor. The best model used for this analysis explained 71% of variance. The results yielded a main effect of group (χ2(1) = 14.48, p < 0.001), a main effect of vowel context (χ2(3) = 91.58, p < 0.001), and a group by vowel context interaction (χ2(3) =
32.23, p < 0.001). Table 18 shows the results for the mixed-effects generalized linear regression model test for the Variance measure. Figure 3 plots the Variance values for group (Manacor, Artà) and vowel context (/a/, /i/, /o/, /ə/). Table (c), in Appendix 6, shows mean and standard deviation values for the Variance measure.
In order to further explore the group by vowel context interaction, several pairwise comparisons were conducted, which compared group productions per each vowel context separately. The results showed that for vowel context /a/, the Variance measure was 584.73 units (± 86.54 se) higher in Manacor than in Artà (t = 6.76, p <
0.001); for vowel context /i/, it was 284.50 units (± 82.23 se) higher in Manacor than in
Artà (t = 3.46, p < 0.002); and for vowel context /ə/, it was 587.05 units (± 63.19 se) higher in Manacor than in Artà (t = 9.29, p < 0.001). Differently, the test comparing only group productions for /o/ showed that the Variance measure did not vary as a function of group for vowel context /o/ (t = 0.21, p > 0.05).
92
Table 18. Variance results for the mixed-effects generalized linear regression model test for group (Artà, Manacor) and vowel context (/a/, /i/, /o/, /ə/) for Catalan. Baseline levels are Artà (group) and /a/ (vowel context). Non-significant values are marked as “n.s.”
Estimate SE Df t-value Probability
(Intercept) 1429.87 61.19 40.03 23.36 p < 0.001
Manacor 584.71 86.53 40.01 6.75 p < 0.001
/i/ 387.85 66.63 39.94 5.82 p < 0.001
/o/ −606.46 81.33 40.08 −7.45 p < 0.001
/ə/ −167.31 42.22 40.13 −3.95 p < 0.05
Manacor: /i/ −300.31 94.24 39.94 −3.18 p < 0.001
Manacor: /o/ −570.44 115.00 40.05 −4.96 p < 0.001
Manacor: /ə/ 2.38 59.77 40.11 0.04 n. s.
Figure 3. Catalan Variance for group (Artà, Manacor) and vowel context (/a/, /i/, /o/, /ə/)
4000
3000
Group Artà Manacor Variance 2000
1000
0 /a/ /i/ /o/ /!/ Vowel 93
As seen in Figure 3, the more fronted the vowel context is (/i/ > /a/ > /ə/), the higher its Variance values are; the less fronted the vowel context is (/o/), the lower
Variance values it has for both groups of speakers. Even if this effect is observed in both groups of speakers, it is more marked for the Manacor group than for the Artà speakers.
Manacor speakers show even higher Variance values for /k/ in the context of /i/, /a/ and
/ə/ (in that order) values than the group from Artà does. Importantly, /k/ for vowel context /o/ did not vary as a function of group. Again, this suggests that Manacor Catalan does not contain /k/ with generally higher Variance values than Artà does: Manacor
Catalan does present a /k/ with higher Variance values than Artà Catalan, but this happens only when /k/ is produced in the vowel context that triggers palatalization, that is, before /i/, /a/ and /ə/.
Summing up the results for the Variance measure, there were significant differences between groups for vowel contexts /a/, /i/ and /ə/. Productions for /o/ vowel context did not vary as a function of group. Additionally, for both groups, the more fronted vowel contexts (/i/ > /a/ > /ə/) presented higher values than the vowel context
(/o/), a difference that was larger for the Manacor group with even higher Variance values for the more fronted vowel contexts.
4.2.1.4. Summary of results for /k/ in Catalan
For each acoustic measure separately (Center of Gravity, Relative Intensity,
Variance), a general (4) × 2 analysis was conducted with vowel context as the within- subject variable (/a/, /i/, /o/, /ə/) and with group (Manacor, Artà) as the between-subject variable. For each dependent variable, a series of pairwise comparison tests were 94 conducted comparing group for each vowel context separately. All three analyses rendered similar results: /k/ productions were significantly different between the two groups of participants for vowel contexts /a/, /i/ and /ə/: Manacor speakers presented a more fronted /k/ (i.e., higher Center of Gravity, Relative Intensity and Variance values) for these three vowel contexts than Artà speakers did. Importantly, no statistical difference was found for vowel context /o/ between groups for any of the three acoustic measures. This specific finding allows us to discard the idea that Manacor speakers might be producing generally more fronted /k/ (when compared to Artà speakers) independently from the vowel context in which it is produced. Rather than this being the case, it seems that Manacor speakers have a /k/ that is governed by allophonic variation, that is, speakers produce a different version of /k/ depending on the vowel that comes after it.
Table 19 summarizes the reported differences between dialects for each of the vowel contexts within each acoustic measure.
Table 19. Crossdialectal difference in Catalan per vowel context (/a/, /i/, /ə/, /o/) for each of the acoustic measures (Center of Gravity, Relative Intensity, Variance). A check (✔) means there are differences between speakers from Manacor and Artà; a cross (✖) means there are no differences between speakers from Manacor and Artà.
/a/ /i/ /ə/ /o/
Center of Gravity ✔ ✔ ✔ ✖
Relative Intensity ✔ ✔ ✔ ✖
Variance ✔ ✔ ✔ ✖
95
Apart from the crossdialectal differences observed in Table 19 for the three first vowel contexts, another important piece of information stands out when observing the values for /k/ in the context of each vowel context. Specifically, /k/ in the context of /i/ is more fronted in Manacor speakers than in the other vowel contexts (followed by /a/, /ə/, and finally, /o/, in that order) for all three acoustic measures. This is expected since vowel context /i/ is the one that normally triggers a higher degree of palatalization of the previous consonant, due to regressive vowel-to-consonant coarticulation. In fact, this vowel context also triggers a more fronted /k/ production for Artà speakers. However, when comparing both groups, Artà /k/ produced in the context of /i/ does not present values that are as high as the ones presented by Manacor speakers for this same vowel context. This suggests that the gesture for the articulation of /k/ in the context of /i/ is much more exaggerated (i.e., is brought even to a more fronted position) by Manacor speakers than it is by Artà speakers. For Artà speakers, it seems that this fronting is simply a result of coarticulation, but for Manacor speakers, it is an intended gesture, suggesting the phonologization of this velar palatalization in Manacor speech.
4.2.2. Analysis of /a/ in Catalan
This section is concerned exclusively with an analysis of /a/ production in
Catalan. Productions of the low vowel by the two groups of participants (Manacor, Artà) are compared with each other for each of two consonant contexts (produced after /k/ and produced after /p/). These data were analyzed using mixed-effects generalized linear regression models. Data were analyzed for the F1 and F2 measures separately. F1 and F2 values were normalized (as explained in Chapter 3). As in the previous analyses, 96 participants were random intercepts, with random slopes for each consonant. Main effects and interactions were assessed using hierarchical partitioning of variance via nested model comparisons. Simultaneous Tests for General Linear Hypotheses analyzed pairwise comparisons using Tukey Contrasts with adjusted p-levels. Figure 4 plots the F2 values for group (Manacor, Artà) and consonant context (/k/, /p/). Table (d), in Appendix
6, presents mean and standard deviation values for normalized F1 and F2 data per group and consonant context. The following subsections are divided according to each of the two acoustic measures.
Figure 4. Normalized F1 and F2 values for /a/ productions in the context of /k/ (red) and in the context of /p/ (blue) for Artà (left) and Manacor (right) groups
Art´a Manacor
a -2
-2 Context /ka/ /pa/ -1 -1 a a a a
0 a a a a 0 a a a a a a a a aa a a a aaa a a a a a aaa a a a a a a a a a aa a a a aaa aa aaa a a a aaa a a a a aa a aa aa aa aa a a a a a a a a a aa a a a a a a a aa a aa a aaaa a a aaa a a aaaaaaaaaaa a aa a a aaa aa aa aa aa aa aa aaaa aaaa a aaaa a aaa aaaaa a a a aa a a a a aa aa aaa aaa aa aa a a aaaa aa a aa a aaaaaaaaaa aaa aa aa aaa aaaa a aa a a aaa a aaaaaaaaaaaaa a aaaaaa aaa aa aaaaa aaa aaaa aa aaa aaa a aaaa aaaa aa aa a a a aa aaaaaaa aaaaaaaaaaaa aaaaa aaaa aaaaaa aa aa a aaa a aa aaaaaaaa a aa aaaaaaaa a a aaa a 1 a a a a a a a a a a a a a a aa a a aaaaaaaaaaaaaaaaaaaa aaaaaaaaaaaaaaaaa aaaa aaaa a a a a aaa aaaa a aaaaaa aa a aaaa aaaaaaaaa aaaaaaaa a a a a a aaaa a aa aaaaaaaaaaaaaaaaaaaaa aaaaaaaaaaa aaaaaaa a aa a a aa a a a aa aaaaaaaaaaaaaaaaaaaaaa aaaaaaaaaa aa aa a a a a a a aaaa aaaaaaaaaa aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa a a a a a aa aa aaaa aaaa aaaaaaaaaaaaaaaaaa aaaa aaaa aa a aa aaa aa a a a a a a a aa a aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa a a a a a a aa a aaaaaa aaaaa aaaaaaaaaaaaaaaaaaaaaa aaaaaaaaa a aa aaaa a 1 a aaa aa aaaaaaaaaaaaaaaaaaaaaaaaaaaa a a a aa aa a a a a a aa a aaaaaaaaaaaa aaaaaaaaaaaa aaaaa aaaaa aa a a a a a aaa aaa aaaaaaaaaaaaa aa a aa a aa a aa a a a aa a aaaaa aa aaaaa a a aaa aaa a aa a a aa a a a aaaa a a aaaaa aa aaaa aa a aa a a a a aaaaa aa a aaa aa aa aaaaaaa a aa a a F1 (z-scores) aaa a aaa aa aa aaaaaaaaaa a a a a a a a aaaa aaaaaaaaaaaaaaaaaa a aa a aaaaaaa aa aa a a a aa a aaaaaaaa aaaa a aaa aa a aa a a a aaaa aaaa aaaaaaaaaaaaaaaaa aaaa aaa a a aaaaaa a a a aaaaaa a a a a a a a aaaaaa aaa aa aaaa aaaa aaaaaaa aaaa aa a a aa a a aaa a aa a a a aa aaa aaa aa aaaaa aa aaaaa a aa a a a a a a a a a aaaa a a a a a a a aaa aaa aaaaa aa a a a a a a a aa a a aaaaa a aaaa a a a aa a a a a aaaaaa aa aa a 2 a aa a a a a aaa a a a a a a a 2 a a 3
a 3
1.0 0.5 0.0 -0.5 -1.0 -1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 F2 (z-scores) F2 (z-scores)
4.2.2.1. F1 measure
The acoustic F1 data were submitted to a mixed-effects generalized linear regression model. The analysis had a (2) × 2 mixed design, with consonant context (/k/,
/p/) as the within-subject variable, and with group (Manacor, Artà) as the between-subject 97 variable. The independent variable was the normalized F1 value. The best model used for this analysis accounts for 36% of the variance. The results yielded a main effect of consonant (χ2(1) = 29.26; p < 0.001) and a group by consonant context interaction (χ2(1)
= 25.56; p < 0.001). There was no main effect of group (χ2(1) = 1.50; p > 0.05). These results suggest that /a/ is significantly higher in the context of /k/ (M = 0.96, sd = 0.32) than in the context of /p/ (M = 1.27, sd = 0.24) for the Manacor group. Note that, in this context, /a/ has a lower F1, which means that, on an inverted F1 x F2 formant chart, this vowel is “higher”. For the Artà group, both the /k/ context (M = 1.09, sd = 0.37) and the
/p/ context (M = 1.15, sd = 0.28) render results that are slightly different between them but that are much closer together than those observed for Manacor productions (/a/ in the context of /k/ and /p/). Artà /a/ is found in between the production of Manacor /a/ in the context of /k/ and /a/ in the context of /p/; that is, whereas Artà speakers produce a more central /a/ (which slightly goes up or down depending on the consonant context),
Manacor speakers produce two different allophones of /a/, one exaggeratedly higher /a/ in the context of /k/ and one exaggeratedly lower /a/ in the context of /p/. The fact that the difference for Artà speakers is much smaller than the difference for Manacor speakers suggests that such an effect is simply due to coarticulatory forces for Artà speakers. Table
20 in the next page shows the results for the mixed-effects generalized linear regression model test for the F1 measure.
98
Table 20. Results for the mixed-effects generalized linear regression model test for the F1 measure for group (Artà, Manacor) and consonant context (/k/, /p/). The baseline levels are Artà (for group) and /k/ (for consonant). Non-significant values are marked as “n. s”.
Estimate SE Df t-value Probability
(Intercept) 1.09353 0.04489 37.99 24.358 p < 0.001
Manacor −0.12478 0.06348 37.97 −1.966 n. s.
/p/ 0.06155 0.02946 38.70 −2.089 p < 0.05
Manacor:/p/ 0.24186 0.04142 37.89 5.839 p < 0.001
4.2.2.2. F2 measure
The acoustic F2 data were submitted to a mixed-effects generalized linear regression model. The analysis had a (2) × 2 mixed design, with consonant context (/k/,
/p/) as the within-subjects variable, and with group (Manacor, Artà) as the between- subjects variable. The independent variable was the normalized F2 value. The best model used for this analysis explained 49% of the variance. The results yielded a main effect of consonant (χ2(1) = 46.11; p < 0.001) and a group by consonant context interaction (χ2(1)
= 28.17; p < 0.001). There was no main effect of group (χ2(1) = 1.85; p > 0.05). These results suggest that /a/ is significantly more fronted in the context of /k/ (M = −0.005, sd
= 0.26) than in the context of /p/ (where it is more posterior) (M = −0.41, sd = 0.21) for the Manacor group. This vowel has higher F2 values in the context of /k/ than in the context of /p/. To the extent that F2 is an acoustic correlate of vowel fronting, it can be inferred that /a/ is more fronted in the context of /k/ than in the context of /p/ for the 99
Manacor group. For the Artà group, on the other hand, both consonant /k/ context (M =
−0.19, sd = 0.29) and /p/ context (M = −0.34, sd = 0.25) render similar results, with slightly more fronting of /a/ in the context of /k/ than in the context of /p/. However, this difference is much smaller than the one observed for Manacor /a/ productions. Artà /a/ (in both consonant contexts) is found in between Manacor /a/ in the context of /k/ and /a/ in the context of /p/. Therefore, whereas Artà /a/ is also central in terms of F2, Manacor /a/ presents an exaggeratedly fronted allophone in the context of /k/ and an exaggeratedly posterior allophone in the context of /p/. The fact that the difference for Artà speakers is much smaller than the difference for Manacor speakers suggests that such an effect is simply due to coarticulatory forces for Artà speakers. Table 21 shows the results for the mixed-effects generalized linear regression model test for the F2 measure.
Table 21. Results for the mixed-effects generalized linear regression model test for the F2 measure for group (Artà, Manacor) and consonant context (/k/, /p/). The baseline levels are Artà (for group) and /k/ (for consonant).
Estimate SE Df t-value Probability
(Intercept) −0.19484 0.03531 37.01 −5.518 p < 0.001
Manacor 0.19091 0.04993 37.89 3.824 p < 0.001
/p/ −0.14373 0.03052 38.08 −4.709 p < 0.001
Manacor: /p/ −0.26851 0.04300 37.54 −6.245 p < 0.001
100
Table 22. Crossdialectal difference in Catalan /a/ production per consonant context (/k/,
/p/) for each of the acoustic measures (F1, F2). A check (✔) means there are differences between speakers from Manacor and Artà; a cross (✖) means there are no differences between speakers from Manacor and Artà.
/k/ context /p/ context
F1 ✔ ✖
F2 ✔ ✖
4.2.2.3. Summary of results for /a/ in Catalan
Several analyses examined the effects of consonant context (/k/, /p/) and group
(Manacor, Artà) on F1 an F2 separately. Analyses for F1 data showed that the effects of consonant context differed for the two dialects: /a/ was significantly higher in the context of /k/ and it was significantly lower in the context of /p/ for the Manacor group. The Artà group presented a very slight difference between /a/ in the context of /k/ and /a/ in the context of /p/ when compared to the difference observed in Manacor. Additionally, both variants of /a/ in Artà are very central and lie in between Manacor /a/ in the context of /k/ and Manacor /a/ in the context of /p/. Analyses for F2 data showed differences for /a/ productions between the two groups depending on the consonant context: /a/ was significantly more fronted in the context of /k/ than in the context of /p/ (where it was significantly more posterior) for the Manacor group than it was for the Artà group. Again,
Artà speakers presented a very slight difference between /a/ in the context of /k/ and /a/ in the context of /p/. The difference observed for Artà speakers, when compared to the much 101 larger difference observed in Manacor speakers, can therefore be connected to coarticulation. Differently, it seems that bringing /a/ to a more fronted or to a more posterior place of articulation is an intended gesture for Manacor speakers. To sum up, it is not true that /a/ is generally more fronted in Majorcan Catalan as a whole. Instead, it seems that in Manacor speech, /a/ is more fronted and higher when produced after /k/ than when produced after /p/, in which it is brought to a much lower and a much more posterior place of articulation. In Artà, /a/ is central in general, with a slightly fronted and higher /a/ in the context of /k/ and a slightly more posterior and lower /a/ in the context of
/p/. Artà /a/ lies in between (both in terms of height and frontedness) the two allophones of Manacor /a/. Therefore, the only case in which /a/ is exaggeratedly fronted and higher is in Manacor productions after consonant context /k/. This result, together with the fact that /k/ is palatalized for this same community in certain contexts (/a/ being one of them), allows for a relationship to be established between the two processes (/a/-fronting and /k/- palatalization), but the order of the influence cannot be established. This is further discussed in the following section.
4.3. Summary of Catalan findings
The purpose of the present chapter was twofold: (1) analyze the acoustic characteristics of /k/ production and (2) analyze /a/ production in two different groups of
Majorcan Catalan speakers. For the first part, three acoustic measures were used (in three different analyses): Center of Gravity, Relative Intensity, Variance. These analyses examined the presence of group differences per each vowel context separately, that is, it analyzed whether speakers from Manacor and Artà produced Catalan /k/ differently for 102 each of the vowel contexts separately. All three analyses rendered very similar results: the two groups differed regarding their production of /k/ in vowel contexts /a/, /i/ and /ə/, but not in vowel context /o/. Participants from Manacor showed higher Center of Gravity,
Relative Intensity and Variance values for /k/ in the context of /i/, /a/ and /ə/ (in that order) than participants from Artà did. The two groups did not differ in their production of /k/ in the context of /o/. These results suggest, firstly, that Manacor and Artà speakers present a different distribution (regarding vowel context) of /k/ in their Catalan.
Specifically, Manacor speakers present velar palatalization of /k/ in the context of /i/, /a/ and /ə/ but not in the context of /o/, which indicates that Manacor /k/ is not more fronted in general, but its fronting depends on the vowel context in which it appears. Differently,
Artà speakers present one generally non-fronted /k/, which is slightly fronted in the context of /i/ only. Secondly, these results also indicate the presence of differences between the two dialects in terms of the degree of palatalization observed for /k/.
Specifically, Manacor fronted /k/ observed before /i/, /a/ and /ə/ vowel contexts displays a much more anterior realization than Artà fronted /k/ observed in the /i/ context.
Importantly, /k/ for the /i/ vowel context is much more exaggeratedly fronted in Manacor speakers than it is in Artà speakers. This exaggeration in the articulation of /k/ in
Manacor speech suggests that the palatalizing gesture is no longer the result of coarticulation, but it has been reinterpreted as intended and, therefore, phonologized.
Velar palatalization of /k/ is a phonological process in Manacor speech: Manacor speakers present two different allophones of /k/, a palatalized and a non-palatalized allophone, which they use in complementary distribution. Differently, the fact that the fronting of /k/ is very subtle when compared to the degree of fronting observed for 103
Manacor speakers suggests that fronting simply indicates an effect of vowel-to-consonant regressive coarticulation in Artà speech.
The second block of analyses in this chapter concentrated on the low vowel, /a/.
Two acoustic measures were used (in two different blocks of analyses): F1 and F2. These analyses examined the group differences per each consonant context separately, that is, it examined whether Manacor and Artà speakers produced /a/ with different F1 and F2 values for each consonant context. The results of the F1 analysis showed a difference between the two groups in their production of /a/ depending on consonant context: /a/ was much higher after /k/ and much lower after /p/ for the Manacor group when compared to Artà data. Differently, participants from Artà showed very slight differences between the two consonant contexts, with values that were in between Manacor /a/ in the context of /k/ and in the context of /p/. The results from the F2 analysis showed a difference between the two groups based on consonant context: /a/ was much more fronted in the context of /k/ and much more posterior in the context of /p/ for participants from Manacor, but not for participants from Artà. Participants from Artà presented central F2 values for /a/ in both consonant contexts, with only a slight anterior articulation in the context of /k/ and a slight posterior articulation in the context of /k/.
The difference between consonant contexts was much larger for Manacor speakers than for Artà speakers. To sum up, Manacor participants presented a fronted and high /a/ when produced after /k/ and a posterior and low /a/ when produced after /p/. Artà participants presented a central /a/ with frontedness and height values lying in between the ones observed for Manacor. Context-dependent variation in Artà central /a/ was very subtle when compared to the one presented by Manacor speakers. These results indicate that, 104 first, /a/ is not generally more fronted and higher in Majorcan Catalan in general. For
Manacor speakers, it is exaggeratedly more fronted and higher when produced after /k/ than when it is produced after /p/. For Artà speakers, it is generally central, with a variation that is very slight and due to consonant-to-vowel coarticulation. When putting these results together with the results from the analysis of /k/, we can observe a relationship between the process of velar palatalization and of /a/-fronting. It is only
Manacor speakers that palatalize /k/ in certain contexts (/a/ being one of them) and that significantly front /a/ in the context of /k/. The two processes, therefore, seem to clearly have been related in its inception. However, the fact that only the community of
Manacor, but not the community of Artà (remember that this community only presents a very slight variation of /a/ due to coarticulation), presents evidence for the two processes makes it impossible to determine the order of the influence between the two processes.
There remain, therefore, three possibilities for the origin of the two changes: (a) /k/- palatalization might have triggered /a/-fronting initially; (b) /a/-fronting might have triggered /k/-palatalization; (c) both processes might have begun and developed simultaneously, thus influencing each other. These options are explored in the discussion section below.
Let us now return to the first research question established at the beginning of this chapter, i.e., does Majorcan Catalan maintain a different distribution of /k/ dependent on the vowel context in which it appears? The answer seems to be that it depends on the community. Speakers from Manacor have two different allophones of /k/, that are used in complementary distribution depending on the following vowel: they have a non-fronted allophone of /k/ for back vowel /o/ and a fronted allophone of /k/ for front vowel /i/ and 105 central vowels /a, ə/. Speakers from Artà have one single allophone of /k/ that is used in all vowel contexts (and which is slightly fronted before front vowels). The second research question relates to the acoustic nature of Majorcan Catalan /a/, i.e., is Majorcan
Catalan fronted and high in general or are there acoustic differences between the two communities regarding its quality? The results point towards a clear pattern: whereas
Manacor speech presents a more fronted and higher /a/ in the context of /k/ and a more posterior and lower /a/ in the context of /p/, Artà /a/ is more central in all consonant contexts (with a slight coarticulatory effect). Therefore, the consonant context /k/ is directly related to a more fronted and higher articulation of /a/ for Manacor speakers.
Since both processes (/k/-palatalization and /a/-fronting) are only observed in Manacor
Catalan, but neither one is observed in Artà Catalan, a relationship between both processes can be established, but the order of the influence cannot be determined.
In sum, the analysis of production of /k/ in Majorcan Catalan has indicated that there are, in fact, cross-dialectal differences in both the distribution of /k/-palatalization
(dependent on vowel context) and in the degree of fronting of /k/. First, Manacor speakers have two different allophones of /k/ in their Catalan: a palatalized and a non- palatalized allophone, which they use in complementary distribution. Differently, Artà speakers only present one allophone of /k/ in their Catalan. Regarding the degree of fronting, Manacor fronted /k/ in the context of /i/ has a much more anterior articulation than Artà /k/ in the context of /i/, which is only slightly fronted. This difference in the degree of articulation between the two communities suggests that what is merely the consequence of vowel-to-consonant coarticulation in Artà speech is a phonologized process in Manacor speech. Additionally, the analysis of Catalan /a/ has determined that 106 the low vowel is not generally more fronted and higher in Majorcan Catalan, but it is only like this when produced in the context of /k/ by Manacor speakers: Manacor Catalan /a/ is fronted and high when it is produced after /k/ and it is posterior and low when it is produced after /p/; Artà Catalan /a/ is central in general, lying in between the two allophones of Manacor /a/. Therefore, whereas Manacor speakers present evidence for both processes (velar palatalization and /a/-fronting) Artà speakers do not present evidence of either of them (remember that coarticulation does not necessarily entail palatalization or fronting). The fact that for Manacor speakers /a/-fronting only occurs in the context of /k/ suggests that whereas there is a direct relationship between the two processes, the order of the influence cannot be established.
107
CHAPTER 5
SPANISH
5.1. Introduction
This chapter analyzes the acoustic characteristics of the burst of /k/ in the Spanish spoken in two areas of the island of Majorca (Manacor, Artà). This chapter, therefore, focuses on production in Spanish. The goal of this chapter is to analyze how Catalan-
Spanish bilinguals dominant in Catalan produce the velar stop consonant /k/ in different vowel contexts in their non-dominant language (i.e., in Spanish). Specifically, it analyzes whether Catalan-dominant speakers who manifest the application of a phonological rule leading to the palatalization of /k/ in Catalan (i.e., Manacor speakers) transfer this phonological rule of their Catalan onto their Spanish. Previous literature on consecutive bilingualism suggests that even with an early and extensive exposure to the L2, bilinguals face difficulties to acquire native-like patterns of the L2 (Guion, Flege, & Loftin, 2000;
Fowler et al., 2008; Mack, 1989; Pallier et al., 1997; Piske et al., 2002; Sebastián-Gallés
& Soto-Faraco, 1999; Simonet, 2010, 2011a). Most of the literature on the Catalan-
Spanish bilingual community suggests that the presence of crosslinguistic transfer is an inevitable consequence of acquiring a second language after the native language, even if the acquisition of the L2 has taken place during childhood (Lleó et al., 2009; Ramírez &
Simonet, accepted; Simonet, 2010, 2011a, 2011b). Contrastingly, other studies note that the acquisition of L2 phonemic contrasts can take place, at least, to a certain extent
(Amengual, 2011, 2016; Simonet, 2014). Note that in these later studies, even if the non- native phonemic categories have been acquired, these categories are not produced with 108 identical acoustic-phonetic features to those produced by speakers who are dominant in that language. Transfer from the L1 to the L2 (even if it is transfer of phonetic features, rather than phonological categories) seems to be a consistent consequence of early sequential bilingual acquisition. The present dissertation analyzes whether speakers from
Manacor transfer the phonological rule observed in their Catalan (i.e., phonologization of
/k/ in certain contexts) onto their speech in Spanish. It does so by conducting the following analyses:
1. The acoustic characteristics of /k/ as produced, in Spanish, by two groups of
Catalan speakers are compared. In particular, the acoustics of /k/ as produced by
speakers of a palatalizing dialect (Manacor) are compared with those produced by
speakers of a non-palatalizing dialect (Artà). Various vowel contexts are
examined: /a, i, o/. This analysis addresses the following question: Do speakers
from palatalizing and non-palatalizing Catalan areas differ in the way they
produce /k/ in their second, non-dominant language (Spanish)? Regarding this
question, our hypotheses are as follows: a) Since speakers from Manacor and Artà
do not differ in how they produce Catalan /k/ in the context of /o/, a difference
between these two groups in terms of how they produce Spanish /k/ should not be
expected; b) On the other hand, since Catalan /k/ is palatalized in the context of
both /i/ and /a/ in the speech of the Manacor speakers but not in that of the Artà
speakers, if Catalan-dominant speakers transfer the acoustic characteristics of
their native language to their non-native language, a difference between these two
groups in terms of how they produce Spanish /k/ in the context of non-back
vowels should be expected. On the other hand, if Manacor speakers do not 109
transfer the acoustics of their Catalan /k/ to their Spanish, no difference between
Manacor and Artà speakers, in terms of how they produce Spanish /k/, is
necessarily expected.
2. The following comparison analyzes the acoustic characteristics of /k/ as produced
in Catalan vs. Spanish in the speech of Catalan-dominant speakers of a
palatalizing (Manacor) dialect and a non-palatalizing (Artà) one. Thus, the
acoustics of /k/ in various vowels contexts (/a, i, o/) are analyzed as a function of
the language spoken: Catalan vs. Spanish. This analysis goes beyond the analysis
explained above by addressing directly the following question, which the
preceding analysis could address only indirectly: To what extent do Catalan-
dominant Catalan-Spanish bilingual speakers transfer the acoustic characteristics
of their Catalan /k/ to their Spanish /k/? In particular, do Catalan-dominant
speakers of a palatalizing dialect (Manacor) transfer a phonological rule of their
first, dominant language to their second, non-dominant one? The hypotheses are
as follows: a) A difference between Spanish and Catalan /k/ is not expected for
the non-palatalizing speakers (Artà); b) To the extent that Catalan-dominant
speakers transfer the acoustic characteristics of their native language to their non-
native language (or a phonological rule of their native language to their non-
native language), the prediction is that Manacor speakers will show evidence of
palatalization in both their Catalan and their Spanish /k/ (in the context of /a/ and
/i/, but not /o/), perhaps using a single phonetic category in both languages. On
the other hand, if the phonetics of Catalan /k/ are not transferred to Spanish /k/,
Manacor speakers should use different phonetic categories for /k/ in their two 110
languages: a palatalized version in Catalan and a non-palatalized version in
Spanish.
The following subsection (5.2) presents the results for the Spanish production data. First, the results for the analysis of Spanish /k/ are presented (5.2.1). Second, the results for a comparison between Catalan and Spanish productions are provided (5.2.2).
Finally, the findings for the Spanish data are summarized (5.2.3).
5.2. Results
5.2.1. Analysis of /k/ in Spanish
This section is concerned exclusively with an analysis of the Spanish /k/ production. The acoustic data were analyzed using mixed-effects generalized linear regression models to test the effect of group (Manacor, Artà) and vowel context (/a/, /i/,
/o/) on different acoustic measures (Center of Gravity, Relative Intensity, Variance). For all models, participants were random intercepts (Baayen et al., 2008), and there were random slopes for each vowel context (Barr et al., 2013) in all models. Statistical significance of group and vowel context, as well as group by vowel context interactions, were analyzed using hierarchical partitioning of variance via nested model comparisons.
Pairwise comparisons were analyzed through Simultaneous Tests for General Linear
Hypotheses, by using Tukey Contrasts with adjusted p-levels. The subsequent subsections are divided according to each of the acoustic measures.
111
5.2.1.1. Center of Gravity
The acoustic Center of Gravity data were submitted to a mixed-effects generalized linear regression model, which tests main effects using a nested model comparison. The analysis had a (3) × 2 mixed design, with vowel context (/a/, /i/, /o/) as the within-subjects factor, and with group (Manacor, Artà) as the between-subjects factor.
The best model used for this analysis explained 81% of variance. The results yielded a main effect of vowel context (χ2(2) = 103.74, p < 0.001). There was no main effect of group (χ2(1) = 1.21, p > 0.05) and no group by vowel context interaction (χ2(2) = 3.57, p
> 0.05). There is no motivation to check effects of group for the three vowel contexts separately. Table 23 shows the results for the mixed-effects generalized linear regression model test for the Center of Gravity measure for vowel contexts (/a/, /i/, /o/) excluding group and group by vowel interactions. Figure 5 plots the Center of Gravity values for group (Manacor, Artà) and vowel context (/a/, /i/, /o/). Table (e), in Appendix 7, shows mean and standard deviation values for the Center of Gravity measure for group
(Manacor, Artà) and vowel context (/a/, /i/, /o/).
Table 23. Results for the mixed-effects generalized linear regression model test for the
Center of Gravity measure for vowel context (/a/, /i/, /o/) for Spanish productions only.
The baseline level is vowel context /a/. Group and interactions are excluded.
Estimate SE Df t-value Probability (Intercept) 1574.58 52.25 39.98 30.13 p < 0.001 /i/ 1368.31 87.97 40.02 15.55 p < 0.001 /o/ −577.00 41.19 39.94 −14.01 p < 0.001
112
Figure 5. Center of Gravity (Hz) for group (Artà, Manacor) and vowel context (/a/, /i/,
/o/) for Spanish data only
6000
4000 Group Art`a Manacor
2000 Center of gravity
0 /a/ /i/ /o/ Vowel
Summing up the results for the Center of Gravity measure for the Spanish data, these results report no differences between the two groups of speakers in the production of /k/: both Artà and Manacor produce a similar /k/ in the context of /a/, a similar one in the context of /i/ and a similar one in the context /o/ in Spanish. The results, however, do suggest differences in the phonetic quality of /k/ as produced in different vowel contexts, which indicates that /k/ presents variability in Spanish depending on the vowel context in which it is produced. Importantly, /k/ in the context of /i/ is produced with higher Center of Gravity values than /k/ in the other two vowel contexts. These results suggest that there is a vowel-to-consonant coarticulation effect for /k/ for both groups of speakers in
Spanish. These coarticulatory effects are of comparable size across the two dialects.
113
5.2.1.2. Relative Intensity
The acoustic Relative Intensity data were submitted to a mixed-effects generalized linear regression model, which tests main effects using a nested model comparison. The analysis had a (3) × 2 mixed design, with vowel context (/a/, /i/, /o/) as the within-subjects factor, and with group (Manacor, Artà) as the between-subjects factor.
The best model used for this analysis explained 47% of variance. The results yielded a main effect of group (χ2(1) = 4.97, p < 0.03) and a main effect of vowel context (χ2(2) =
14.45, p < 0.001). There was no interaction between the two factors (χ2(2) = 4.31, p >
0.05). Table 24 shows the results for the mixed-effects generalized linear regression model test for the Relative Intensity measure for vowel contexts (/a/, /i/, /o/) and group
(Manacor, Artà). Figure 6 plots the Relative intensity values for group (Manacor, Artà) and vowel context (/a/, /i/, /o/). Table (f), in Appendix 7, shows mean and standard deviation values for the Relative Intensity measure for group (Manacor, Artà) and vowel context (/a/, /i/, /o/).
Table 24. Relative Intensity results for the mixed-effects generalized linear regression model test for group (Artà, Manacor) and vowel context (/a/, /i/, /o/) for Spanish productions only. The baseline levels are Artà (for group) and /a/ (for vowel context)
Estimate SE Df t-value Probability
(Intercept) 9.44 0.49 48.00 19.23 p < 0.001
Manacor 1.49 0.61 40.00 2.42 p < 0.001
/i/ 1.30 0.35 40.03 3.66 p < 0.001
/o/ 0.83 0.26 40.08 3.13 p < 0.001 114
Figure 6. Relative Intensity for group (Artà, Manacor) and vowel context (/a/, /i/, /o/) for
Spanish data only
20
10 Group Art`a Manacor 0 Relative intensity
-10
/a/ /i/ /o/ Vowel
The main effect for the group factor indicates that there are overall higher values in the Manacor productions than in the Artà productions. The main effect of vowel context indicates that overall Relative Intensity values (across all three vowel contexts) were slightly higher/greater for Manacor than they were for Artà speakers. There is no motivation to check effects of group for the three vowel contexts separately.
Summing up the results for the Relative Intensity measure for the Spanish data, there were significant overall differences between groups: Manacor productions had higher Relative Intensity values in general (for all three vowel contexts together) than
Artà speakers did. Note that, however, the Relative Intensity measure is not as reliable as the Center of Gravity measure, since it can measure place of articulation in a more indirect and obscure matter. Note also that, whereas the model with Center of Gravity as response predicted 81% of the variance, this model with Relative Intensity as response only predicts 43% of the variance. Therefore, the results for the Center of Gravity metric 115 are more reliable, as the fit model (the explained variance) is much higher. For the present model (with Relative Intensity as the metric), much variance remains unaccounted for, thus making any significant findings somewhat less relevant. This will be discussed further in the summary subsection below.
5.2.1.3. Variance
Spectral Variance data were submitted to a mixed-effects generalized linear regression model, which tests main effects using a nested model comparison. The analysis had a (3) × 2 mixed design, with vowel context (/a/, /i/, /o/) as the within- subjects factor, and with group (Manacor, Artà) as the between-subjects factor. The best model used for this analysis explained 72% of variance. The results yielded a main effect of vowel context (χ2(3) = 93.88, p < 0.001). There was no main effect of group (χ2(1) =
1.13, p > 0.05) and no group by vowel context interaction (χ2(3) = 3.93, p > 0.05). Table
25 shows the results for the mixed-effects generalized linear regression model test for the
Variance measure for vowel contexts (/a/, /i/, /o/) excluding group and group by vowel context interactions. Figure 7 plots the Variance values for group (Manacor, Artà) and vowel context (/a/, /i/, /o/). Table (g), in Appendix 7, shows Variance mean and standard deviation values for group (Manacor, Artà) and vowel context (/a/, /i/, /o/).
The results for the Variance measure for the Spanish data suggest that the presence of differences is only based on vowel context, but not based on dialect. This means that both groups (Manacor and Artà) produce /k/ similarly between them and that
/k/ only changes depending on the vowel context in which it is produced. Production of
/k/ only varied as a function of vowel context, since, as Figure 7 suggests, vowel context 116
/i/ triggers higher Variance values (higher /k/ fronting) than the other two vowel contexts do. This fronting of /k/ in the context of /i/ can be interpreted as evidence of coarticulation. Importantly, the effects of coarticulation are comparable, in terms of their size, across the two dialects. Additionally, note that the fit of this model is relatively high, accounting for 72% of the variance. This fit is not as high as that for Center of Gravity as metric (81%), but it is much higher than that with Relative Intensity as metric (43%).
Table 25. Results for the mixed-effects generalized linear regression model test for the
Variance measure for vowel contexts /a/, /i/, /o/ for Spanish productions only. The baseline level is vowel context /a/. Group and interactions are excluded.
Estimate SE Df t-value Probability (Intercept) 1367.21 50.14 40.01 27.26 p < 0.001 /i/ 606.50 67.99 40.02 8.92 p < 0.001 /o/ −487.45 36.26 39.98 −13.44 p < 0.001
Figure 7. Variance values for group (Artà, Manacor) and vowel context (/a/, /i/, /o/) for
Spanish data only
4000
3000
Group 2000 Artà Manacor Variance
1000
0
/a/ /i/ /o/ Vowel 117
5.2.1.4. Summary of results for /k/ productions in Spanish
An analysis was conducted for each of the three acoustic measures (Center of
Gravity, Relative Intensity, Variance) regarding the burst of word-initial /k/ as produced in different vowel contexts (/a/, /i/, /o/) and by two groups (Manacor, Artà). In other words, these analyses looked at potential effects of group on the production of /k/ in different vowel contexts. For the Center of Gravity and Variance measures, the results were similar: /k/ production did not vary as function of group: both Manacor and Artà speakers produced similar values for /k/ in the context of /a/, similar ones in the context of /i/ and similar ones in the context of /o/. For the Relative Intensity measure, however,
Manacor /k/ productions seemed to show higher values in general (when compared to
Artà productions). Table 26 summarizes the reported differences between dialects for each of the vowel contexts within each acoustic measure.
Table 26. Crossdialectal difference in Spanish per vowel context (/a/, /i/, /o/) for each of the acoustic measures (Center of Gravity, Relative Intensity, Variance). A check (✔) means there are differences between speakers from Manacor and Artà; a cross (✗) means there are no differences between speakers from Manacor and Artà.
/a/ /i/ /o/
Center of Gravity ✗ ✗ ✗
Relative Intensity ✔ (higher values in general for Manacor)
Variance ✗ ✗ ✗
118
Note that the difference reported for the Relative Intensity measure should be taken with care, since this difference is much smaller than the between-group differences in the Catalan data reported in Chapter 4. This statistical finding, which reports a crossdialectal difference for these two vowel contexts, could actually be due to a Type-I error. If this were the case, the statistical model might be reporting a significant difference when, in reality, there is no such difference. In any case, as stated in the
Methods section of this dissertation, even though all three acoustic measures are indices of place of articulation, the most accurate measure in the indication of fronting is Center of Gravity. As seen in these results, analyses for this measure indicate that there is no difference between the two communities for the production of /k/ in any of the three vowel contexts. This finding is supported by the Variance acoustic measure. It is only one of the three acoustic measures, the Relative Intensity measure, which is an indirect and less well-known measure of place of articulation, that provides contradicting findings.
Additionally, the Relative Intensity model has a low fit compared to the other two models: whereas the models with Center of Gravity and Variance as responses predicted
81% and 72% of the variance, respectively, the model with Relative Intensity as response accounted for only 43% of the variance, making this a less reliable model. Therefore, as stated in the Methods section, since the Center of Gravity measure detects no difference between communities and is supported by the Variance measure with higher fits (than the one for the Relative Intensity model), these results will be interpreted as there being no difference between communities in their production of Spanish /k/.
Additionally, for the Center of Gravity and the Variance measures, there were main effects of vowel context, suggesting that /k/ in the context of /i/ is produced with 119 higher values than /k/ in the other vowel contexts; and that /k/ in the context of /o/ seems to be produced with lower values than /k/ in the context of /a/. Importantly, these findings on /k/ when produced in vowel context /i/ suggest that there is in fact some degree of /k/- fronting in this vowel context. This is expected, even for Spanish, since this is occurring for the vowel context that triggers fronting of the previous consonant due to coarticulation. This fronting in the Manacor and Artà Spanish data, however, when compared to the fronting observed in the Manacor Catalan data is clearly smaller in size:
/k/-fronting for the /i/ context is much more exaggerated in Catalan than in Spanish. This difference suggests that whether this effect is a phonologized rule in Catalan, it is simply the result of coarticulation in Spanish.
When looking at these results and comparing them to the Catalan results observed in the previous chapter, the logical conclusion seems to be that Manacor speakers have a different contextual distribution and a different degree of fronting for /k/ for each of their two languages: whereas they seem to considerably front /k/ before contexts /a, i, ə/ but not /o/ in Catalan, they only show effects of vowel-to-consonant coarticulation in their
Spanish. Even if they show more fronted articulations of /k/ for /i, a/ vowel contexts than for /o/ in their Spanish, this fronting does not differ from the one observed for Artà speakers. Therefore, whereas Manacor speakers palatalize in their Catalan (in various contexts), they do not do it in Spanish. In Spanish, by not differing from the Artà group, they simply show effects of regressive vowel-to-consonant coarticulation. If Manacor speakers differ from the Artà speakers in Catalan but not in Spanish, it will indicate that they actually have two different phonological systems for their two different languages and that they do not transfer the phonological feature observed in their Catalan (i.e., /k/- 120 palatalization) onto their Spanish. Even though the figures suggest these findings, the statistical models in this section were not able to address this question (i.e., comparison of /k/ production between the two languages of the speakers). The following subsection provides an analysis using a design that answers this question via a within-subject comparison analysis.
5.2.2. Comparing Catalan and Spanish /k/
In the preceding chapter, it was reported that the production of /k/ in Catalan differed for the two groups of participants: participants from Manacor had a significantly different /k/ than participants from Artà did in the context of /a/, /i/ and /ə/ (but not for
/o/) for all three acoustic measures analyzed (Center of Gravity, Relative Intensity and
Variance). The preceding section of this same chapter reported that the production of /k/ in Spanish by the same two groups of participants did not differ between groups for two of the acoustic measures (Center of Gravity and Variance). The only difference based on group was found for the Relative Intensity measure, since participants from Manacor showed higher values in general (i.e., across vowel contexts) when compared to Artà productions. It seems, therefore, that, in general, the two groups of speakers produced similar velar consonants that only differed in terms of vowel context. These findings suggest that these bilinguals are maintaining two separate distributions for their two languages. Importantly, Manacor speakers do not seem to be transferring the phonological pattern of Catalan onto their Spanish. Any comparison between languages, however, was not directly answered by the statistical models presented in the previous subsection. Thus, conclusions such as the one just provided are not warranted by the 121 results presented so far. (For instance, it is possible that even the Artà speakers, and not only the Manacor speakers, maintain separate phonetic categories for their Catalan and
Spanish /k/.) The following subsection reports on within-speaker analyses that compare the Catalan and Spanish data of speakers. These analyses were conducted separately for the two dialectal groups. This comparison was conducted in order to see whether there is an actual transfer of phonological patterns or whether bilinguals maintain separate patterns for their L1 and L2. The following subsections are divided by group of speakers
(Manacor productions in Catalan and Spanish compared first, Artà productions compared after).
5.2.2.1. Manacor (Catalan-Spanish comparison)
The present section reports on an analysis of /k/ productions by participants from
Manacor for both languages (Catalan, Spanish). Comparisons are conducted for all three acoustic measures (Center of Gravity, Relative Intensity, Variance). The acoustic data were analyzed using mixed-effects generalized linear regression models to test the effect of language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) on the three acoustic measures. Notice that the only vowel contexts compared are /a/, /i/ and /o/ (and not /ə/) because these were the three vowel contexts shared by both languages (i.e., Spanish does not have /ə/). For all models, participants were random intercepts (Baayen et al., 2008), with random slopes for each vowel context (Barr et al., 2013). Statistical significance of language, vowel context and language by vowel context interactions were assessed using hierarchical partitioning of variance via nested model comparisons. Simultaneous Tests for General Linear Hypotheses analyzed pairwise comparisons using Tukey Contrasts 122 with adjusted p-levels. The following subsections report on the results per acoustic measure separately.
5.2.2.1.1. Center of Gravity
The acoustic Center of Gravity data from the Manacor group were submitted to a mixed-effects generalized linear regression model. The analysis had a (2) × (3), with language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) as within-subject factors. The best model used for this analysis explained 81% of the variance. The results yielded a main effect of vowel context (χ2(2) = 31.98, p < 0.001), and a language by vowel context interaction (χ2(2) = 34.18, p < 0.001). There was no main effect of language (χ2(1) = 0.13, p > 0.05). Table 27 shows the results for the mixed-effects generalized linear regression model test for the Center of Gravity measure for the Manacor group. Figure 8 plots the
Center of Gravity values for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/).
Table (h), in Appendix 8, shows mean and standard deviation values for the Center of
Gravity measure for the Manacor group.
In order to further explore the language by vowel context interaction, several pairwise comparisons were conducted, which compared language productions per each vowel context separately. For vowel context /a/, the results showed that the Center of
Gravity measure was 1196.25 units (± 33.33 se) higher in Catalan than in Spanish (t =
−35.90, p < 0.001); it was 287.21 units (± −41.53 se) higher in Catalan than in Spanish (t
= −6.92, p < 0.001) for vowel context /i/; and it did not vary as a function of language (t =
0.30, p > 0.05) for vowel context /o/. 123
Table 27. Center of Gravity results for the mixed-effects generalized linear regression model test for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Manacor productions. The baseline levels are Catalan (for language) and /a/ (for vowel context)
Estimate SE Df t-value Probability
(Intercept) 2740.46 119.42 20.02 22.94 p < 0.001
Spanish −1193.89 128.75 20.01 −9.27 p < 0.001
/i/ 594.70 94.32 19.95 6.30 p < 0.001
/o/ −1728.72 122.23 20.06 −14.14 p < 0.001
Spanish: /i/ 911.01 119.93 20.00 7.59 p < 0.001
Spanish: /o/ 1199.05 130.18 20.14 9.21 p < 0.001
Figure 8. Center of Gravity (Hz) for language (Catalan, Spanish) and vowel context (/a/,
/i/, /o/) for Manacor group only
6000
4000 Manacor Catalan Spanish
Center of2000 gravity
0 /a/ /i/ /o/ Vowel
124
These results for the Center of Gravity measure for the Manacor group suggest that there were significant differences between languages for vowel contexts /a/ and /i/, but /k/ did not vary in the context of /o/ as a function of language. It seems that Manacor speakers are resorting to a different phonological pattern depending on the language they are speaking. In Catalan, they are producing a fronted allophone of /k/ in vowel contexts
/a/, /i/ (and /ə/) and a non-fronted allophone of /k/ before /o/. In Spanish, they are producing a non-fronted allophone of /k/ in general, which is slightly fronted for vowel context /i/ only. These differences between the two languages of Manacor speakers have been corroborated by this within-subject statistical model. The results suggest, therefore, that any vowel context effects on Spanish /k/ are due to coarticulation while the vowel context effects on Catalan /k/ are due to the application of a phonological rule: while
Manacor speakers palatalize in Catalan, they do not palatalize in Spanish. Thus, they do not transfer a phonological rule from Catalan onto Spanish.
5.2.2.1.2. Relative Intensity
The acoustic Relative Intensity data from the Manacor group were submitted to a mixed-effects generalized linear regression model. The analysis had a (2) × (3), with language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) as within-subject factors. The best model used for this analysis explained 49% of the variance. The results yielded a main effect of language (χ2(1) = 4.55, p < 0.05), and a language by vowel interaction
(χ2(2) = 7.96, p < 0.05). There was no main effect of vowel context (χ2(2) = 2.68, p >
0.05). Table 28 shows the results for the mixed-effects generalized linear regression model test for the Relative Intensity measure for the Manacor group. Figure 9 plots the 125
Relative Intensity values for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/).
Table (i), in Appendix 8, shows mean and standard deviation values for the Relative
Intensity measure for the Manacor group.
In order to further explore the language by vowel-context interaction, three pairwise comparisons were conducted, which compared language productions per each vowel context separately. The data containing only vowel context /a/ indicated that the
Relative Intensity measure was −2.25 units (± 0.66 se) higher in Catalan than in Spanish
(t = −3.41, p < 0.003); for vowel context /i/, it was −1.53 units (± −0.57 se) higher (t =
−2.68, p < 0.02); and for vowel context /o/, it did not vary as a function of language (t =
−1.19, p > 0.05).
Table 28. Results for the mixed-effects generalized linear regression model test for the
Relative Intensity measure for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Manacor productions only. The baseline levels are Catalan (for language) and /a/ (for vowel). Non-significant values are marked as “n. s.”
Estimate SE Df t-value Probability
(Intercept) 13.23 0.68 20.00 19.30 p < 0.001
Spanish −2.24 0.66 20.00 −3.40 p < 0.05
/i/ 0.59 0.45 19.96 1.30 n.s.
/o/ −1.60 0.73 20.01 −2.20 p < 0.05
Spanish: /i/ 0.71 0.57 20.00 1.24 n.s.
Spanish: /o/ 1.93 0.64 20.02 3.02 p < 0.05
126
Figure 9. Relative Intensity for language (Catalan, Spanish) and vowel context (/a/, /i/,
/o/) for the Manacor group
30
20
Manacor Catalan Spanish 10 Relative intensity
0
/a/ /i/ /o/ Vowel
These results for the Relative Intensity measure for the Manacor group support the findings reported for the Center of Gravity measure: Manacor speakers have a different pattern of /k/ production for each one of their languages. In Catalan, they produce a more fronted /k/ for the /a/ and /i/ vowel contexts, and a less fronted /k/ for the
/o/ vowel context. In Spanish, they produce /k/ with similar phonetic qualities independently of position (even though their /i/ is slightly more fronted due to coarticulation).
5.2.2.1.3. Variance
The acoustic Variance data from the Manacor group were submitted to a mixed- effects generalized linear regression model. The analysis had a (2) × (3), with language 127
(Catalan, Spanish) and vowel context (/a/, /i/, /o/) as within-subject factors. The best model used for this analysis explained 75% of the variance. The results yielded a main effect of vowel context (χ2(2) = 20.40, p < 0.001), and a language by vowel context interaction (χ2(2) = 29.11, p < 0.001). There was no main effect of language (χ2(1) = 3.37, p > 0.05). Table 29 shows the results for the mixed-effects generalized linear regression model test for the Variance measure for the Manacor group. Figure 10 plots the Variance values for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/). Table (j), in
Appendix 8, shows mean and standard deviation values for the Variance measure for the
Manacor group.
In order to further explore the language by vowel interaction, several pairwise comparisons were conducted, which compared language productions per each vowel context separately: for vowel context /a/, the Variance measure was −683.74 units (±
103.18 se) higher in Catalan than in Spanish (t = −6.63, p < 0.001); and it did not vary as a function of language for vowel context /i/ (t = −1.73, p > 0.05) or for vowel context /o/
(t = −1.86, p > 0.05).
The Variance measure for the Manacor group suggests that there were differences between languages only for vowel context /a/, but /i/ and /o/ did not lead to differences as a function of language in the acoustic characteristics of /k/. The findings for the Variance measure do not fully corroborate the findings indicated by the Center of Gravity and the
Relative Intensity measure: the previous two measures indicated that there were differences between the two languages of Manacor speakers for /k/ in the context of /i/ and /a/, but not /o/. Regarding Variance, it was found that the /i/ context did not lead to any language differences. 128
Table 29. Results for the mixed-effects generalized linear regression model test for the
Variance measure for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for
Manacor productions only. The baseline levels are Catalan (for language) and /a/ (for vowel). Non-significant values are marked as “n. s.”
Estimate SE Df t-value Probability
(Intercept) 2014.42 64.32 20.00 31.31 p < 0.001
Spanish −684.46 103.22 19.99 −6.62 p < 0.001
/i/ 87.87 53.29 19.95 1.64 n.s.
/o/ −1177.23 94.84 20.02 −12.41 p < 0.001
Spanish: /i/ 594.98 113.45 20.00 5.24 p < 0.001
Spanish: /o/ 756.82 96.25 20.02 7.86 p < 0.001
Figure 10. Variance for language (Catalan, Spanish) and vowel context (/a, i, o/) for
Manacor productions
4000
3000
Manacor Catalan Spanish Variance 2000
1000
0
/a/ /i/ /o/ Vowel 129
5.2.2.1.4. Summary: Manacor Catalan and Spanish /k/
For each acoustic measure separately (Center of Gravity, Relative Intensity,
Variance), an analysis was conducted in order to examine potential effects of language on the production of /k/ in different vowel contexts (/a/, /i/, /o/). The analyses for the Center of Gravity measure and the Relative Intensity measure rendered similar results: /k/ productions were significantly different between the two languages for /k/ in the context of /a/ and /i/, but not for /k/ in the context of /o/. The analysis for the Variance measure showed that /k/ productions were significantly different between the two languages only for vowel context /a/, but not for /i/ or /o/. Table 30 shows a summary of these results for the Manacor production.
Table 30. Crosslinguistic difference per vowel context (/a/, /i/, /o/) for each of the acoustic measures (Center of Gravity, Relative Intensity, Variance) for Manacor productions only. A check (✔) means there are differences between languages; a cross (✗) means there are no differences between languages.
/a/ /i/ /o/
Center of Gravity ✔ ✔ ✗
Relative Intensity ✔ ✔ ✗
Variance ✔ ✗ ✗
These results suggest that, for two of the three acoustic measures (Center of
Gravity and Relative Intensity), Manacor speakers maintain two separate distributions for vowel contexts /a/ and /i/, but not for vowel context /o/. If we consider, as has been 130 established in the methods chapter, that the Center of Gravity measure is the clearest indicator of place of articulation, it seems that Manacor speakers maintain two separate phonological distributions (one per language). Importantly, this finding is supported by another acoustic measure, Relative Intensity. In Manacor Catalan, speakers present a fronted /k/ for vowel contexts /a, i, ə/ and a less fronted /k/ in the context of /o/. In their
Spanish, they present a non-fronted /k/ in general, which is slightly fronted before producing /i/ as a consequence of coarticulation. These findings suggest that there is no evidence of transfer of the Catalan phonological rule onto their Spanish. As a final note, the findings reported for the Variance measure can be the result of a Type II error, by means of which the statistical model does not capture a real difference. Since we consider the Center of Gravity measure to be the most reliable cue for place of articulation, and since this measure (and a second one, Relative Intensity) indicate the existence of a phonetic difference between languages, it will be interpreted as there actually being crosslinguistic differences in the quality of /k/ for both the /a/ and the /i/ contexts for
Manacor speakers.
5.2.2.2. Artà (Catalan-Spanish comparison)
The present section reports on an analysis of /k/ productions by participants from
Artà for both languages (Catalan, Spanish). Comparisons are conducted for all three acoustic measures (Center of Gravity, Relative Intensity, Variance). The acoustic data were analyzed using mixed-effects generalized linear regression models to test the effect of language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) on the three acoustic measures. The same statistical models that were used for the previous analyses were used 131 for this analysis. The following subsections report on the results per acoustic measure separately.
5.2.2.2.1. Center of Gravity
The acoustic Center of Gravity data for the Artà group were submitted to a mixed- effects generalized linear regression model. The analysis had a (2) × (3), with language
(Catalan, Spanish) and vowel context (/a/, /i/, /o/) as within-subject factors. The best model used for this analysis explained 72% of the variance. The results yielded a main effect of vowel context (χ2(2) = 3989.52, p < 0.001). There was no main effect of language (χ2(1) = 53.39, p > 0.05), and no language by vowel interaction (χ2(2) = 3.01, p
> 0.05). Table 31 shows the results for the mixed-effects generalized linear regression model test for the Center of Gravity measure for the Artà group. Figure 11 plots the
Center of Gravity values for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/).
Table (k), in Appendix 8, shows mean and standard deviation values for the Center of
Gravity measure for the Artà group. Since there is no interaction between language and vowel context, there is no justification to conduct any further pairwise comparisons.
The results for the Center of Gravity measure for the Artà group suggest that there were no significant differences between languages for any of the vowel contexts.
Consonant /k/ had similar Center of Gravity values in both languages for the Artà group:
Artà speakers produce /k/ with similar values in their two languages. Additionally, Artà speakers produced /k/ in the context of /i/ more fronted than in the other two contexts in both of their languages, due to coarticulation. We know this is due to coarticulation when a comparison is made between the fronting of /k/ in this context for Artà speakers in any 132 of their two languages and the fronting of /k/ for /i/ in Manacor Catalan, since the latter is much more fronted in this context than the former.
Table 31. Results for the mixed-effects generalized linear regression model test for the
Center of Gravity measure for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Artà productions only. The baseline levels are Catalan (for language) and /a/ (for vowel context). Non-significant values are marked as “n. s.”
Estimate SE Df t-value Probability
(Intercept) 1708.36 74.99 22.00 22.78 p < 0.001
Spanish −106.67 45.66 39.00 −2.33 p < 0.05
/i/ 1098.62 30.06 3510.00 36.54 p < 0.001
/o/ −698.93 30.05 3510.00 −23.2 p < 0.001
Spanish: /i/ 132.99 42.32 3510.00 3.14 p < 0.05
Spanish: /o/ 75.45 42.28 3510.00 1.78 n.s.
Figure 11. Center of Gravity values for language (Catalan, Spanish) and vowel context
(/a, i, o/) for Artà productions only
6000
4000 Art´a Catalan Spanish
Center of2000 gravity
0 /a/ /i/ /o/ Vowel 133
5.2.2.2.2. Relative Intensity
The acoustic Relative Intensity data for the Artà group were submitted to a mixed-effects generalized linear regression model. The analysis had a (2) × (3), with language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) as within-subject factors. The best model used for this analysis explained 45% of the variance. The results yielded a main effect of vowel context (χ2(2) = 7.15, p < 0.05) and a language by vowel context interaction (χ2(2) = 8.55, p < 0.05). There was no main effect of language (χ2(1) = 0.09, p
> 0.05). Table 32 shows the results for the mixed-effects generalized linear regression model test for the Relative Intensity measure for the Artà group. Figure 12 plots the
Relative Intensity values for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for the Artà group. Table (l), in Appendix 8, shows mean and standard deviation values for the Relative Intensity measure for the Artà group.
In order to further explore the language by vowel context interaction, several pairwise comparisons were conducted, comparing language productions per each vowel context separately. The Relative Intensity measure did not vary as a function of language for vowel contexts /a/ (t = −2.04, p > 0.05), /i/ (t = 1.60, p > 0.05), or /o/ (t = 0.32, p >
0.05). The interaction captured by the omnibus model, which, upon running post-hoc analyses, is found to be uninterpretable, may have come about by the difference in the direction of the (negligible) effects of language for the /i/ and /a/ vowel contexts.
These results for the Relative Intensity measure for the Artà group indicate that there was no effect of language on any of the vowel contexts. Consonant /k/ had equal crosslinguistic Relative Intensity values for all three vowel contexts for the Artà group: 134
Artà speakers produce a similar Spanish and Catalan /k/ before /a/, a similar Spanish and
Catalan /k/ before /i/ and a similar Spanish and Catalan /k/ before /o/.
Table 32. Results for the mixed-effects generalized linear regression model test for the
Relative Intensity measure for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Artà productions only. The baseline levels are Catalan (for language) and /a/ (for vowel context). Non-significant values are marked as “n. s.”
Estimate SE Df t-value Probability
(Intercept) 10.05 0.55 20.01 18.56 p < 0.001
Spanish −0.65 0.32 20.01 −2.04 n.s.
/i/ 0.16 0.60 19.98 0.26 n.s.
/o/ 0.59 0.39 19.95 1.50 n.s.
Spanish: /i/ 1.13 0.36 20.00 3.18 p < 0.05
Spanish: /o/ 0.75 0.32 19.99 2.32 p < 0.05
Figure 12. Relative Intensity results for language (Catalan, Spanish) and vowel context
(/a, i, o/) for Artà productions only
20
10 Art´a Catalan Spanish 0 Relative intensity
-10
/a/ /i/ /o/ Vowel 135
5.2.2.2.3. Variance
The acoustic Variance data from the Artà group were submitted to a mixed-effects generalized linear regression model. The analysis had a (2) × (3), with language (Catalan,
Spanish) and vowel context (/a/, /i/, /o/) as within-subject factors. The best model used for this analysis explained 69% of the variance. The results yielded a main effect of vowel context (χ2(2) = 48.36, p < 0.001), and a language by vowel context interaction
(χ2(2) = 9.72, p < 0.01). There was no main effect of language (χ2(1) = 0.15, p > 0.05).
Table 33 shows the results for the mixed-effects generalized linear regression model test for the Variance measure for the Artà group. Figure 13 plots the Variance values for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for the Artà group. Table (m), in Appendix 8, shows Variance mean and standard deviation values for the Artà group.
Table 33. Results for the mixed-effects generalized linear regression model test for the
Variance measure for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Artà productions only. The baseline levels are Catalan (for language) and /a/ (for vowel context). Non-significant values are marked as “n. s.”
Estimate SE Df t-value Probability
(Intercept) 1429.42 57.84 20.00 24.71 p < 0.001
Spanish −25.90 49.06 19.94 −0.52 n.s.
/i/ 388.11 77.77 19.98 4.99 p < 0.001
/o/ −605.78 64.89 20.02 −9.33 p < 0.001
Spanish: /i/ 141.99 42.40 19.97 3.35 p < 0.05
Spanish: /o/ 51.19 51.66 20.04 0.99 n.s. 136
Figure 13. Variance for language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for the Artà group
3000
2000 Artà Catalan Spanish Variance
1000
/a/ /i/ /o/ Vowel
In order to further explore the language by vowel context interaction, several pairwise comparisons were conducted, which compared language productions per each vowel context separately. The first test compared language productions for a subset of data containing vowel context /a/ only. The results showed that the Variance measure did not vary as a function of language for vowel context /a/ (t = −0.53, p > 0.05) or for vowel context /o/ (t = 1.09, p > 0.05). It did vary, however, for vowel context /i/, which was
115.93 units (± 39.70 se) higher in Catalan than in Spanish (t = 2.92, p < 0.01).
Summing up the results for the Variance measure for the Artà group, there were no crosslinguistic differences for vowel contexts /a/ and /o/: Artà speakers produced /k/ in 137 the context of /a/ similarly for both languages, and /k/ in the context of /o/ with similar crosslinguistic values too. The only crosslinguistic difference reported was for Artà /k/ in vowel context /i/ for the Variance measure: /k/ in the context of /i/ presented higher values for the Spanish data than for the Catalan data. Note that the direction of this effect is in the opposite direction from what would be expected if there was any phonetic transfer from Catalan into Spanish: a transfer effect would be observed by a larger degree of palatalization in Catalan than in Spanish (even though the analysis for Artà Catalan productions showed that there was no palatalization for Artà Catalan /k/). This finding seems to suggest that, again, this statistical difference might be found due to a Type I error. As stated in the methods section, a difference between communities or languages
(languages for the current analysis) will be interpreted as such only when at least two out of the three acoustic measures detect such a statistical difference. The current analysis for the Artà crosslinguistic data does not follow this condition since only the Variance measure leads to statistical differences, whereas the other two measures (Center of
Gravity and Relative Intensity) determined that Artà speakers produced similar values for
/k/ between their two languages for all three vowel contexts. Thus, the evidence cannot be used to claim that there is a language difference for Artà productions as any effects are tenuous at best (found in only one metric).
5.2.2.2.4. Summary: Artà Catalan and Spanish /k/
The results of the acoustic properties of the burst of /k/ were examined by means of the Center of Gravity, Relative Intensity, and Variance measures. The potential effects of language on the production of /k/ in different vowel contexts were analyzed. The 138 analyses for the Center of Gravity measure and the Relative Intensity measure rendered similar results: /k/ productions were not significantly different between the two languages for any of the three vowel contexts. The analysis for the Variance measure showed that
/k/ productions were significantly different between the two languages only for vowel context /i/, but not for /a/ or /o/. Table 34 summarizes the findings for the Spanish-
Catalan within-speaker comparison for the Artà group.
Table 34. Artà crosslinguistic difference per vowel context (/a/, /i/, /o/) for each of the acoustic measures (Center of Gravity, Relative Intensity, Variance). A check (✔) means there are differences between languages; a cross (✗) means there are no differences between languages.
/a/ /i/ /o/
Center of Gravity ✗ ✗ ✗
Relative Intensity ✗ ✗ ✗
Variance ✗ ✔ ✗
Note that the findings for the /i/ context in the Variance measure suggest something different from what is indicated by the more reliable measure, Center of
Gravity (and supported by the Relative Intensity measure). In the Methods section, it was determined that a statistical difference in /k/ production between languages would be determined only when at least two out of the three acoustic measures would indicate the presence of such a difference. Since the two out of the three measures indicate a lack of difference for the /i/ context, the finding for the Variance measure will be not taken into 139 consideration. The results for the Artà group, therefore, suggest that these speakers do not present differences in their production of /k/ between their two languages. In other words,
Artà speakers produce the same /k/ independently of language: they produce a non- fronted /k/ in general for both languages, which is slightly fronted only when it is produced in the vowel context /i/. The fact that this fronting is very subtle suggests that it is simply due to coarticulation. Note that this fronting is much less exaggerated than the fronting observed in the Catalan of Manacor speakers for vowel contexts /a/, /i/, /ə/. Artà and Manacor speakers, thus, differ in their crosslinguistic distribution of /k/: whereas
Manacor speakers clearly maintained two separate distributions –one per language– of /k/ depending on the vowel context (in the context of /i/ and /a/, but not /o/), Artà speakers maintain a similar acoustic distribution of /k/ across languages (or at least, show no evidence of differences in /k/ production between the two languages).
5.3. Summary of Spanish findings
The purpose of the present chapter was to analyze whether Catalan-Spanish bilinguals who are dominant in Catalan transferred the phonological/phonetic features from their dominant language onto their non-dominant language. Specifically, it examined whether Manacor speakers transferred the phonological rule regarding velar palatalization observed in their Catalan onto their speech in Spanish. This was analyzed in terms of the acoustic features of the burst of /k/ as displayed by Center of Gravity,
Relative Intensity and Variance values. This analysis was done via two different sets of analyses. For the first set, the presence or absence of group differences per each vowel context separately was investigated. Only the Spanish productions were explored. This 140 entailed an analysis of whether speakers from Manacor and Artà produced Spanish /k/ differently for each of the vowel contexts separately. The second set examined the presence or absence of differences between the productions of each participant in their two languages for all vowel contexts.
The findings for the two sets of analyses point towards the same direction: 1) The speakers of the two dialectal groups do not differ in terms of how they produce /k/ in
Spanish (whereas they did on their production of Catalan /k/, as shown in Chapter 4); 2)
The Artà speakers use the same phonetic category (very similar acoustic tokens) for both of their languages, Catalan and Spanish; and 3) The Manacor speakers use two different phonetic categories (or allophones) for their two languages, at least in the context of /i/ and /a/, but not in that of /o/. It seems that the Catalan-dominant bilinguals from Manacor do not transfer the phonological rule from their Catalan onto their Spanish; they have two allophones of /k/ in Catalan and only one in Spanish. Any effect that could indicate the presence of crosslinguistic transfer is very small and is only found for one acoustic measure of the three examined. Such effects, therefore, are considered negligible. Thus, it may be concluded that Manacor speakers maintain one Spanish /k/ distribution, which is different from their Catalan /k/ (in terms of Center of Gravity, Relative Intensity and
Variance).
In their Spanish, both groups of participants seem to produce a slightly more fronted /k/ in the context of /i/, but this is due to coarticulation, as suggested by the difference between these values and the ones observed for the production of /k/ in the context of /i/ in Manacor Catalan. For the Manacor speech community, the fronting of 141
Catalan /k/ denotes a phonologization of this feature, whereas for their Spanish (and for
Artà speech in both languages), it may be a mere coarticulatory consequence.
Let us now return to the research question concerning this section, i.e., do these bilinguals maintain two distributions for this phenomenon (one per language) or is there crosslinguistic transfer in any of the two groups? The answer seems to clearly be that there is no crosslinguistic transfer of the phonological rule observed in the Catalan of the bilinguals from Manacor. Speakers from Manacor maintain two separate distributions since they do not transfer the palatalizing distribution or the degree of palatalization observed in their Catalan onto their Spanish. Artà speakers seem to maintain the same or a similar distribution for both languages, since they have similar features regarding production of /k/ in each vowel context: /k/ is generally non-fronted except for the /i/ context, in which it is slightly fronted due to coarticulation.
In sum, this analysis of Spanish production of /k/ in Catalan-Spanish bilinguals has indicated that there is no transfer of the Catalan phonological rule of palatalization onto Spanish speech. Catalan-dominant speakers from Manacor maintained two different distributions for /k/, each one corresponding to each language: whereas they fronted /k/ before /a, i, ə/ in Catalan and did not front it before /o/, they only showed a slight degree of fronting for Spanish /k/ before the /i/ vowel context due to coarticulation. In other words, they have a palatalized and a non-palatalized allophone in Catalan, which they use in complementary distribution. They have a single allophone in Spanish. Artà speakers have a single /k/ allophone, and they use it in both of their languages.
142
CHAPTER 6
DISCUSSION AND CONCLUSION
6.1. Summary of aims and experiment
At a larger scale, the goal of this project was to contribute to the existing literature on velar palatalization, on phonologization of phonetic patterns, and on the phonetics and phonology of early bilingual speakers in a situation of language contact. Specifically, the aims of this project were threefold. Firstly, it investigated, through acoustic analyses, a phonological process of Majorcan Catalan that had only been reported through descriptive impressionistic accounts. Certain areas of Majorca had been reported to produce a velar /k/ with a more fronted place of articulation before front vowels /i, e, ɛ/
(Alcover, 1920; Barnils, 1933; Bernat i Baltrons, 2001; Bibiloni, 1983; Moll, 1952; Moll,
1980; Recasens, 1991; Veny, 1978). Note that these vowel contexts are expected contexts to trigger a fronting of /k/, since these vowels have a fronted articulation, which brings the articulation of /k/ to a more fronted place too (than in other vowel contexts).
Therefore, it would be reasonable to attribute the fronting of /k/ before front vowels to coarticulation. Other areas of Majorca, on the other hand, had been reported to palatalize
/k/ before those same contexts, /i, e, ɛ/, but also before /a, ə/ and in word-final position
(Alcover, 1920; Barnils, 1933; Bernat i Baltrons, 2001; Bibiloni, 1983; Moll, 1952; Moll,
1980; Recasens, 1991; Veny, 1978). These contexts are not common triggers –thus, unexpected contexts– of palatalized articulations. Note that these previous accounts only report a dialectal distinction in the phonological distribution of palatalized /k/, but do not make any references to potential differences between these two dialectal areas with 143 regards to their degree of palatalization. The present project determines what it is that characterizes palatalizing dialects; that is, it analyzes how /k/ differs in the two languages in terms of the contextual distribution of palatalized /k/ and in terms of their degree of palatalization. In yet other words, it examines whether the traditionally reported phonological distribution can actually be supported with acoustic data and whether there is not only a difference in the contextual distribution but also in the degree of palatalization (specifically, before the vowel context /i/).
Velar palatalization in unexpected positions usually starts with the exaggeration of a phonetic feature in an expected context such as /i/ (which should trigger /k/-fronting in both areas, at least due to coarticulation) (Mielke, 2004). By analyzing the dialectal differences in the degree of velar fronting before this vowel context, it could be seen whether this palatalization is simply a coarticulatory-phonetic consequence for both areas equally, or whether it has been phonologized in one of the two dialects. Secondly, this project aimed at determining the origin of the process of velar palatalization in Majorcan
Catalan. It has been claimed that velar palatalization in the unexpected contexts occurs in languages or dialects that present an especially fronted /a/ vowel (e.g. French) (Buckley,
2003). This project examines the nature of /a/ in the two different dialects of Majorca in order to see whether the palatalizing area would produce a more fronted /a/, which might have triggered /k/-palatalization. The third and final goal of the present dissertation was to examine how Catalan-dominant Catalan-Spanish bilinguals produce the velar stop consonant /k/ in different vowel contexts in their non-dominant speech (i.e., in Spanish).
Specifically, it analyzed whether Catalan-dominant speakers who palatalized in both the 144 unexpected and expected contexts (i.e., Manacor speakers) would transfer this phonological rule of their Catalan onto their Spanish.
These research questions were addressed by conducting an experiment designed to examine the production of 40 Catalan-dominant bilinguals (equally divided by sex).
Half of these participants (n = 20) were from Manacor, a Majorcan town taken to represent the traditionally palatalizing area (/k/-fronting before the expected as well as before the unexpected vowel contexts). The other half (n = 20) was from Artà, a small
Majorcan village taken to represent the traditionally non-palatalizing area (/k/-fronting before the expected vowel contexts only). All participants were asked to read a list of
Catalan words containing /k/ in word-initial position before vowel contexts /a, i, o, ə/.
They were also asked to read a list of Spanish words containing /k/ in word-initial position before vowel contexts /a, i, o/.
6.2. Summary of results
This section presents the main findings of the experiment by providing results for
Catalan data first, followed by Spanish data and a comparison between Catalan and
Spanish data.
6.2.1. Catalan data
The results from the Catalan data indicate that there is, in fact, a distinction between the two dialects of Majorcan Catalan under study, and thus corroborate previous descriptive accounts that reported this distinction. Manacor speakers produce a more fronted /k/ before the expected context /i/, as well as before the unexpected contexts /a,
ə/, but produce a non-fronted velar /k/ before /o/. Differently, Artà speakers produce a 145 more fronted /k/ exclusively before /i/, but not before /a, ə, o/, probably due to coarticulatory forces (i.e., since /i/ has a more fronted place of articulation than other vowels, it triggers /k/ to be produced with a more fronted place of articulation).
Importantly, the results suggest that there are not only differences in the distribution of velar palatalization across vowels, but there are also differences in the degree of velar palatalization for the /i/ vowel context: even though Artà speakers do front /k/ before /i/, this /k/-fronting is even more exaggerated in Manacor speakers than in Artà speakers, i.e.,
/k/ in the /i/ context is more fronted in Manacor than in Artà. These findings suggest that whereas /k/-fronting before /i/ is due to coarticulatory forces in Artà, this pattern has been phonologized for Manacor speakers. Regarding production of /a/, Manacor and Artà speakers presented differences in the quality of this vowel and the distribution of its variants. Manacor speakers presented two clearly different allophones of /a/: one fronted allophone after /k/ and one posterior allophone after /p/. Differently, Artà speakers presented one generally central allophone, which was in between the two allophones of
Manacor /a/ both in terms of height and in terms of frontedness. This central allophone was slightly more fronted after /k/ and slightly less fronted after /p/, but this difference was much smaller than the one observed for Manacor data.
6.2.2. Spanish data (and Catalan-Spanish comparison)
The results for the Spanish data indicate that there is no difference between the two dialects in the production of /k/ in different vowel contexts. It seems that both groups of speakers have the same distribution of /k/ in different vowel contexts, which is a more fronted /k/ before certain vowels, such as /i/, and less fronted before other vowels, such as 146
/a/, /o/. It seems, therefore, that there is evidence of coarticulation in Spanish, but there is no evidence of phonologized palatalization. When comparing Catalan and Spanish productions by each group of speakers, the results suggest the following. Artà speakers seem to maintain the same distribution of /k/ for all vowel contexts in their two languages, Spanish and Catalan. In other words, the same pattern and degree of coarticulation is found in all instances of /k/ in this dialect. Differently, Manacor speakers display a different distribution of /k/ depending on the language they are speaking: in
Catalan, they are producing a more fronted /k/ in the context of /i, a, ə/ and a less fronted
/k/ in the context of /o/; in Spanish, they are producing a less fronted /k/ for all vowel contexts than they are doing in Catalan. The fact that they are maintaining two separate distributions (one per language) suggests that they are not transferring a phonological process of their Catalan into their Spanish. Specifically, they are not transferring velar palatalization before the unexpected contexts onto their Spanish.
6.3. Discussion
6.3.1. Velar palatalization in different Majorcan Catalan areas
Most Catalan dialects underwent two palatalization processes (First and Second palatalizations), as the rest of Romance languages. Whereas some languages, such as
French, underwent a further palatalization process affecting velar consonants (Third
Velar Palatalization), Catalan did not undergo such a change. Present-day Majorcan
Catalan has been reported to present a velar palatalization process affecting /k/ before /a/,
/ə/ and in word-final position in Majorcan Catalan (Alcover, 1920; Barnils, 1933; Bernat i Baltrons, 2001; Bibiloni, 1983; Moll, 1952; Moll, 1980; Recasens, 1991; Veny, 1978). 147
This suggests that this dialect might be at an initial stage of what might have happened in
French prior to the velar softening process known as the Third Velar Palatalization: /k/ >
/kj/ > /c/ > /tʃ/ > /ʃ/. This feature of Majorcan Catalan, however, has not been reported to occur in all varieties of Majorcan Catalan. A geographical division of the island has been proposed to account for the difference in the articulation of Catalan velar stops.
According to traditional impressionistic descriptions (Alcover, 1920; Barnils, 1933;
Bernat i Baltrons, 2001; Bibiloni, 1983; Moll, 1952; Moll, 1980; Recasens, 1991; Veny,
1978), the island of Majorca is divided between a non-palatalizing area (e.g. Bunyola,
Esporles, Sa Pobla, Alcúdia, Sineu, Artà, Santa Margalida, Sóller, Campos) and a palatalizing area (e.g. Palma, Valldemossa, Pollença, Son Servera, Felanitx, Manacor,
Calvià, Algaida, Santanyí). The non-palatalizing area is said to front /k/ only before the expected contexts, that is, before front vowels, /i/, /e/, /ɛ/. Note that these contexts are expected because their fronted articulation automatically brings the articulation of /k/ to a more fronted position too. The palatalizing area, on the other hand, is said to present a palatalized articulation of /k/ before the front vowels, as well as before /a/, /ə/ and in word-final position (unexpected contexts). Neither of the two areas is said to palatalize before back vowels /o/, /ɔ/ and /u/. Note that these traditional descriptive accounts only report (impressionistic, descriptive) differences between the two areas regarding the contextual distribution of palatalized /k/, but do not provide any type of information regarding the degree of palatalization (irrespective of position; or in the palatalizing contexts). Additionally, note that this dialectal difference had only been reported impressionistically through descriptive accounts. Whereas some studies (Recasens &
Espinosa, 2006; Recasens, 2014) provided acoustic and articulatory characteristics of /k/ 148 in palatalizing dialects of Majorcan Catalan, these studies have not provided acoustic data for the two different dialects of Majorcan Catalan. Therefore, the palatalizing dialectal difference had yet to be acoustically proved. In other words, relative to non-palatalizing speakers, the degree to which palatalizing speakers front their /k/ has not been explored in the literature. Therefore, prior to the present dissertation, it was not possible to assess whether the palatalization of /k/ results from the phonologization of a coarticulatory pattern or can be explained by attributing this pattern to mere coarticulatory forces.
The results of the present dissertation (1) corroborate the differences between the two areas in terms of the contextual distribution of palatalization and (2) provide data to show differences in the degree of palatalization. Let us concentrate first on the analysis of the contextual distribution of palatalized /k/. Manacor speakers, who were taken to represent the palatalizing area, did in fact produce more palatalized versions of /k/ before vowel contexts /i, a, ə/ (in that order), and did not produce palatalized versions of /k/ before vowel context /o/. Differently, Artà speakers produced fronted versions of /k/ before vowel context /i/ only, but not before vowel contexts /a, ə, o/. These results support, through acoustic data, the descriptive accounts that report a dialectal difference between palatalizing and non-palatalizing dialects; that is to say, in Artà, /k/ is fronted only before front vowels while, in Manacor, it is palatalized before front and central vowels (Alcover, 1920; Barnils, 1933; Bernat i Baltrons, 2001; Bibiloni, 1983; Moll,
1952; Moll, 1980; Recasens, 1991; Veny, 1978). Additionally, it supports the categorization of Manacor as a town that belongs to the palatalizing dialect and Artà as a town that belongs to the non-palatalizing dialect. 149
The current dissertation, therefore, corroborates the presence of a distinction between the two dialects with regards to the contextual (vowel) distribution of /k/- palatalization. The question remains as to whether the process of velar palatalization in these dialects is merely a phonetic consequence of coarticulation or whether it is a phonological feature for any of the two dialects. Let us concentrate on this in the following subsection.
6.3.2. Velar palatalization: Coarticulation or phonologization?
In order to analyze whether velar palatalization in the two dialects under study is merely due to coarticulation or whether it is in fact a phonological effect, it was necessary to compare the degree of palatalization of /k/ between the two communities. Vowel context /i/ serves as the perfect context to analyze this. As mentioned before, the articulation of vowel /i/, which is far more fronted than other vowels, brings the articulation of /k/ to a more fronted place of articulation. This coarticulatory effect is universally observed (Öhman, 1966). Importantly, it is observed in both dialects under study: for both dialects, /k/ in the context of /i/ is more fronted than /k/ in any other vowel context. Even if this is true for both dialects, it is relevant to observe whether the degree of /k/-palatalization in this context is different in the two dialects, i.e., whether the palatalizing dialect palatalizes /k/ even more exaggeratedly than the non-palatalizing dialect does for this vowel context. The present findings suggest that this is precisely the case: Manacor speakers produced a far more fronted /k/ than Artà speakers did for vowel context /i/. These results indicate that for Artà speakers, fronting of /k/ in the context of
/i/ (remember that it is not observed for any other vowel context in this community) is a 150 mere consequence of the articulation of two neighboring sounds: the frontedness of /i/ is bringing the articulation of /k/ to a more fronted place of articulation, and this effect is relatively modest. This contrasts with the articulation of /k/ in the context of /i/ for
Manacor speakers, which has been exaggeratedly brought to an extremely fronted place of articulation. These results suggest that, instead of being a phonetic consequence of articulation, this is a phonological effect for Manacor. An automatic phonetic pattern that could previously be predicted by universal phonetic principles has become unpredictable
(Hyman, 1976) for this community. The phonetic gesture of palatalizing before front vowels has been exaggerated and reinterpreted as intended, thus leading to the phonologization of the process for these speakers. The gesture can no longer be attributed to universal phonetics (coarticulation), but it is now used as an intended cue to identify a phonological element.
Once a phonetic tendency has become phonologized in one context, it is a possibility that it becomes spread analogically to other segments. Consider the case of the low vowel /a/. As mentioned before, /a/ is not a common context for velar palatalization to occur. However, once the gesture has become phonologized in an expected context, such as /i/, it is probable that the same feature is exaggerated –and later phonologized– in other contexts (Mielke, 2004). For instance, consider the case of vowel nasalization in
French. Vowel nasalization tends to occur more often for lower vowels instead of higher vowels due to articulatory forces, i.e., tongue lowering facilitates velum lowering. It is proposed that vowel nasalization in Old French began in the 11th century for vowel /a/ due to articulatory forces and that it was spread to other vowels in the 14th century (Chen,
1973). The change, therefore, had its roots on an expected context due to coarticulation, it 151 was then exaggerated (and, later, phonologized) in this context, and later extended to (and phonologized in) other unexpected contexts.
Something similar could have happened to velar palatalization in Majorcan
Catalan. For something to be phonologized in an unexpected context, it is probable that the feature has already been exaggerated (and phonologized) in an expected context. It could have been the case that /k/ was first phonologized in the context of /i/ initially triggered by articulatory forces, later phonologized in this vowel context. Then, it would be reasonable to think that it was spread to /k/ in other less expected contexts, such as /a/.
Whereas velar palatalization before front vowels is a common process observed in multiple language families, such as Indo-European, Salish, Mayan, Sino-Tibetan, Niger-
Congo, among many others (Bhat, 1974; Bloomfield, 1933; Buckley, 2003; Chen, 1973;
Halle, 2005; Hock, 1986; Guion, 1996; Ladefoged & Maddieson, 1996), velar palatalization before the low vowel /a/ is typologically rare. Vowel context /a/ is not a crosslinguistically common trigger for velar palatalization (Buckley, 2003; Guion, 1998;
Jakobs & Berns, 2013), but it has been found in other languages such as French. Mielke
(2004) proposes that the spreading of a feature to other segments or contexts is more often found when such segments or contexts already present the feature that has been phonologized at least to a lesser degree, but robustly. This suggests that there should be some sort of frontedness in /a/ for /k/ to become palatalized in this context. In fact,
Buckley (2003) proposes that palatalization of /k/ in the context of /a/ occurs in varieties in which the low vowel has an especially fronted allophone. This is exactly what has been proposed for Majorcan Catalan. This dialect of Catalan has been described to present a more fronted (and higher) articulation for /a/ than other Catalan dialects (Recasens, 1991, 152
2011, 2014; Recasens & Espinosa, 2006), with F1 and F2 values that are higher than for other varieties. Note, however, that such studies did not mention the subdialect of
Majorcan Catalan participants belonged to, which made it impossible to know whether
/a/-fronting was in any way related to /k/-palatalization. Therefore, Majorcan Catalan was taken as a whole when describing the acoustic qualities of /a/. It could have been the case that Majorcan Catalan as a whole presented a more fronted and higher /a/, but it could also be the case that there were subdialectal differences that had not been previously reported. If there were differences in the quality of /a/ between two communities, this could be an answer to why there were differences in the quality of /k/ too. It could be the case that, similar to what happened in French, a more fronted realization of the low vowel triggered this change only in those varieties of Majorcan Catalan that presented a more fronted allophone. The present dissertation found that, in fact, Manacor speakers had a more fronted and higher /a/ for the context of /k/ than Artà speakers did. This was true only when /a/ was in the context of /k/, but not when /a/ was found in other consonant contexts, such as /p/, in which case /a/ was even more posterior and lower than the one reported for Artà. This finding establishes a clear relationship between the process of velar palatalization and /a/ fronting, which will be discussed in the following subsection.
6.3.3. Two phonological changes: Can we establish the order of the influence?
We now know that Manacor speakers present differences in the quality of /k/ (i.e., their /k/ is more fronted for /a, i, ə/) and in the quality of /a/ (i.e., their /a/ is more fronted and higher for consonant context /k/ than for /p/) when compared to productions by Artà speakers (whose /a/ is more central and in between the two variants of Manacor /a/). 153
Knowing this information, the next logical step is to understand the relationship between the two sound changes: Do we have enough information to determine the order of the influence? Did the fact that /a/ was already fronted for this community trigger the palatalization of /k/? Or did the fact that /k/ was being palatalized in other contexts such as /i/, and was then spread to vowel context /a/, trigger a more fronted articulation of /a/?
Since both sounds are different in the dialect of Manacor, a relationship can be established between the two processes, but the order of the influence cannot be established as a one-way type of influence, i.e., there is no clear evidence pointing at which process, velar palatalization or /a/-fronting, came first and triggered the other one.
Thus, we may consider the possibility of a third type of historical interaction. This is the case of a relationship based on the interaction between the two processes towards their phonologization.
Take as an example the ongoing tonogenetic sound change in Seoul Korean
(Bang, Sonderegger, Kang, Clayards, & Yoon, 2015). This variety presents a system of three voiceless stops (tense, lax, aspirated), which were initially contrasted mainly through VOT differences, and secondarily by F0 differences. Recently, this traditional consonantal VOT cue has been replaced by changing the F0 of the following vowel. An initially articulatory gesture (differences in F0) seems to have become the cue that displays contrast between phonemes: this gesture has been phonologized. While this process, known as tonogenesis, is a common type of sound change, not much is known about how this change unfolds over time: Does the exaggeration of the vocalic cue lead to the collapse of the consonantal cue (hyperarticulation F0 distinction)? Or does the reduction of the consonantal cue (hypoarticulation of VOT distinction) lead to the 154 exaggeration of the vocalic cue? Bang et al. (2015) discuss the possibility that both processes occur at the same time, in tandem: the loss of the VOT contrast seemed to have happened at the same time as the enhancement of the F0 contrast. This finding is consistent with theories of adaptive-behavior sound change (Diehl, 2008; Lindblom,
1990) that argue that “languages develop in an economical way through continuous negotiation between articulatory ease and perceptual contrast” (Bang et al., 2015: 4). It seems that as the precision of a contrast is reduced, cues are enhanced to compensate for this reduced precision. Kirby (2013) proposes that the degree of enhancement is “a probabilistic function to contrast precision, while the probability with which a given cue is enhanced is related directly to its informativeness, the degree to which it contributes to accurate identification of a speech sound” (Kirby, 2013: 3). Importantly, he proposes that in Seoul Korean, probabilistic enhancement (of F0) and loss of contrast precision (in
VOT) must interact simultaneously for phonologization to take place. In other words, in
Seoul Korean, contrast in voiceless stops is maintained by a simultaneous contrast reduction in VOT and an adaptive expansion in F0.
Let us now go back to the issue of Majorcan Catalan /k/-palatalization. The results indicate that Manacor speakers present evidence for two sound changes: (1) a process of velar palatalization in the /a/ context, and (2) a process of /a/-fronting in the /k/ context.
Note that this is not observed in Artà speakers, since their fronting of /a/ and /k/ is extremely subtle when compared to the one presented by Manacor speakers in certain contexts: the size of the effect determines that, whereas both processes have been phonologized in the speech of Manacor, they are merely due to coarticulation in the speech of Artà. For Manacor speakers, according to our data, the relationship between the 155 two processes seems to be robust. As mentioned above, it could be the case that one of the processes triggered the other process, but our data do not allow us to see which phenomenon may have preceded the other one. A third interpretation of the origin of these changes relies on a parallel influence between the two, as discussed for Seoul
Korean tonogenesis (Bang et al., 2015). At the same time that /a/ was being fronted in certain dialects of Majorcan Catalan (i.e., Manacor), /k/ could have also been moving towards a more forward place of articulation. Rather than one process initiating the other process, both processes could have proceeded simultaneously. Thus, for the speech community of Manacor, /a/ might have become more fronted at the same time that /k/ was becoming palatalized. This does not mean that both changes occurred drastically at the same time, but that the initiation and development of the sound changes might have occurred gradually step by step by influencing one another. Perhaps the fact that front vowels had already triggered a fronted allophone of /k/ due to coarticulation triggered the same slightly fronted version of /k/ in other vowel contexts such as /a/. Then these two sounds, /k/ and /a/, might have slightly brought each other to a more fronted position: gradually, the /a/ sound might have become fronted and /k/ may have become palatalized, which, at the same time, might have triggered more frontedness and more palatalization.
Finally, at a certain moment in time, the fronting of the /a/ and the palatalization of the /k/ must have become exaggerated, the articulatory movements must have become intended and, therefore, the processes must have been phonologized.
156
6.3.4. Non-dominant speech of Catalan-Spanish bilinguals: Evidence of phonetic
transfer?
In consecutive bilingual acquisition, it has been suggested that the longer the asynchrony between the onset of the acquisition of a first language and the onset of the acquisition of a second language, the larger the consequences for the acquisition of phonetic native-like patterns of the L2. This has been suggested to be true both for perception of L2 sounds (Bosch et al., 2000; Flege & MacKay, 2004; Goto, 1971; Hojen
& Flege, 2006; Pallier et al., 1997; Sebastián-Gallés & Bosch, 2005; Sebastián-Gallés &
Soto-Faraco, 1999; Werker & Tees, 1984) as well as for production of L2 sounds (Flege,
Munro, & MacKay, 1995; Flege et al., 1999; Fowler, Sramko, Ostry, Rowland, & Hallé,
2008; Long, 1990; Oyama, 1976, Piske, Flege, MacKay, & Watt, 2002; Simonet, 2010,
2011a). Many times, these difficulties in the acquisition of L1 patterns entail the transfer of a phonetic element from the speaker’s L1 onto their L2 (Flege, 1991; Long, 1990). It seems that linguistic sequential acquisition makes the L1 phonological patterns to act as a filter that interferes with the acquisition of the phonetic patterns of the L2. Some studies have suggested that, even with an early and extensive exposure to the L2, there are difficulties to acquire L2 patterns as if this was an L1 (Guion, Flege, & Loftin, 2000;
Fowler et al., 2008; Mack, 1989; Pallier et al., 1997; Piske et al., 2002; Sebastián-Gallés
& Soto-Faraco, 1999; Simonet, 2010, 2011a). Some other studies have shown that such difficulties can be overcome (or reduced) and that bilinguals can acquire near-native phonetic patterns of their L2 (Amengual, 2011, 2016; Best & Strange, 1992; Flege, 1995,
2007; Herrick, 2006). Importantly, even these second types of studies show effects of phonetic transfer: bilinguals might acquire the phonological categories of their non- 157 dominant language, but these categories present phonetic-acoustic differences from those produced by bilinguals for whom that language is their dominant language.
The community of Catalan-Spanish bilinguals has been previously analyzed in order to understand the intricacies of bilingual production patterns. Most studies have been analyzed from the perspective of Catalan, that is, they have analyzed the influence of Spanish over Catalan by Spanish-dominant subjects. The present dissertation does the opposite: it analyzes the potential influence of Catalan over Spanish by Catalan-dominant bilinguals. The findings of this dissertation contrast with previously reported findings for this same community of bilinguals: no evidence of phonetic transfer was found in the present data. Specifically, the research question aimed at analyzing whether Catalan- dominant bilinguals who showed evidence of velar palatalization in their Catalan speech
(i.e., speakers from Manacor) would transfer such phonological rule onto their non- dominant speech in Spanish. The results showed that, whereas they did show velar palatalization in Catalan, there was no trace of it in their Spanish. It is true that a tendency could be observed for /k/-fronting in the context of /i/ in their Spanish, but this was merely due to coarticulatory forces: their Spanish /k/ in the context of /i/ was definitely not as fronted as their Catalan /k/ in the context of /i/. Importantly, no evidence of /k/-fronting was found in the context of /a/ in their Spanish speech (as was found in their Catalan speech). Therefore, the results seem to robustly indicate that there is no evidence of transfer of velar palatalization onto non-dominant speech in Spanish for these speakers. It seems logical, then, to determine that Catalan-dominant speakers of the town of Manacor have two different distributions for the articulation of /k/: one distribution for their Spanish /k/ and one distribution for their Catalan /k/. In Spanish, they produce /k/ 158 with a certain degree of contextual variability, especially for vowel context /i/, which clearly brings the articulation of /k/ to a more fronted position. Their articulation of /k/ in
Spanish, however, does not reach the degree of fronting that is observed for their articulation of /k/ in Catalan. This suggests that in Spanish, Manacor speakers are bringing the articulation of /i/ to a more fronted position simply because the fronted articulation of /i/ automatically brings the articulation of /k/ to a more fronted position too. Differently, in Catalan, they produce an especially fronted /k/ in the context /i/, a very fronted /k/ in the contexts of /a/ and /ə/ and a velar /k/ in the context of /o/. The distribution of /k/ between languages and communities seems to be different. To sum up, all speakers manifest the consequences of the same coarticulatory forces in their speech, both in Catalan and in Spanish. This is the extent of the facts for the Artà speakers. The speakers from Manacor, on the other hand, show a different pattern: they have a phonological process in their L1 that they do not transfer to their L2. Therefore, Manacor speakers have a rule (a phonologized process) that they apply only in their L1. They, thus, have language-specific phonologies.
The question arises now as to why it is that participants from the Catalan-Spanish community of Majorca do not show transfer effects when previous studies on this same community have shown that phonetic transfer is a common outcome of –even early– consecutive bilingualism. Most studies on this community have concentrated on analyzing production of Catalan contrasts by Spanish-dominant speakers (Amengual,
2011, 2016; Herrick, 2006; Lleó et al., 2009; Ramírez & Simonet, accepted; Simonet,
2011a, 2014) and have shown that there are phonetic transfers from Spanish (L1) onto
Catalan (L2). The few studies that have looked at the production of Spanish features by 159
Catalan-dominant bilinguals have also proved the presence of phonetic transfer effects in non-dominant speech (Spanish in this case). For instance, Simonet (2011b), in a study about sentence-final F0 contour in declarative sentences in Spanish and Catalan as produced by Majorcan Catalan- or Spanish-dominant bilinguals, finds effects of phonetic transfer in L2 intonation patterns: Catalan-dominant females produce Spanish with a
Catalan-like intonation. Similarly, in Simonet (2010), alveolar laterals in both languages are analyzed and phonetic transfer effects are also found: laterals produced in the L2 are different from laterals produced by dominant speakers of that L2. Findings for the
Spanish literature on Catalan-Spanish contact for this community, though scarce, points at phonetic transfer of features from the L1 (Catalan) onto the L2 (Spanish) as an inevitable consequence of sequential acquisition. Therefore, transfer effects from the L1 onto the L2 seems to be a robust finding for the Catalan-Spanish community both in those studies that look at production of Spanish by Catalan-dominant bilinguals and in those studies that examine production of Catalan by Spanish-dominant bilinguals.
Transfer of (at least minor) phonetic features has been shown to occur robustly in this community.
Then, why is it that participants with a very Catalan-dominant linguistic profile show no effects of phonetic transfer onto their Spanish? These participants reported a past and present lower use of Spanish in their daily lives (much lower than their use of
Catalan), in terms of work, family or friends. They also reported to have a high proficiency in Spanish, even though they hardly use it. Perhaps such proficiency and knowledge of Spanish also entails an awareness of the sociolinguistic connotations that pronouncing velar stops as palatalized velars in Spanish has. Manacor speakers are very 160 aware that velar palatalization is a very distinguishable feature. More importantly, they are very aware that velar palatalization is an idiosyncratic feature of their community, namely, the town of Manacor. In fact, this characteristic has a tradition of being perceived as non-standard and has even been stigmatized: “Almost all features that the linguistic conscience within Majorcan people rejects are present in the speech of
Manacor. (…) the horrible palatalization of velar sounds, which is so characteristic of speech from the city (Palma) and which is so socially telling, which becomes even more intense the lower you go in the social ladder, to the point of being caricaturesque14”
(Vidal Alcover, 1993: 20). Personal observations, however, suggest that people from
Manacor have turned what was a traditionally stigmatized feature around the island into a feature of pride and identity, at least for their Catalan. Manacor is the second largest town in Majorca, after Palma, the capital city. The fact that it has always been a secondary town might have triggered this pride in all that is characteristic of Manacor. In situations of linguistic rivalry or conflict it is common for speakers of a language to attribute a higher status and prestige to a language variety as a sign of identity (Weinreich, 1968).
Motivated by the need to create a sense of community, a linguistic group assigns a certain degree of prestige to their own stigmatized linguistic variety. This has been termed as covert prestige (Trudgill, 1972). Personal observations of the community of Manacor suggest this is a community that takes pride in their language and, specifically, in the features that make Manacor Catalan different. These features might have acquired social
14 Original quote: “Quasi tots els trets que la consciència lingüística latent dels mallorquins rebutja són presents en el parlar de Manacor. (…) la terrible palatalització dels sons velars, tan característics de la parla ciutadana i tan determinants socialment, que així que davallam en l’escala social es fa més intensa, fins a fer-se caricaturesca” 161 saliency or social meaning and they use it as a sign of identity. It is not surprising, then, to observe the maintenance of this feature in their Catalan speech in the present data.
When moving onto non-dominant speech in Spanish, however, the issue is different. This community might also be aware that heavy Catalan-accented speech in
Spanish is extremely stigmatized in Majorca (Pieras, 1999). Specifically, producing palatalized velars in Spanish can be perceived as sounding rural or uneducated. It could be the case that Manacor speakers, when using Spanish, try to avoid such forms in order to avoid such negative or stigmatized perceptions. Therefore, what seems to be a marker of identity in Catalan, is a marker of rurality and lack of education in Spanish, which makes it undesirable and therefore avoided in this language. They seem to be using this phonetic feature differently in each language to modulate others’ perceptions of them.
Other studies within the same community have shown that the speakers’ awareness of the stigmatization of certain features might be the cause to change their behavior regarding that feature. For instance, the aforementioned study by Simonet
(2010) finds that Catalan-dominant females, instead of producing velarized laterals in their Catalan speech, produce alveolar non-velarized laterals. It is possible that they are aware that velarized laterals are interpreted as rural or uneducated, which might be the reason why they avoid them in their speech. Note that even if these findings are found in
L1 speech, rather than L2 speech (as in the present dissertation), the factor of sociolinguistic awareness can also play a role in L2 production. Pieras (1999) studied the same feature in the same population and found that younger Catalan-dominant speakers did not transfer velarized laterals from their Catalan onto their Spanish. According to this study, velarization of laterals is a very notorious and easily perceptible feature commonly 162 used, together with intonation, for imitations and mocks of Majorcan Catalan-dominant bilinguals speaking Spanish (Pieras, 1999). Therefore, these younger Catalan-dominant speakers seem to have been consciously avoiding the transfer of velarized laterals into their speech in Spanish in order to modulate others’ perceptions of them.
The conscious use of certain phonetic features over others to display or avoid displaying an identity has been deemed like a plausible explanation for speech patterns that have been observed in the Catalan-Spanish bilingual community (Pieras, 1999;
Simonet, 2010), as well as in many others. One of the most famous studies was conducted by Labov (1972) on the island of Martha’s Vinyard, Massachusetts. He reported on a shift towards the centralization of the first item of the diphthongs ay and aw. This change was marked by the social attitudes of the speakers. At that time (1930s to 1970s), the island was receiving migrants from the mainland. In order to emphasize a difference from the immigrants, the English Yankee settlers, who were mainly fishermen, began emphasizing the first elements in the diphthongs as centralized, a feature that people from the mainland did not display in their speech. This triggered the beginning of the change, which was later adopted by other ethnic groups, such as the aboriginal
Indians and more recent Portuguese settlers. Importantly, people who wanted to leave the island and go to the mainland soon did not centralize their diphthongs as much as people who wanted to stay on the island did. It seems that in order to identify more deeply with the speech community of the island, those who wanted to stay on Martha’s Vineyard consciously adopted the centralized diphthongs, whereas people who wanted to maintain a distance with the island’s speech identity, maintained non-centralized diphthongs.
Changes in phonetic-phonological features in a certain dialect, therefore, were directly 163 linked to the social attitudes of the speakers: those who wanted to be perceived as part of the community adopted their feature; those who did not want to be perceived as part of the community refrained from doing so.
A final interpretation of the lack of transfer in our Spanish data could be related to the nature of the task conducted. Participants were asked to read a series of sentences that contained the target words. It could be the case that creating a formal context for data collection has affected their speech production, which, in this case, is careful speech. As
Labov (1972) mentioned, access to the vernacular, that is, access to informal unselfconscious language used in everyday situations is highly jeopardized when the participant is aware of being recorded, thus potentially leading to the modification of speech patterns (commonly known as the observer’s paradox). Thus, the Spanish spontaneous speech of this community might actually present velar palatalization that has been transferred from Catalan. The nature of the task at hand might simply not be capturing such transfer of this phonetic feature. This interpretation, however, should be taken with care. If this is the case, this dissertation contrasts with other studies conducted in a phonetic laboratory that did find effects of transfer. For instance, most of the aforementioned studies on the Catalan-Spanish community of Majorca that found (at least minor) phonetic transfer effects were conducted on a similar type of research setting
(Amengual, 2011, 2016; Ramírez & Simonet, accepted; Simonet, 2010, 2011a, 2014).
Why is it, then, that transfer of other phonetic features has been found in studies that have collected laboratory speech but not in this one? It could have to do, perhaps, with the idea that speakers might be even more aware of the social saliency of this feature than of other previously studied features and therefore decide not to produce them in such a formal 164 context. It could also be that, since there is not a clear Catalan monolingual norm (due to the lack of Catalan monolinguals), transfer of Spanish features onto Catalan is socially acceptable (and so Spanish-dominant speakers do not have to work hard enough to prevent this from happening), while transfer of Catalan features onto Spanish is mostly unacceptable (and so Catalan-dominant speakers avoid this in their Spanish speech). This might explain why it is that the literature provides much less evidence of transfer of
Catalan features onto Spanish speech than vice versa. Finally, it could be due to the fact that mere statistical distributions of input or intake of the language make it a likely finding: whereas everybody has access to monolingual Spanish speakers (even monolingual Spanish-speaking Majorcans), nobody has access to monolingual Catalan speakers (anywhere in the world). A major accessibility to input in Spanish might be leading to a smaller degree of accented speech in this language; a smaller accessibility to input in Catalan might be leading to a higher degree of accented speech in Catalan. In any case, collection of data in other more informal settings might help answer these questions.
6.4. Future directions
The present dissertation only presented participants from two different towns or speech communities to represent the dialectal division regarding velar palatalization mentioned in previous descriptive accounts. Future research should include data from many other areas within each subdialect to see if they also correspond with what traditional descriptions reported. Specifically, the capital city of the island, Palma, should be analyzed in terms of velar palatalization. Even if it has been traditionally described as 165 being a palatalizing area (Alcover, 1920; Barnils, 1933; Bernat i Baltrons, 2001; Bibiloni,
1983; Moll, 1952; Moll, 1980; Recasens, 1991; Veny, 1978), personal observations suggest that speakers from Palma might be moving towards a non-palatalization norm, thus approaching the production observed in Artà speakers, rather than Manacor speakers. Several socio-historical events that have occurred during the last decades in
Majorca might have contributed to a (hypothetical) loss of the palatalization of /k/ in
Palma. Since the 1960s, the island has received multiple waves of tourism –usually concentrated in Palma and on the island’s coasts– as well as many waves of immigrants from mainland Spain and from abroad. These immigrant families, many of them monolingual Spanish speakers, have established themselves on the capital city. These changes in the social panorama of the capital city have affected the linguistic abilities reported for the capital city residents: 83% report understanding, 55% report speaking,
68% report reading, and 42% report writing in Catalan. These percentages for Palma are much lower than the ones presented in Manacor and Artà, both towns within the eastern territory (or "Comarca") of Llevant: 87% reports understanding, 72% reports speaking,
78% reports reading, and 44% reports writing ability in Catalan (IBE, 2010). The lower reported abilities in Catalan in Palma are related to a lower use of Catalan in the capital city (31% of Catalan use) when compared to the other areas of Majorca such as Llevant
(42%). It seems reasonable to conclude that, if there is any loss in the idiosyncratic features of Majorcan Catalan, these might occur first in the capital city, Palma, where there is a lower daily use of Catalan and where contact with Spanish is more intense.
Future studies should collect data from speakers from the capital city in order to see whether velar palatalization is still a characteristic of Palma or whether it is changing in 166 this speech community. Data from diverse age groups would be helpful to see if there are any differences between the generations and thus report a potential language change in progress.
Another point of interest is related to the lack of phonetic transfer found in the present dissertation. This work found no trace of L1-to-L2 transfer of the velar palatalization feature. As mentioned above, this lack of transfer might be due to an actual lack of transfer for this feature into the Spanish of Manacor speakers. This could also be due to a higher degree of awareness of the markedness of this feature and of the stigmatization of Catalan-accented speech in general. However, there is a possibility that this lack of transfer evidence might be due to the nature of the experiment conducted for this dissertation. Laboratory speech has been defined as careful speech that might not necessarily reflect how speakers of a community really produce language in their daily lives (Labov, 1972). Future research on the same community but with a different data collection method might corroborate the lack of transfer for this feature or might provide evidence for transfer in more natural speech. Even though many previous studies that collected data in the phonetic laboratory showed effects of transfer from the L1 onto the
L2 (Amengual, 2011, 2016; Ramírez & Simonet, accepted; Simonet, 2010, 2011a, 2014), the current data found that bilinguals can actually maintain their two languages separate without the effect of phonetic/phonological transfer.
As mentioned above, the lack of transfer effects has been explained through the idea that Manacor speakers are highly aware of this feature of Majorcan Catalan and of its stigmatization when produced in Spanish. Future research might benefit from perception studies that actually prove that this feature is highly noticeable and that 167
Majorcan speakers are highly aware of it. Additionally, sociolinguistic studies of language attitudes (in the form of matched-guise experiments or in the form of qualitative interviews) might provide an insight into the perception of this feature in both Catalan and Spanish. Studies such as these might shed light on the lack of phonetic transfer observed in the present dissertation.
Finally, most of the existing acoustic studies on the Catalan-Spanish bilingual community analyze the production and perception of Catalan contrasts by Spanish- dominant speakers. Future research should analyze the perception and production of
Spanish contrasts by Catalan-dominant speakers in order to understand better the linguistic dynamics of this bilingual community as a whole.
6.5. Concluding remarks
The present dissertation investigated the process of velar palatalization in the
Catalan and Spanish speech of Majorcan Catalan-dominant bilinguals. Specifically, it provided acoustic data for a dialectal distinction in Majorcan Catalan that had only been reported through descriptive impressionistic accounts. Acoustic analyses allowed to confirm that speakers from the Majorcan town of Artà produce the voiceless velar stop consonant /k/ with a similar place of articulation when produced before vowel contexts
/i, a, ə, o/. Only for vowel context /i/ did /k/ appear to have a relatively more fronted place of articulation, which may be explained as a phonetic consequence of coarticulation: the more fronted articulation of vowel /i/ brings the articulation of /k/ to a more fronted place of articulation. More importantly, the present dissertation provided acoustic data showing that speakers from the town of Manacor produce /k/ with a more 168 fronted place of articulation (they palatalize /k/) not only before an expected context such as /i/, but also before vowel contexts /a, ə/. The same speakers produced /k/ with a less fronted place of articulation (probably as a velar) before vowel context /o/. Apart from these dialectal differences in the distribution of velar palatalization in Catalan, the present dissertation reported on differences in the degree of palatalization between the two dialectal groups. Specifically, Manacor speakers showed a higher degree of palatalization before vowel context /i/ than Artà speakers did. This was explained as being evidence of the phonologization of velar palatalization in Manacor speakers. What is simply an articulatory-phonetic consequence of coarticulation in Artà speech is an intended articulatory gesture in Manacor speech. In sum, the data in this dissertation support the following conclusion: in addition to coarticulatory forces, which apply to all speakers
(including the Manacor and Artà speakers in this dissertation), there is evidence for postulating a phonological (linguistic, cognitive) process for only one of the dialects under investigation. That is, the Manacor speakers have, in their phonologies, an allophonic rule: a complementary distribution between palatalized and non-palatalized
/k/. The Artà speakers have no such rule in their phonologies: their production patterns can be entirely attributed to the phonetic implementation component.
The second contribution of the present dissertation was related to the nature of
Majorcan /a/ and its relationship with the process of velar palatalization. An analysis of the acoustic features of /a/ as produced by both dialects under study showed that the low vowel is fronted and high in the speech of Manacor when it is produced after /k/ and it is posterior and low when produced after /p/ (as opposed to Artà /a/, which is generally central in both contexts). Previous claims stated that palatalization of /k/ before /a/ 169 occurred in languages or dialects that present an especially fronted /a/ (such as French).
The present dissertation provided evidence of a clear relationship between palatalization of /k/ and fronting of /a/. Rather than pointing at one of the two processes triggering the other one, the present data are in consonance with a theory of parallel influence between the two, as has been observed for processes in other languages, such as the process of tonogenesis in Seoul Korean. This dissertation proposes that for the speech community of
Manacor, /a/ became more fronted at the same time that /k/ became palatalized. This did not necessarily have to happen drastically, but both sounds might have changed step-by- step influencing one another.
Finally, the production of Spanish by the same groups of Catalan-dominant speakers showed that there was no phonetic transfer from Catalan into Spanish.
Specifically, Manacor speakers, who palatalized /k/ in their Catalan, did not transfer this phonological process/rule onto their Spanish (their non-dominant language), that is,
Manacor speakers maintain language-specific phonological rules. This contrasts with previous literature based on the Catalan-Spanish bilingual community, which has robustly shown that transfer of (at least minor) phonetic features is a common consequence of even early consecutive bilingual acquisition. This difference with previous studies has been explained in terms of the potential social saliency of this feature. In Manacor Catalan, this feature is perceived as a sign of identity. In Spanish, however, producing highly Catalan-accented speech (as is producing velar palatalization in Spanish) is perceived as a marker of rurality or lack of education. Therefore, these bilinguals might be avoiding the production of a stigmatized feature in their non- dominant speech. This might also have to do with the social acceptability of transferring 170
Spanish features into Catalan and the unacceptability of transferring Catalan features into
Spanish.
To sum up, this project has contributed to the existing literature on velar palatalization and on the phonologization of such process. By hypothesis, it has captured a snapshot of a process that affected Gallo-Romance around the late 5th century AD or the early 6th century AD (Buckley, 2003; Pope, 1934; Price, 1971) and is affecting some dialects of present-day Majorcan Catalan, thus offering some lessons on sound change actuation (Weinreich, Labov & Hertzog, 1968). Additionally, it has provided further understanding of the phonetics and phonology of early consecutive bilingual speakers in a situation of societal language contact.
171
APPENDIX 1. List of Catalan stimuli (central column) divided by vocalic context (/a/, /ə/, /o/, /i/) (left column), and translation into English (right column) Vocalic context Catalan stimuli Translation /k/ + /a/ cada Each casa House cara Face canvi Change cambra Room carta Letter calma Calm canta Sing (3rd person sg) canya Rod càstig Punishment càlcul Calculation càntar Pitcher cala Cove cabra Goat càlid Warm /k/ + /ə/ camí Road cavall Horse cafè Coffee calor Heat cançó Song casal House cabell Hair canal Canal calent Hot cabal Honest callat Quiet cambrer Waiter captar To capture camell Camel candor Candor /k/ + /o/ conya Joke coca Cake cola Glue coixa lame cota duck (3rd person sg) conte story compte check coma coma copa glass colze elbow coco coconut cofre chest coba suck up to (donar coba) 172
colma deluge (3rd person sg) compra buy (3rd person sg) /k/ + /i/ quina which (plural, fem.) quinze fifteen química chemistry químic chemist quinta fifth quilo kilo químicament chemically quilla keel quins which (plural, masc.) quístic cystic quíntuple quintuple quin which (sing, masc.) quimio chemo (-therapy) quica hen kiwi kiwi
173
APPENDIX 2. List of Catalan control items for production of /a/ (after /p/) (central column) and its translation into English (right column) Context Catalan control item Translation /p/ + /a/ pala Shovel parles Speak (2nd person sg.) pasta Pasta pare Father passi Come in (3rd person sg.)
174
APPENDIX 3. List of Spanish stimuli (central column) divided by vocalic context (/a/, /o/, /i/) (left column), and translation into English (right column) Vocalic context Spanish stimuli Translation /k/ + /a/ cable cable cafre savage calca trace (3rd person sg.) calcio calcium calva bald spot calza wear shoes (3rd person sg.) cámping camping campus campus cana white hair cáncer cancer canon canon cansa get tired (3rd person sg.) casta chaste castra castrate cata try (3rd person sg.) /k/ + /o/ cobra earn (3rd person sg.) coce boil (3rd person sg.) colmo too much colon colon comba jump rope como eat (1st person sg.) cóndor condor contra against copio copy (1st person sg.) copla copla cosa thing Cosme Cosme (proper name) costa coast costo cost coto property /k/ + /i/ Kike Kike (proper name) Kiko Kiko (proper name) quimba type of shoes quimo chyme quino cinchona tree quinqui delinquent quinto fifth quise want (1st person sg. pret.) quiso want (3rd person sg. pret.) quiste cyst quita remove (3rd person sg.) quítala remove it (imperative, sg.) quite remove (1st sg. pres. subj.) 175
quítelos remove them (imper., pl.) quito remove (1st sg. pres.)
176
APENDIX 4. List of Catalan filler words for production experiment in Catalan (original word in Catalan on the left column, translation on the right column) Word in Catalan Translation into English arbre Tree baixi Go down (3rd person sg) balsa Raft banya Horn bateig Baptism bidet Bidet bossa Bag bota Boot botó Button bruta Dirty cerca Search (3rd person sg) ceres Waxes dàtil Date dutxa Shower escacs Chess estoig Pencil case figa Figue gata Drunk gerra Jar goma Eraser grisa Gray groga Yellow llibre Book lloro Parrot lupa Magnifying glass marró Brown mira Look (3rd person sg.) moixa Female cat morter Mortar muda Mute negre Black neta Clean nina Girl olla Pot pactat Agreed padró Census pagar To pay piadós Compassionate pianista Pianist pis Apartment pita Honk (3rd person sg.) piti Honk (3rd person sg. subj.) Pola Pola (proper name) 177 poma Apple poro Blackhead pota Vomit (3rd person sg.) potra Luck prevé Foresee (3rd person sg.) raspall Brush sastre Tailor savi Wise sentir To feel sobre Envelope soler Tend to Sóller Sóller (name of town) sostre Ceiling sucre Sugar suro Cork susto Fright talar To cut down tallar To carve tassa Cup taula Table tengui Have (3rd person sg. subj.) tigre Tiger tira Strip torre Tower tosca Rough vaga Lazy vàter Toilet vela Candle venir To come vermell Red vine Come (2nd person sg. imp.)
178
APPENDIX 5: List of Spanish filler words for production experiment in Spanish (original word in Spanish on the left column, translation on the right column) Word in Spanish Translation into English adiós Good-bye algo Something antes Before basta Enough bebé Baby boca Mouth bolsa Bag brujo Wizard cesto Basket dedos Fingers foto Photo fuera Outside gato Cat gota Drop gripe Flu hombre Man inca Incan letra Letter lote set mansa docile manta blanket mata kill (3rd person sg., pres.) mejor better mete put in (3rd person sg., pres.) miró look (3rd person sg., pret.) mismo same mosca Fly mundo World muro Wall muslo Thigh nada Nothing nunca Never palpa Feel (3rd person sg.) pana Velveteen papa Pope parte Part pato Duck Pili Pili (proper name) písala Step on it (imperative) píselos Step on them (imperative) pite Whistle (3rd sg., pres. subj.) pito Whistle poda Trim (3rd sg., pres.) 179
polo Pole pone Put (3rd sg., pres.) porto Carry (1st sg., pres.) postor Bidder rato While risa Laugh ristra String rojo Red rosco Bread roll rota Broken saber To know saldré Go out (1st sg., fut. simple) salir To go out salsa Sauce samba Samba santo Saint sarna Mange sienes Temple sobra Leftover tasar To rate taza Cup terrón Cube todo Everything tordo Thrush trazo Stroke trompa Trunk tropa Troop turca Turkish tuvo Have (3rd sg., pret.) uñas Nails vaca Cow vestir To dress
180
APPENDIX 6. Mean and standard deviation values for all analyses conducted on Catalan data
Table (a). Mean (M) and standard deviation (sd) Center of Gravity values (in Hz) per group (Manacor, Artà) and vowel context (/a/, /i/, /o/, /ə/) for Catalan productions only Manacor Artà /a/ M = 2744.77, sd = 807.97 M = 1705.82, sd = 500.49 /i/ M = 3340.38, sd = 833.63 M = 2803.82, sd = 896.47 /o/ M = 1011.86, sd = 273.04 M = 1007.08, sd = 250.24 /ə/ M = 2717.39, sd = 766.12 M = 1488.5, sd = 481.52
Table (b). Mean (M) and standard deviation (sd) Relative Intensity values per group (Manacor, Artà) and vowel context (/a/, /i/, /o/, /ə/) Manacor Artà /a/ M = 12.26, sd = 3.96 M = 10.06, sd = 3.46 /i/ M = 13.88, sd = 3.61 M = 10.22, sd = 3.28 /o/ M = 11.71, sd = 3.63 M = 10.66, sd = 3.52 /ə/ M = 11.79, sd = 3.33 M = 8.52, sd = 3.72
Table (c). Mean (M) and standard deviation (sd) Variance values per group (Manacor, Artà) and vowel context (/a/, /i/, /o/, /ə/) Manacor Artà /a/ M = 2015.73, sd = 481.20 M = 1432.90, sd = 420.09 /i/ M = 2102.18, sd = 448.45 M = 1818.48, sd = 413.63 /o/ M = 838.54, sd = 407.48 M = 824.01, sd = 320.06 /ə/ M = 1847.90, sd = 437.78 M = 1262.91, sd = 352.07
Table (d). Mean (M) and standard deviation (sd) for normalized F1 and F2 data per group (Manacor, Artà, collapsed groups) and consonant context (/k/, /p/) Manacor Artà Collapsed groups /k/ F1 M = 0.96, sd = 0.32 M = 1.09, sd = 0.37 M = 1.03, sd = 0.35 F2 M = −0.005, sd = 0.26 M = −0.19, sd = 0.29 M = −0.09, sd = 0.29 /p/ F1 M =1.27, sd = 0.24 M = 1.15, sd = 0.28 M = 1.21, sd = 0.27 F2 M = −0.41, sd = 0.21 M = −0.34, sd = 0.25 M = −0.37, sd = 0.23
181
APPENDIX 7. Mean and standard deviation values for all analyses conducted on Spanish data
Table (e). Mean (M) and standard deviation (sd) Center of Gravity values (in Hz) per group (Manacor, Artà) and vowel context (/a/, /i/, /o/, /ə/) for Spanish data only Manacor Artà /a/ M = 1549.10, sd = 496.18 M = 1602.86, sd = 472.59 /i/ M = 3053.77, sd = 832.67 M = 2833.85, sd = 895.01 /o/ M = 1017.90, sd = 333.81 M = 978.02, sd = 243.70
Table (f). Mean (M) and standard deviation (sd) Relative Intensity values per group (Manacor, Artà) and vowel context (/a/, /i/, /o/) for Spanish data only Manacor Artà /a/ M = 10.99, sd = 3.47 M = 9.39, sd = 3.64 /i/ M = 12.29, sd = 3.08 M = 10.68, sd = 3.56 /o/ M = 11.32, sd = 3.47 M = 10.73, sd = 3.84
Table (g). Mean (M) and standard deviation (sd) Variance values per group (Manacor, Artà) and vowel context (/a/, /i/, /o/, /ə/) for Spanish data only Manacor Artà /a/ M = 1331.06, sd = 478.80 M = 1403.47, sd = 403.05 /i/ M = 2013.26, sd = 452.34 M = 1933.41, sd = 444.24 /o/ M = 913.08, sd = 490.43 M = 849.47, sd = 357.46
182
APPENDIX 8. Mean and standard deviation values for all within-speaker Catalan- Spanish analyses
Table (h). Mean (M) and standard deviation (sd) Center of Gravity values (in Hz) per language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Manacor productions only Catalan Spanish /a/ M = 2744.77, sd = 807.97 M = 1549.11, sd = 496.18 /i/ M = 3340.38, sd = 833.63 M = 3053.77, sd = 832.67 /o/ M = 1011.86, sd = 273.04 M = 1017.90, sd = 333.81
Table (i). Mean (M) and standard deviation (sd) Relative Intensity values per language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Manacor productions only Catalan Spanish /a/ M = 13.26, sd = 3.96 M = 10.99, sd = 3.47 /i/ M = 13.88, sd = 3.61 M = 12.29, sd = 3.08 /o/ M = 11.71, sd = 3.63 M = 11.32, sd = 3.47
Table (j). Mean (M) and standard deviation (sd) Variance values per language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Manacor productions only Catalan Spanish /a/ M = 2015.73, sd = 481.20 M = 1331.06, sd = 478.80 /i/ M = 2102.18, sd = 448.45 M = 2013.26, sd = 452.34 /o/ M = 838.54, sd = 407.48 M = 913.08, sd = 490.43
Table (k). Mean (M) and standard deviation (sd) Center of Gravity values (in Hz) per language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Artà productions only Catalan Spanish /a/ M = 1705.82, sd = 500.49 M = 1602.86, sd = 472.59 /i/ M = 2803.82, sd = 896.47 M = 2833.85, sd = 895.01 /o/ M = 1007.08, sd = 250.24 M = 978.02, sd = 243.70
Table (l). Mean (M) and standard deviation (sd) Relative Intensity values per language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Artà productions only Catalan Spanish /a/ M = 10.06, sd = 3.46 M = 9.39, sd = 3.64 /i/ M = 10.22, sd = 3.28 M = 10.68, sd = 3.52 /o/ M = 10.66, sd = 3.52 M = 10.73, sd = 3.84
Table (m). Mean (M) and standard deviation (sd) Variance values per language (Catalan, Spanish) and vowel context (/a/, /i/, /o/) for Artà productions only Catalan Spanish /a/ M = 1432.90, sd = 420.09 M = 1403.47, sd = 403.05 /i/ M = 1818.48, sd = 412.63 M = 1933.41, sd = 444.24 /o/ M = 824.01, sd = 320.06 M = 849.47, sd = 357.46 183
REFERENCES
Alcover, A. M. (1920). La darrera eixida de conjugació i altres erbes a diferents pobles de
Mallorca. Bolletí del Diccionari de la Llengua Catalana, 11, 34–59.
Amengual, M. (2011). Spanish and Catalan in Majorca: Are there contact-induced
changes in the Majorcan Catalan vowel system? In L. Ortiz (Ed.), Selected
Proceedings of the 13th Hispanic Linguistics Symposium (pp. 214–223).
Somerville, MA: Cascadilla Proceedings Project.
Amengual, M. (2013). An experimental approach to phonetic transfer in the production
and perception of early Spanish-Catalan bilinguals. University of Texas at
Austin: Doctoral dissertation.
Amengual, M. (2016). The perception and production of language-specific mid-vowel
contrasts: Shifting the focus to the bilingual individual in early language input
conditions. International Journal of Bilingualism, 13(1), 55–79.
Asher, J. J., & García, R. (1986). The optimal age to learn a foreign language. Modern
Language Journal, 53, 334–341.
Baayen, H., Davidson, D., & Bates, D. (2008). Mixed-effects modeling with crossed
random effects for subjects and items. Journal of Memory and Language, 59,
390–412.
Badía Margarit, A. M. (1951). Gramática histórica catalana. Barcelona: Noguer.
Bang, H., Sonderegger, M., Kang, Y., Clayards, M., & Yoon, T. (2015). The effect of the
time course of tonogenesis in Seoul Korean. Proceedings of the 18th ICPhS.
184
Barnils, P. (1933). L’articulació de la k i la g mallorquines. Bolletí del Diccionari de la
Llengua Catalana, 3, 74–79.
Barr, D., Levy, R., Scheepers, C., & Tily, H. (2013). Random effects structure for
confirmatory hypothesis testing: Keep it maximal. Journal of Memory and
Language, 68, 255–78.
Bernat i Baltrons, F. (2001). La fonètica dels parlars baleàrics segons les notes de
Manuel Milà i Fontanals. Interpretació i estudi lingüístic. Universitat de
Barcelona: Doctoral dissertation.
Best, C. T., & Strange, W. (1992). Effects of phonological and phonetic factors on cross-
language perception of approximants. Journal of Phonetics, 20, 305–330.
Bhat, D. N. S. (1974). A general study of palatalisation. In J. H. Greenberg, C. Ferguson
& E. Moravcsik (Eds.), Universals of human language (pp. 47–92). Stanford:
Stanford University Press.
Bibiloni, G. (1983). La llengua dels mallorquins. Anàlisi sociolingüística. Universitat de
Barcelona: Doctoral dissertation.
Birdsong, D., Gertken, L. M., & Amengual, M. (2012). Bilingual Language Profile: An
easy-to-use instrument to assess bilingualism. COERLL, University of Texas at
Austin.
Blas-Arroyo, J. L. (2007). Spanish and Catalan in the Balearic Islands. International
Journal of the Sociology of Language, 79–93.
Bloomfield, L. (1933). Language. New York: University of Chicago Press.
Boersma, P., & Weenink, D. (2011). Praat: doing phonetics by computer [Computer
program]. 185
Bosch, L., & Sebastián-Gallés, N. (2003). Simultaneous bilingualism and the perception
of a language-specific vowel contrast in the first year of life. Language and
Speech, 46(2–3), 217–243.
Bosch, L., Costa, A. & Sebastián-Gallés, N. (2000). First and second language vowel
perception in early bilinguals. The European Journal of Cognitive Psychology,
12, 189–221.
Boyd-Bowman, P. (1954). From Latin to Romance in sound charts. Washington DC:
Georgetown University Press.
Bradshaw, M. (1999). A crosslinguistic study of consonant-tone interaction. Ohio State
University. Doctoral dissertation.
Buckley, E. (2003). The phonetic origin and phonological extension of Gallo-Roman
palatalization. Philadelphia: University of Pennsylvania.
Bybee, J. (1977). Substantive evidence for linearity: Vowel length and nasality in
English. In W. A. Beach, S. Fox, & S. Philosoph (Eds.), Papers from the 13th
regional meeting (pp. 152–64). Chicago: Chicago Linguistic Society.
Bybee, J. (2001). Phonology and language use. Cambridge: Cambridge University Press.
Caimari Frau, F. (1983). El vocalisme de Sineu. Maina, 8.
Calabrese, A. (1993). Palatalization processes in the history of the Romance languages: A
theoretical study. In W. J. Ashby, M. Mithun, G. Perissinotto, & E. Raposo (Eds.),
Linguistic perspectives on the Romance languages (pp. 65–83). Amsterdam: John
Benjamins Publishing Company.
Carignan, C., Shosted, R., Shih, C., & Rong, P. (2011). Compensatory articulation in
American English nasalized vowels. Journal of Phonetics, 39, 668–682. 186
Carrasco, P., Hualde, J. I., & Simonet, M. (2012). Dialectal differences in Spanish
voiced obstruent allophony: Costa Rican versus Iberian Spanish. Phonetica, 69,
149–179.
Casillas, J. V., & Simonet, M. (2016). Production and perception of the English /æ/-/a/
contrast in switched-dominance speakers. Second Language Research, 32(2),
171–195.
Chen, M. (1970). Vowel length variation as a function of the voicing of the consonant
environment. Phonetica, 22, 129–59.
Chen, M. (1973). Predictive power in phonological description. Lingua, 32, 173–191.
Cicres, J. (2011). Los sonidos fricativos sordos y sus implicaciones forenses. Estudios
filológicos, 48, 33–48.
Denes, P. (1955). Effect of duration on the perception of voicing. Journal of the
Acoustical Society of America, 27, 761–64.
Diehl, R. L. (2008). Acoustic and auditory phonetics: The adaptive design of speech
sound systems. Philosophical transactions of the royal society, 363, 965–978.
Dupoux, E., Pallier, C., Sebastián-Gallés, N. & Mehler, J. (1997). A Destressing
“Deafness” in French? Journal of Memory and Language, 36, 406.
Erdozia Mauleón, J. L. (2001). Sakana erdialdeko euskara. Nafarroako Gobernua,
Pamplona/Iruña.
Fathman, A. (1975). The relationship between age and second language productive
ability. Language learning, 25, 245–253.
Flege, J. E. (1988). Factors affecting degree of perceived foreign accent in English
sentences. Journal of the Acoustical Society of America, 84, 70–79. 187
Flege, J. E. (1991). Age of learning affects the authenticity of voice onset time (VOT) in
stop consonants produced in a second language. Journal of the Acoustical Society
of America, 89, 395–411.
Flege, J. E. (1995). Two procedures for training a novel second language phonetic
contrast. Applied Psycholinguistics, 16, 425–442.
Flege, J. E. (2007). Language contact in bilingualism: Phonetic system interactions. In J.
Cole, & J. I. Hualde (Eds.), Laboratory Phonology, 9 (pp. 353–382). Berlin:
Mouton de Gruyter.
Flege, J. E., & Fletcher, K. L. (1992). Talker and listener effects on degree of perceived
foreign accent. Journal of the Acoustical Society of America, 9(1), 370–389.
Flege, J. E. & MacKay, I. R. A. (2004). Perceiving vowels in a second language. Studies
in Second Language Acquisition, 26, 1–34.
Flege, J. E., Munro, M., & MacKay, I. R. A. (1995). Factors affecting strength of
perceived foreign accent in a second language. Journal of the Acoustical Society
of America, 97, 3125–3134.
Flege, J. E., Yeni-Komshian, G. & Liu, S. (1999). Age Constraints on Second-Language
Acquisition. Journal of Memory and Language, 41.
Flege, J. E., Schirru, C., & MacKay, I. R. A. (2003). Interaction between the native and
second language phonetic subsystems. Speech Communication, 40, 467–491.
Forrest, K., Weismer, G., Milenkovic, P., & Dougall, R. N. (1988). Statistical analysis of
word-initial voiceless obstruents: Preliminary data. Journal of the Acoustical
Society of America, 84, 115–123.
Fouché, P. (1924). Phonétique historique du roussillonnaise. Toulouse: Privat. 188
Fowler, C., & Housum, J. (1987). Talkers’ signaling of ‘new’ and ‘old’ words in speech
and listeners’ perception and use of the distinction. Journal of Memory and
Language, 26, 489–504.
Fowler, C., Sramko, V., Ostry, D., Rowland, S., & Hallé, P. (2008). Cross language
phonetic influences on the speech of French–English bilinguals. Journal of
Phonetics, 36, 649–663.
Garrett, A., & Blevins, B. (2009). Analogical morphophonology. In K. Hanson, & S.
Inkelas (Eds.), The nature of the word: Studies in honor of Paul Kiparsky (pp.
527–545). Cambridge, Mass.: MIT Press.
Garrett, A., & Johnson, K. (2013). Phonetic bias in sound change. In A. Yu (Ed.), Origins
of sound change: Approaches to phonologization (pp. 51–97). Oxford, UK:
Oxford University Press.
Gordon, M., Barthmaier, P., & Sands, K. (2002). A cross-linguistic acoustic study of
voiceless fricatives. Journal of the International Phonetic Association, 32, 141–
174.
Goto, H. (1971). Auditory perception by normal Japanese adults of the sounds “L” and
“R”. Neuropsychologia, 9– 317–323.
Grosjean, F. (1999). The bilingual’s language modes. In J. Nicol (Ed.), One mind, two
languages: Bilingual language processing (pp. 1–22). Oxford, Blackwell.
Guion, S. G. (1996). Velar palatalization: Coarticulation, perception, and sound change.
University of Texas at Austin: Doctoral dissertation.
Guion, S. G. (1998). The role of perception in the sound change of velar palatalization.
Phonetica the International Journal of Phonetic Science, 55, 18–52. 189
Guion, S., Flege, J. E., & Loftin, J. (2000). The effect of L1 use on pronunciation in
Quichua–Spanish bilinguals. Journal of Phonetics, 28, 27–42.
Hall, R. (1976). Proto-Romance Phonology. New York: American Elsevier Publishing
Company.
Halle, M. (2005). Palatalization/velar softening: What it is and what it tells us about the
nature of language. Linguistic Inquiry, 36, 23–41.
Harrington, J. (2010). The phonetic analysis of speech corpora. Oxford: Wiley-
Blackwell.
Herrick, D. (2006). Mid vowels and schwa in Eastern Catalan: Five non-Barcelona
dialects. In J. P. Y. Montreuil (Ed.), New Perspectives on Romance Linguistics
(pp. 113–126). Amsterdam: John Benjamins.
Hickey, R. (1985). Velar segments in Old English and Old Irish. In J. Fisiak (Ed.),
Proceedings of the Sixth International Conference on Historical Linguistics (pp.
267–279). Amsterdam: John Benjamins.
Hock, H. (1986). Principles of Historical Linguistics. Berlin: Mouton de Gruyter.
Hoffmann, K. (1982). Avestan language. In E. Yarshater (Ed.), Encyclopaedia Iranica
(pp. 47–62). London.
Hojen, A. & Flege, J. (2006). Early learners’ discrimination of second-language (L2)
vowels. Journal of the Acoustical Society of America, 119, 3072–3084.
Hualde, J. I. (1990). Vowel lengthening in Basque. Folia Linguistica, 24, 269–288.
Hualde, J. I., Simonet, M., & Nadeu, M. (2011). Consonant lenition and phonological
recategorization. Laboratory Phonology, 2(1), 301–329.
Hualde, J. I., Simonet, M., & Torreira, F. (2008). Postlexical contraction of nonhigh 190
vowels in Spanish. Lingua, 118(12), 1906–1925.
Hyman, L. M. (1972). Nasals and nasalization in Kwa. Studies in African Linguistics, 3,
167–206.
Hyman, L. M. (1973). The role of consonant types in natural tonal assimilations. In L. M.
Hyman (Ed.), Consonant types and tone (151–179). Southern California
Occasional Papers in Linguistics 1. Los Angeles: University of Southern
California.
Hyman, L. M. (1976). Phonologization. In A. Juilland (Ed.), Linguistic studies presented
to Joseph H. Greenberg (pp. 407–418). Saratoga: Anma Libri.
Hyman, L. M. (2008). Enlarging the scope of phonologization. UC Berkeley phonology
lab annual report, 382–408.
Hyman, L. M., & Shuh, R. G. (1974). Universals of tone rules: Evidence from West
Africa. Linguistic Inquiry, 5, 81–115.
IBE (Institut Balear d'Estadística). (2010). Encuesta modular de hábitos sociales. Módulo
de conocimientos y usos lingüísticos.
Jacobs, H. M. G. M., & Berns, J. K. M. (2013). Perception, production and markedness in
sound change: French velar palatalization. In D. L. Arteaga (Ed.), Research on
Old French: The state of the art (Studies in natural language and linguistic theory
88) (pp. 107–122). Dordrecht: Springer.
Jakobson, R. (1931). Principles of historical phonology. In A. R. Keiler (Ed.), A reader in
historical and comparative linguistics (pp. 121–138). New York: Holt, Rinehart
& Winston.
Janda, R. D. (2000). Beyond ‘pathways’ and ‘unidirectionality’: On the discontinuity of 191
language transmission and the counterability of grammaticalization. Language
Sciences, 22(4), 265–340.
Janda, R. D., & Joseph, B. D. (2001). Reconsidering the canons of sound-change:
Towards a Big Bang Theory, m.s. OSU.
Janda, R. D., & Joseph, B. D. (2003). On language, change, and language change – Or, of
history, linguistics and historical linguistics. In B. Joseph & R. Randa (Eds.),
Handbook of historical linguistics (pp. 3–180). Oxford: Blackwell.
Johnson, K. (2011). Acoustic and auditory phonetics. Oxford, UK: Wiley-Blackwell.
Jones, D., & Ward, D. (1969). The phonetics of Russian. Cambrige: Cambridge
University Press.
Jongman, A., Blumstein, S. E., & Lahiri, A. (1995). Acoustic properties for dental and
alveolar stop consonants – A cross-language study. Journal of Phonetics, 13,
235–251.
Keel, W. (1982). Atomic phonology and phonological variation. Tübingen: Gunter Narr.
Kerswill, P. (2003). Dialect levelling and geographical diffusion in British English. In D.
Britain & J. Cheshire (Eds.), Social dialectology: In honor of Peter Trudgill (pp.
223–243). Amsterdam: John Benjamins Publishing.
Kirby, J. (2013). The role of probabilistic enhancement in phonologization. In A. Yu
(Ed.), Origins of sound patterns: Approaches to phonologization (pp. 228–246).
Oxford: Oxford University Press.
Kuhl, P. K. (1994). Learning and representation in speech and language. Current Opinion
in Neurobiology, 4, 812–822.
Kuhl, P. K., Williams, K. A., Lacerda, F., Stevens, K. N., & Lindblom, B. (1992). 192
Linguistic experience alters phonetic perception in infants by 6 months of age.
Science, 255(5044), 606–608.
Kuhl, P. K., Stevens, E., Hayashi, A., Deguchi, T., Kiritani, S., & Iverson, P. (2006).
Infants show a facilitation effect for native language phonetic perception between
6 and 12 months. Developmental Science, 9(2), F13–F21.
Labov, W. (1972). Sociolinguistic patterns. Philadelphia: University of Pennsylvania
Press.
Ladefoged, P. (2003). Phonetic data analysis. An introduction to fieldwork and
instrumental techniques. Oxford: Blackwell Publishing.
Ladefoged, P., & Maddieson, I. (1996). The sounds of the world's languages. Oxford,
UK: Blackwell.
Lieberman, P. (1963). Some effects of semantic and grammatical context on the
production and perception of speech. Language Speech, 6, 172–187.
Lindblom, B. (1990). Explaining phonetic variation: A sketch of the H & H theory. In W.
J. Hardcastle & A. Marchal (Eds.), Speech production and speech modeling (pp.
403–439). Kluwer, Dordrecht.
Lleó, C., Benet, A., & Cortés, S. (2009). Límits de la normalització lingüística: Vocals
vulnerables en el català de Barcelona. In J. Kabatek, & C. D. Pusch (Eds.),
Variation, Polyglossie und Standardisierung (pp. 157–180). Aachen: Shaker
(Biblioteca Germànica 7).
Llompart, M. & Simonet, M. (2015). Differential positional neutralization of back vowels
in two Majorcan Catalan subdialects. In The Scottish Consortium for ICPhS 193
(Eds.), Proceedings of the 18th International Congress of Phonetic Sciences.
Glasgow, Scotland.
Lobanov, B. (1971). Classification of Russian vowels spoken by different listeners.
Journal of the Acoustical Society of America, 49, 606–08.
Long, M. H. (1990). Maturational constraints on language development. Studies in
Second Language Acquisition, 12, 251–85.
Lubker, J. (1968). An EMG-cinefluorographic investigation of velar function during
normal speech production. Cleft Palate Journal, 5(1).
Luce, P. A., & Luce J. C. (1985). Contextual effects on vowel duration, closure duration
and the consonant-vowel ratio in speech production. Journal of the Acoustical
Society of America, 78, 1949–1957.
Mack, M. (1989). Consonant and vowel perception and production: Early English–
French bilinguals and English monolinguals. Perception y psychophysics, 46,
187–200.
Malécot, A. (1960). Vowel nasality as a distinctive feature in American English.
Language, 36, 222–29.
Matthews, W. K. (1967). Russian historical grammar. London: Athlone.
Maye, J., Weiss, D. J., & Ashlin, R. N. (2008). Statistical phonetic learning in infants:
Facilitation and feature generalization. Developmental Science, 11, 122–134.
Mayrhofer, M. (1965). Sanskrit grammatik mit sprachvergleichenden erläuterungen.
Berlin.
Melià, J. (1998). Palma i pobles: Les dues cares de la situació lingüística a Mallorca.
Actes de la cinquena Trobada de Sociolingüistes Catalans (Barcelona, 24 i 25 194
d’abril de 1997), 114–131.
Mielke, J. (2004). The emergence of distinctive features. The Ohio State University,
Columbus. Doctoral dissertation.
Miralles i Montserrat, J. (1989). La llengua catalana a Mallorca (1229–1986). In I. Marí
(Ed.), II Congrès Internacional de la Llengua Catalana, vol. 8 (pp. 111–157).
València: Institut de Filologia Valenciana.
Moll, F. de M. (1952). Gramática histórica catalana. Madrid.
Moll, F. de M. (1960). Estática y dinámica del catalán en Mallorca. Papeles de Son
Armadans, L, 161–175.
Moll, F. de M. (1968). Gramàtica catalana referida especialment a les Illes Balears.
Palma: Moll.
Moll, F. de B. (1980). El parlar de Mallorca. Barcelona: Moll.
Moll, K. L. (1962). Velopharyngeal closure in vowels. Journal of Speech and Hearing
Research, 5, 30–37.
Moulton, W. (1967). Types of phonemic change. In P. Friedrich (Ed.), To honor Roman
Jakobson: Essays on the occasion of his seventieth birthday, 2 (pp. 1393–1407).
The Hague: Mouton.
Moyer, A. (1999). Ultimate attainment in L2 phonology: The critical factors of age,
motivation, and instruction. Studies in Second Language Acquisition, 21, 81–108.
Odden, D. (2005). The verbal tone system of Zina Kotoko. In B. K. Schmidt, D. Odden &
A. Homberg (Eds.), Aspects of Zina Kotoko grammar. München: Lincom
Europa. 195
Odden, D. (2005). Introducing phonology. Cambridge introductions to language and
linguistics. Cambridge: Cambridge University Press.
Öhman, S. E. G. (1966). Coarticulation in VCV utterances: Spectrographic
measurements. Journal of the Acoustical Society of America, 39, 151–168.
Otake, T., Hatano, G., Cutler, A., & Mehler, J. (1993). Mora or syllable? Speech
segmentation in Japanese. Journal of memory and Language, 32(2), 258–278.
Oyama, S. (1976). A sensitive period for the acquisition of a nonnative phonological
system. Journal of Psycholinguistic Research, 5, 261–283.
Pallier, C., Bosch L., & Sebastian-Gallés, N. (1997). A limit on behavioral plasticity in
speech perception. Cognition, 64, B9–B17.
Patkowski, M. S. (1990). Age and accent in a second language: A reply to James Emil
Flege. Applied Linguistics, 11(1), 73–89.
Penny, R. (1991). A history of the Spanish language. Cambridge: Cambridge University
Press.
Pieras, F. (1999). Dinámica social del contacto lingüístico en Palma de Mallorca: Actitud
y transferencia fonológica. Doctoral dissertation.
Piper, T., & Cansin, D. (1988). Factors influencing the foreign accent. The Canadian
Modern Language Review, 44(2), 334–342.
Piske, T., Flege, J. E., MacKay, I. R., & Watt, D. (2002). The production of English
vowels by fluent early and and late Italian–English bilinguals. Phonetica, 59, 49–
71.
Pope, M. K. (1934). From Latin to Modern French with especial consideration of Anglo-
Norman. Manchester, England: University of Manchester Press. 196
Port, R. F., & Dalby, J. (1982). Consonant/vowel ratio as a cue for voicing in English.
Perception and Psychophysics, 32, 141–52.
Price, G. (1971). The French Language Present and Past. London: Edward Arnold.
R Core Team. (2015). R: A language and environment for statistical computing. Vienna,
Austria: R Foundation for Statistical Computing. Retrieved from https://www.R-
project. org/.
Radatz, H-I. (2007). De la Mallorca preturística a la moderna: La desdialectització del
català de Mallorca. Els Marges, 81, 29–41.
Ramírez, M., & Simonet, M. (accepted). Acoustics of Majorcan Catalan /ʒ/ in native and
non-native speech.
Recasens, D. (1991). Fonètica Descriptiva del Català. Barcelona: Biblioteca d’Estudis
Catalans.
Recasens, D. (2011). Velar and dental stop consonant softening in Romance.
Diachronica 28, 2, 186–224.
Recasens, D. (2013). On the articulatory classification of (alveolo)palatal consonants.
Journal of the International Phonetic Association, 43, 1–23.
Recasens, D. (2014). Acoustic characteristics of (alveolo)palatal stop consonants, and
velar softening. Journal of Phonetics, 42, 37–51.
Recasens, D., & Espinosa, A. (2006). Articulatory, positional and contextual
characteristics of palatal consonants: Evidence from Majorcan Catalan. Journal of
Phonetics, 34, 295–318.
Recasens, D., & Espinosa, A. (2009). Acoustics and perception of velar softening for
unaspirated stops. Journal of Phonetics, 37, 189–211. 197
Recasens, D., & Espinosa, A. (2010). Lingual kinematics and coarticulation for
alveolopalatal and velar consonants in Catalan. Journal of the Acoustical Society
of America, 127, 3154–3165.
Recasens, D., & Pallarès, M. D. (2001). De la fonètica a la fonologia. Barcelona: Ariel.
Reetz, H., & Jongman, A. (2009). Phonetics: Transcription, production, acoustics and
perception. Winchester, UK: Wiley-Blackwell.
Sampson, R. (1999). Nasal vowel evolution in romance. Oxford: Oxford University
Press.
Schmid, S. (2011). An acoustic analysis of palatal obstruents in two Romance varieties.
In 17th International Congress of Phonetic Sciences, Hong Kong, 17 August
2015–21 August 2011.
Sebastián-Gallés, N., & Bosch, L. (2005). Phonology and bilingualism. In J. F. Kroll &
A. M. B. de Groot (Eds.), Handbook of bilingualism: Psycholinguistic
approaches (pp. 68–87). New York, NY: Oxford University Press.
Sebastián-Gallés, N., & Soto-Faraco, S. (1999). Online processing of native and non-
native phonemic contrasts in early bilinguals. Cognition, 72, 111–123.
Seliger, H., Krashen, S., & Ladefoged, P. (1975). Maturational constraints in the
acquisition of second languages. Language Sciences, 38, 20–22.
Simonet, M. (2010). Dark and clear laterals in Catalan and Spanish: Interaction of
phonetic categories in early bilinguals. Journal of phonetics, 38, 663.
Simonet, M. (2011a). Production of a Catalan-specific vowel contrast by early Spanish–
Catalan bilinguals. Phonetica, 68, 88–110. 198
Simonet, M. (2011b). Intonational convergence in language contact: Utterance-final F0
contours in Catalan–Spanish early bilinguals. Journal of the International
Phonetic Association, 41, 157–84.
Simonet, M. (2014). Phonetic consequences of dynamic cross-linguistic interference in
proficient bilinguals. Journal of Phonetics, 43, 26–37.
Stoel-Gammon, C., Williams, K., & Buder, E. (1994). Cross-language differences in
phonological acquisition: Swedish and American /t/. Phonetica, 51.
Sundara, M. (2005). Acoustic-phonetics of coronal stops: A cross-language study of
Canadian English and Canadian French. Journal of the Acoustical Society of
America, 118 (2), 1026.
Sundara, M. (2006). Production of coronal stops by simultaneous bilingual adults.
Bilingualism: Language and Cognition: 9(1), 97–114.
Suter, R. (1976). Predictors of pronunciation accuracy in second language learning.
Language, Learning, 26, 233–253.
Tahta, S., Wood, M., & Lowenthal, K. (1981). Foreign accents: Factors relating to
transfer of accent from the first language to the second language. Language &
Speech, 24, 265–272.
Thomason, S. G., & Kaufman, T. (1988). Language contact, creolization, and genetic
linguistics. Berkeley: University of California Press.
Thompson, I. (1991). Foreign accents revisited: The English pronunciation of Russian
immigrants. Language Learning, 41(2), 177–204.
Trudgill, P. (1974). The social differentiation of English in Norwich. Cambridge:
Cambridge University Press. 199
Van Coetsem, F. (1988). Loan phonology and the two transfer types in language contact.
Dordrecht/Providence R.I.: Foris Publications.
Veny, J. (1978). Els parlars catalans. Palma de Mallorca: Raixa.
Vidal Alcover, J. (1993). Estudis de literatura catalana contemporània. Barcelona:
Publicacions Universitat de Barcelona.
Weinreich, U. (1968). Languages in contact. Findings and problems. The Hague:
Mouton.
Weinreich, U., Labov, W., & Herzog, M. (1968). Empirical foundations for a theory of
language change. In W. P. Lehmann & Y. Malkeil (Eds.), Directions for
historical linguistics: A symposium (pp. 95–188). Austin: University of Texas
Press.
Werker, J. F. (1994). Cross-language speech perception: Development change does not
involve loss. In J. C. Goodman & H. C. Nusbaum (Eds.), The development of
speech perception: The transition from speech sounds to spoken words (pp. 93–
120). Cambridge, MA: MIT Press.
Werker, J. F., & Tees, R. C. (1984). Cross-language speech perception: Evidence for
perceptual reorganization during the first year of life. Infant Behavior and
Development Infant Behavior and Development, 7(1), 49–63.
Wheeler, M. W., Yates, A., & Dols, N. (1999). Catalan: A comprehensive grammar.
London: Routledge.
Winford, D. (2005). Contact-induced changes: Classification and processes. Diachronica,
22, 373. 200
Zimmerman, S. A., & Sapon S. M. (1958). Note on vowel duration seen cross-
linguistically. Journal of the Acoustical Society of America, 30, 152–3.