Master’s thesis
Heritage language phonotactics
Word-medial cluster syllabification in Albanian heritage
speakers in Greece
Submitted by
Aikaterini Iliopoulou
Supervisor: Prof. Ioanna Kappa
Rethymno, January 2020
To my grandparents, Panayiotis Iliopoulos and Kaiti Iliopoulou,
My first and forever teachers
Acknowledgments
First of all, I want to thank all my teachers at the University of Crete for the support I received and for all
the knowledge I gained during the course of my studies. It has truly been an amazing journey. Special
thanks go to my supervisor, Prof. Ioanna Kappa, who believed in me and has always been there for me, to
advise, guide, and support.
I would also like to thank my family, Yannis, Thomai and Eva, for being behind every little step in my life
that has brought me where I am today, especially my dear dad, Yannis, who has been an inspiration to
me, and who seems to never get tired of supporting my dreams-and reading my drafts. I am particularly
grateful to Eva, my personal photoshop designer.
A sincere thank you to Orestis who has always offered me space and support to dream big, comforted me
whenever I felt like a failure, and put up with my moods.
Finally, I am deeply indebted to all the amazing heritage speakers and Albanian immigrants who
volunteered their time to participate in this study, which would not have been possible without them.
I am privileged to have met you all. The best is yet to come!
Table of Contents
1.
Introduction ........................................................................................................................ 3
2.
Heritage speakers and their grammars............................................................................. 5
2.1 Who is a heritage speaker ................................................................................................... 5
2.2 Imbalanced bilingualism and incomplete acquisition .......................................................... 6
2.3 Heritage speakers’ phonology and phonetics ...................................................................... 9
2.4 Albanian as a heritage language in Greece ....................................................................... 16
2.5 Conclusion .......................................................................................................................... 19
3.
Syllable and phonotactics ................................................................................................ 20
3.1 The syllable in phonology................................................................................................... 20
3.2 Standard Albanian syllable and phonotactics .................................................................. 24
3.3 Standard Modern Greek syllable and phonotactics ........................................................ 28
3.4 Conclusion .......................................................................................................................... 32
4.
Experimental Design ........................................................................................................ 33
4.1 Goals and predictions ........................................................................................................ 33
4.2 Participants ........................................................................................................................ 33
4.3 Experimental stimuli .......................................................................................................... 36
4.4 Experimental procedure .................................................................................................... 40
4.5 Limitations ......................................................................................................................... 40
5.
Discussing the data .......................................................................................................... 41
5.1 The data for each cluster type ........................................................................................... 41
5.2 Participants and their data ................................................................................................ 55
5.3 Discussion of the data in the framework of Optimality Theory (OT)................................. 61
5.4 Interpretation of the results .............................................................................................. 71
6.
General discussion and concluding remarks-implications for future research .............. 75
References ................................................................................................................................... 77
APPENDIX I - STANDARD ALBANIAN AND STANDARD MODERN GREEK CONSONANTS
(PHONEMES AND ALLOPHONES) ..................................................................................................... 92
APPENDIX II - EXPERIMENTAL STIMULI ...................................................................................... 93
APPENDIX III - TABLE OF DATA ................................................................................................... 95
1
Abstract
This thesis studies the metalinguistic phonotactic knowledge in Albanian heritage speakers whose
dominant language is Greek, aiming to investigate whether phonotactics in a heritage language can
be incompletely acquired and/or attrited. To this end, a group of 6 Albanian heritage speakers who
were raised in Greece and a control group of 2 Albanian immigrants who moved to Greece in
adulthood participated in a three-consonant word-internal cluster syllabification task, syllabifying 66
nonce-words that contained clusters allowed by Albanian phonotactics, but disallowed by the
phonotactics of Modern Greek. The great between-subjects and within-subjects variability in the
results of both groups suggests incomplete acquisition of heritage phonotactics by heritage speakers,
as well as some degree of attrition in the first-generation Albanian immigrants. I argue that this
variability is attributed to the use of Multiple Parallel Grammars (Kiparsky, 1993; Anttila, 2002a,
2002b; Anttila and Cho, 1998; Revithiadou and Tzakosta 2004a, 2004b; Tzakosta, 2004, among others),
which is indicative of incomplete acquisition and non-native ultimate attainment in the phonotactic
knowledge of heritage speakers, while the use of Multiple Parallel Grammars by first generation
immigrants can suggest first language attrition of phonotactics.
2
1. Introduction
Language acquisition and bilingualism have long been of interest in linguistic studies. Heritage
language acquisition, i.e. bilingual acquisition in a context where the one language has a minority
status, has been a growing research field. Heritage speakers, as minority language speakers, are
widely studied by sociolinguistics, while education studies have also shown an interest on migrant
populations and minority languages research, since findings on different linguistic needs are
important for pedagogical practices. In the last decades, there has been a growing interest for heritage
linguistics in the fields of theoretical linguistics and psycholinguistics, since theoretical and empirical
findings on the acquisition of a variety of language pairs, that are studied in heritage language
acquisition, can shed light on language development as well as on universal and language-specific
aspects of language (Montrul, 2016).
Despite the growing interest linguistics have shown in heritage language research, heritage
phonology remains understudied. Findings in heritage phonology indicate impaired segmental and
suprasegmental perception and production and some interference from (and sometimes to) the
heritage speaker’s dominant language. However, very little research has been done to investigate the
nature of heritage phonotactic acquisition. To my knowledge, there has been only one study on
phonotactics so far, carried out by Shelton et al. (2017), who found significant interference of the
dominant language (English) on heritage Spanish diphthong syllabification.
This thesis aims to provide data that will contribute to the elucidation of the nature of heritage
phonotactic acquisition, adding to the findings of this vastly understudied field. To this end, the
present study examined the syllabification of word-medial three-consonant clusters, an issue with no
previous findings that I am aware of. The goal was to test for dominant language interference in the
syllabification patterns of heritage speakers but, also, to provide some primitive understanding of the
nature of phonotactic acquisition in a context of insufficient input, such as the context of heritage
language acquisition, finding evidence that support or refute the claim that heritage languages are
incompletely acquired (e.g Montrul, 2008, 2016). The subjects tested were bilinguals, heritage
speakers of Standard Albanian, who were born in Greece or moved to Greece during childhood, thus
Standard Modern Greek became their dominant language. A control group of two Albanian
immigrants, who immigrated to Greece at the age of 19 and speak Greek as a second language, was
also tested in order to compare findings.
The present thesis has the following outline:
In chapter 2, I present the definition and description of heritage speakers, imbalanced bilingualism
and incomplete acquisition. After that, I present a comprehensive overview of all linguistic research
on heritage phonetics and phonology that has been conducted so far. Finally, I provide some
information on the status of Albanian as a heritage language in Greece today.
In chapter 3, syllable structure and phonotactics of Standard Albanian and Standard Modern Greek
are presented.
In chapter 4, I present the experimental design and procedure, as well as the participants’
background, the stimuli used in the experiment and the limitations which arose.
3
In chapter 5, I discuss the experimental findings in detail, first grouped by cluster type and then
grouped by participant, and present a detailed description and some generalizations within the
theoretical framework of Optimality Theory (Prince and Smolensky, 1993; McCarthy and Prince,
1993a, 1993b), followed by discussion and a possible explanation of the findings, following the
Multiple Parallel Grammars model (Kiparsky, 1993; Anttila, 2002a, 2002b; Anttila and Cho, 1998;
Revithiadou and Tzakosta 2004a, 2004b; Tzakosta, 2004, among others).
Chapter 6 is the conclusion of this thesis, with a general discussion on heritage language
acquisition, the findings of this study and implications for future research.
Finally, there is a comprehensive list of the works cited in this thesis, followed by three Appendices
with comprehensive tables of Standard Albanian and Standard Modern Greek consonants (phonemes
and allophones) (Appendix I), the experimental stimuli (Appendix II) and the experimental data of the
present study (Appendix III).
4
2. Heritage speakers and their grammars
Infants are capable of acquiring as many languages as they hear in their environment and, typically,
the ultimate attainment of first language acquisition is the adult native speaker proficiency level
(Montrul, 2016). Growing up exposed to two or more languages is called bilingual first language
acquisition (De Houwer, 2009; Meisel 1994, 2001), while acquisition of one or more additional
languages that starts later in time than the acquisition of first language(s) is called second language
acquisition (Ellis, 1989). Heritage language 1 acquisition is early bilingual acquisition that takes place in
an environment where the one language is a culturally or ethnolinguistically minority language while
the other language is a majority language, spoken by the larger part in a predominantly monolingual
community (Kondo-Brown, 2006; Silva-Corvalán, 1994; Montrul, 2008, 2016; Polinsky, 2008, 2011,
2018, among others).
2.1 Who is a heritage speaker
According to Polinsky (2018:9; italics mine), “A heritage language speaker (for short, heritage
speaker) is a simultaneous or sequential (successive) bilingual whose weaker language corresponds to
the minority language of their society and whose stronger language is the dominant language of that
society”.
A minority language may be the native language of immigrant families (or of just one of the parents
in bicultural families), or the language of communities who speak an indigenous, national or regional
language in a place where a different majority language is spoken 2 (Montrul, 2016). The minority
language status is ascribed to “lower social, cultural and political status related to factors surrounding
immigration or colonialization” (ibid:14) and not necessarily to demographics. In any case, a language
can have the status of a majority or a minority language, depending on the given social context. For
example, Spanish is a widely spoken language, which is a majority language in Spain as well as in
various Southern American countries. At the same time, Spanish is a minority language in the United
States, spoken by immigrant families and heritage speakers (Montrul, 2016).
Migration naturally leads to bilingualism and language contact. Migrant families who are speakers
of a minority language are subject to language shift, especially when they move to a predominantly
monolingual country (Holmes, 1992). Getting a good command of the majority language is crucial for
successful assimilation, as “immigrants who look and sound ‘different’ are often regarded as
threatening by majority group members” (ibid:56). This means that adult immigrants who have to use
the majority language at work and want to be integrated in society become bilinguals. In
predominantly monolingual societies this can result in their abandoning their native language and
using the host country’s majority language even with interlocutors of the same origin or at home
(Holmes, 1992). This is especially true for languages that have a low sociopolitical status (Montrul,
2016). Child immigrants, who have immigrated with their family at a young age, as well as children
born to migrant families in the host country, may communicate in the minority language at the domain
of home and family, before they start school. However, when they start to interact with peers and
teachers at the domain of school, they will have to use the majority language of the country they live
in, and this will become their dominant language outside home (ibid). Additionally, children often
refuse to interact in their mother tongue, as they feel it alienates them from their peers (ibid). As a
1
Other terms used to describe heritage languages are: international, community, immigrant, ethnic, indigenous,
minority, ancestral, third, non-official (Montrul, 2016:13-14).
2
Returnees and international adoptees are also considered to be heritage speakers (Montrul, 2016).
5
result, even when immigrant parents do not cease to use their native language at home, the majority
language “infiltrates the home through the children” (Holmes, 1992:56).
Apart from the social factors mentioned above, when the immigrant community does not take
measures to ensure language maintenance, language shift is unavoidable in future generations, as
the dominant language is used in every institutional domain (Holmes, 1992). Especially in cases where
the immigrant group is isolated, in the sense that the immigrants cannot find many interlocutors to
use their native language with, language maintenance is difficult to be achieved (ibid). This language
shift can happen over only two generations, as the migrants’ children are bilingual and their
grandchildren tend to be monolingual in the majority language (ibid).
Due to the different frequency of use of the two first languages, the heritage language which,
especially in adolescence, is contained in the domain of family (Oh and Fuligni, 2010) is acquired
incompletely or attrited and heritage speakers do not often reach the level of competence and fluency
of a native speaker who lives in a community where the heritage language is a majority language (cf.
2.2). At the same time, the majority language, spoken in all public settings, tends to become the
dominant language of such bilinguals and is processed with ease, compared to the weaker language
(Montrul, 2016). However, heritage speakers may decide to learn or reacquire and further develop
their family language in order to empower their resume, but also because they may feel that this is
their native language and their cultural heritage (ibid). These speakers affirm that “they have lost parts
of it [the heritage language] as they were growing up” (ibid:4).
Research shows that heritage speakers are proficient in the minority language to different extents.
There can be different degrees of fluency, ranging from full fluency to little or no productive ability.
Also, there are differences regarding the extent of proficiency in formal registers, as well as in the
degrees of literacy (Montrul, 2016). Montrul (2016) suggests that “the vast majority of heritage
speakers […] lies in-between these two extremes” (ibid:17; italics mine) of proficiency. These
differences across speakers of heritage languages correlate to the age at which the majority language
is acquired or learnt, to the extent and frequency of use of the heritage language in the family and in
broader settings, and to the extent of instruction and literacy (ibid:18).
To conclude, a heritage language is a minority language due to sociopolitical factors, hence its
infrequency of use. The home languages of migrant families are usually minority languages, since
migrants move to a region where their home language has a minority status and they are forced into
language shift towards the majority language. Furthermore, migrant children, or children born to
migrant families grow up in an environment where the vast majority of input is in the majority
language, using the minority language only in the setting of family. Such speakers are defined as
heritage speakers and they rarely have an ultimate native-like attainment in the heritage language.
2.2 Imbalanced bilingualism and incomplete acquisition
The idealized representation of a bilingual person is that of an individual who has acquired two
languages in infancy and, as an adult, has a balanced proficiency in both languages (Montrul, 2008,
2016). Even though it is possible for the two (or more) languages to be fully acquired with a nativelike command, this is a very rare outcome (Grosjean 1989, 1998). Typically, bilingual linguistic
knowledge and language use are imbalanced (Montrul, 2008) as the two languages have to be equally
used in various domains. While the amount of use of the minority language in limited, given that the
language is only used in the family setting, there is abundant input in the majority language, which is
the predominant language to which the bilingual is exposed, in a variety of contexts, including school
and media in both oral and written modality. In addition to that, many heritage language speakers
6
have a low degree of motivation, being aware that the heritage language is used only in their home
setting, while the majority language is spoken in nearly any other context (Montrul, 2008) and make
language choices that reflect their perceived value of the heritage language (ibid). Additionally, the
wider quality and quantity of input in the dominant language, alongside the input impediment in the
heritage language and enhanced proficiency in the majority language, can guide language preference,
which leads to language choices that favor the use of the majority language (Meisel, 2007), hence its
dominance. In such cases, the dominant language is the one that is more native-like (Meisel, 2001).
An important factor in heritage language maintenance, fluency and proficiency is the age of onset
of bilingualism (Jia and Aaronson, 2003). Bilingualism can be either simultaneous or sequential.
Simultaneous bilingual acquisition, or bilingual/multiple first language acquisition (see Meisel, 2004;
de Houwer, 1995), occurs when infants are exposed to the input of two (or more) languages in their
environment before the age of 3 (Montrul, 2008, 2016). This is usually the case with children of
immigrants (second generation immigrants), children living in bilingual (or multilingual) communities
(e.g. Montreal, Catalonia, Switzerland) and children raised in a setting where each parent has a
different native language and speaks solely their language to the child (one-parent/one-language
strategy), even in cases when the input of the different language comes from a caregiver (Montrul,
2008). In this case, children get less input in their home language than sequential bilingual children
because in simultaneous bilingual children the quantity of input and the time of use are shared by
their two first languages (Montrul, 2008).
L2 acquisition in children who grow up immersed in the L2 environment after the age of 4, when
the foundations of the basic grammar of the first language are already established, tends to have a
native-like ultimate attainment, although this could lead to L1 attrition (subtractive bilingualism)
(Lambert, 1977). Even though this is technically a case of L2 acquisition, it takes place within the critical
period 3 (Lenneberg, 1967), thus being subject to Universal Grammar (Bley-Vroman, 1989). In the
typical case, children of migrant families who acquire a heritage language are subsequent bilinguals,
who receive plenty of exposure in their heritage language in infancy, but the quantity of this input is
reduced to a large extent when heritage speakers start formal schooling, when societal and peer
influences build up an advantage for the majority language. (O’Grady et al., 2011). As the input in their
L2, which is the majority language, increases in frequency and variety of contexts, the use of L1
progressively decreases and becomes domain specific. Thus, development of their L1 stops and their
ability in the heritage language diminishes or stabilizes at a stage where it is not fully developed (Jia
and Paradis, 2014). This results in incomplete acquisition (Montrul 2008, 2016) or L1 attrition (Cook,
2003; de Bot, 2004; Jessner, 2003; Köpke et al., 2007; Polinsky, 2006, 2011; Montrul, 2008, 2010, 2011,
2012; Bylund, 2009 among others) as “input is relevant not only for language acquisition, but also for
language maintenance” (O’Grady et al., 2011: 29, italics mine), so adult heritage speakers tend to lose
elements of their L1 “as a consequence of disuse” (O’Grady et al., 2011: 35) when they use the
heritage language occasionally and/or in restricted contexts. Additionally, studies comparing children
and adult heritage speakers (O’Grady et al., 2011; Polinsky, 2011) have indicated that there seems to
be attrition over the course of a lifespan, since their data imply more evident language regression in
adults.
The outcome of heritage language acquisition depends on the input the child receives in the
minority language during the critical period (Montrul, 2008, 2016; O’Grady et al., 2011) as well as on
3
The critical period ends at about the age of puberty (12-13 years old), when cerebral lateralization is
established (Lenneberg, 1967: 62, 65).
7
social factors. Both simultaneous and sequential bilinguals are prone to an unbalanced language
development. Sequential bilinguals, who had enough exposure in their L1 before the commencement
of L2 acquisition tend to grow up to be more proficient in the heritage language than simultaneous
bilinguals (Allen, 2007; Allen et al., 2006; Montrul 2002, 2008, 2016, among others) 4. Children who
immigrate with their parents tend to maintain a degree of proficiency in their L1, and different
patterns of attrition or incomplete acquisition are observed (Montrul, 2008, 2016; Polinsky, 2006),
depending on the quality and quantity of input in the L1 that they receive in the community to which
they move in. Nevertheless, they might grow up to be receptive bilinguals or overhearers 5 (Au et al.,
2002).
There are also some studies (Ellis, 2006; Yip and Matthews, 2007) claiming that the language
patterns attested in heritage speakers are the outcome of dominant language interference or transfer.
According to Grosjean (2001), both languages are activated at the same time in bilingualism.
Nevertheless, only one of the languages is used at a given time, while the other one is inhibited.
Building on that, Meisel (2007) suggests that dominant language interference can be ascribed to the
unsuccessful inhibition of the dominant language, which results in the grammars of both languages
being used at the same time.
The Incompleteness Hypothesis, proposed by Schachter (1990) (building on Bley-Vroman, 1989)
and further argued for by Sorace (1993), describes the nonnative-like attainment attested in L2
acquisition and the incompleteness of L2 grammars, ascribing that to maturational effects. However,
Montrul (2008) argues that, although incomplete acquisition is typical to late bilingualism, it is also
possible in early bilingual grammars which are acquired before the end of the critical period. What is
more, some of the structural patterns observed in heritage languages are also observed in L2 learners
of the same languages (Schlyter, 1993; Montrul, 2016), while the dominant language exhibits patterns
found in normal L1 development (Schlyter, 1993). This is typical in heritage language acquisition. As
Montrul (2008, 2016) suggests, because of its particular context, heritage language acquisition follows
developmental paths similar to those observed in both monolingual L1 and child L2 acquisition,
concluding that heritage grammars resemble grammars in early stages of language development.
Research suggests that heritage speakers have acquired the core aspects of the heritage language
grammar and vocabulary. However, it is observed that heritage speakers show a “tendency toward
simplification, reduction and reanalysis” (Montrul, 2016:86) and the less fluent a heritage speaker is
the more nonnative patterns emerge in vocabulary, inflectional morphology, syntax, semantics,
discourse and pragmatics (see Montrul, 2016: §3 for a detailed discussion on the grammar of heritage
speakers and a detailed literature overview).
Several studies indicate that many of the patterns observed are due to indirect (Otheguy and
Zentella, 2012; Silva-Corvalán, 1994) or direct (Albirini and Benmammoun 2014; Montrul and Ionin,
2010) transfer from the majority language, since language contact can lead to simplification of a
language (DeGraff, 1999; McWhorter, 2007; Meisel 2011; Silva-Corvalán, 1994; Thomason and
4
Cf. Kupisch et al. (2017) for a discussion of the contradicting view. Kupisch et al. (2017) argue that type of
bilingualism (subsequent or sequential) does not create a disadvantage and that insufficient input does not
always lead to incomplete acquisition or attrition (at least for definiteness effects in the syntax-semanticsdiscourse interface).
5
i.e. heritage speakers with a passive knowledge of the language, who grew up hearing it but not communicating
or being addressed in it. They only retain some receptive language skills, usually in oral language.
8
Kaufman, 2001). On the other hand, there are studies suggesting that language contact can also lead
to complexification, rather than simplification in grammar (McWhorter, 2007; Keel, 2015; Shin, 2014).
However, the patterns of simplification, reduction and reanalysis attested in heritage grammars are
also found in monolingual first language acquisition, suggesting that the heritage speakers’ divergence
from the native adult grammar is due to incomplete acquisition or first language attrition (Montrul,
2016). Evidence for language attrition indicates that the structures of L1 are yet to be stable at the
age of 4 and they can be lost or incompletely acquired (ibid), hence language acquisition has not been
completed by that age and there is a need for a large amount of input, in various contexts and for
many years to follow, in order for the language to develop and stabilize (ibid).
The findings on the incomplete acquisition of heritage languages challenge the assumption that
child language acquisition results in native competence in adulthood (Chomsky, 1981; Crain and
Thornton, 1998) as well as the assumption of solidity of the structure of the native language,
assumptions that take the ideal monolingual native speaker for granted (Montrul, 2016).
Concluding, typical bilingualism is unbalanced. In the case of heritage speakers, the weaker
language is their home (minority) language, while the majority language of their community is their
dominant language. Regardless of whether heritage speakers are simultaneous or sequential
bilinguals, they tend to display nonnative patterns in their grammars, since insufficient input in the
heritage language and narrow contexts of use result in incomplete acquisition or attrition.
2.3 Heritage speakers’ phonology and phonetics
Linguistic research in the field of heritage languages has shown that there are signs of attrition or
incomplete acquisition in the domains of syntax, inflectional morphology, semantics, pragmatics,
discourse and vocabulary. Heritage speakers display nonnative patterns in these domains and they
display different degrees of fluency 6. However, phonetics and phonology seem to be the least affected
linguistic domains in heritage grammars.
Heritage speakers seem to have native-like phonological skills, when compared to L2 learners
matched for morphosyntactic skills. Even heritage speakers with minimum proficiency in the heritage
language who have no productive skills (overhearers) seem to have strong receptive aural skills. On
the other hand, when heritage speakers are compared to native speaker groups or to the baseline 7,
they seem to be significantly less proficient.
2.3.1 Production
Studies in production have indicated some divergence from native phonology/phonetics. Most
researchers studying segmental production in heritage speakers conclude that there is some
interference from the dominant language.
Godson (2003, 2004) reports signs of incomplete acquisition in ten Western Armenians who
immigrated to the US before the age of 8. Godson’s subjects exhibited affected production in 3 of the
five Western Armenian vowels: the two front vowels /i/ and /ε/ and the central vowel /a/, which were
influenced by English, as the acoustical and statistical analysis showed. Comparing the results of these
6
Cf. Montrul (2016) and Polinsky (2018) for comprehensive literature overviews and an extended discussion.
In the study of heritage language acquisition, first generation immigrants are the baseline and serve as input
source for the heritage speakers. Attrition is possible in the baseline, which means that the input heritage
speakers receive from the baseline is not the same as the input they would receive in the homeland (Polinsky,
2018).
7
9
speakers to those of Western Armenian immigrants, who immigrated to the US after the age of 8, as
well as to data from monolingual Western Armenian speakers, Godson showed that the three vowels
in the group of subjects who immigrated before the age of 8 were more influenced by English than
the outputs of the group who immigrated after the age of 8. Keeping in mind that the target words in
the experiment were expected to be acquired by the age of 5 in monolingual children growing up in a
Western Armenian setting, Godson concludes that the grammar of the adult heritage speakers she
investigated is incomplete and probably influenced by L2, in the phonetics domain. Additionally,
Godson’s findings suggest that the age of immigration, at which L1 gets demoted to a minority
language status, plays key role to the degree of attrition or incomplete acquisition. Bullock and Gerfen
(2004) and Louden and Page (2005) had findings similar to those of Godson (2003, 2004). Bullock and
Gerfen (2004) investigated heritage speakers of French in Frenchville, Pennsylvania and found
convergence to the English vowel system, while Louden and Page (2005) studied vowel production in
German heritage speakers in Pennsylvania and found evidence for both convergence to and
divergence from English. What is more, Mayr et al. (2015) presented evidence for convergence of
some (but not all) Welsh vowels towards English. Similarly, Saadah (2011) investigated production of
Palestinian Arabic vowels in 12 heritage speakers of Palestinian Arabic raised in the US, 12 L2 learners
of Arabic with English as their L1, and 6 native Palestinian Arabic speakers who immigrated to the US
after the age of 20. Saadah examined the acoustic values for the Palestinian Arabic vowels /i, u, a, i:,
u:, a:/ in their plain and pharyngealized forms. Heritage speakers in the study exhibited both native
and nonnative performance, as their values were in proximity to those of the native speakers for the
high front vowels, yet they performed closer to the L2 group for /u:/, falling in between the two
comparison groups for the rest of the vowels /u, a, a:/.
Spanish heritage speakers also exhibited production differences from native speakers in the studies
of Rao (2013, 2014, 2015), who looked into spirantization in Spanish intervocalic voiced stops (/b, d,
g/ turn into [β, ð, ɣ]). Rao’s data suggest that there is individual variability in heritage speakers, as the
allophonic production of more proficient heritage speakers in his experiment was more native-like
than the production of heritage speakers of low proficiency. Moreover, the production of /b/ was not
as native-like as the production of /d/ and /g/. Similarly, Henriksen (2015) and Amengual (2016)
studied the production of trill versus tap in Spanish, both concluding that, while the heritage speakers
have acquired the trill/tap distinction, there are articulatory differences from monolingual speakers.
Tse (2016a, 2016b, 2017a, 2017b) showed that, while heritage speakers of Cantonese in Toronto
produce the same phonemic inventory as the baseline, there is some divergence in vowels. Heritage
speakers analyze vowel distinctions using the contrast between tense and lax vowels, a contrast that
is part of English phonology, which advocates for language shift in the domain of phonological
representations (Polinsky, 2018). Moreover, Ronquest (2013) and Alvord and Rogers (2014) found that
heritage speakers of Spanish in the US produce more centralized and shorter unstressed vowels
(compared to vowels in stressed position), indicating interference from English, since vowel reduction
is typical in it. However, Chang et al. (2011) conducted 3 production experiments, investigating the
production of language-internal phonemic contrasts in 15 Mandarin heritage speakers in the US,
comparing them to a group of 6 native speakers of Mandarin who immigrated to the US from Taiwan
and mainland China after the age of 14. Chang and his colleagues investigated the production of back
vowels /o, u, y/, the production of aspirated versus unaspirated plosives, and the production of
retroflex fricative /ʂ/ and alveo-palatal fricative /ɕ/, asking participants to read the target phonemes
inside words. Unlike the other studies mentioned above, all three experiments concluded that the
10
heritage speakers who participated in the study had not lost any of the language internal contrasts of
Mandarin but, also, they were able to discern crosslinguistic contrasts between Mandarin and English.
In addition to that, there are also studies that have implied that bilingual production is
bidirectionally affected, and that this also applies to the case of heritage speakers. In the study of
Baker and Trofimovich (2005), the production of 6 English and 5 Korean vowels by Korean heritage
speakers dominant in English was acoustically distinct from the production of the same vowels by
English and Korean monolinguals. Guion (2003) presented similar results, studying the production of
vowels in Quechua-Spanish bilinguals, dominant in Spanish.
Barlow et al. (2013) and Barlow (2014) offer additional supporting evidence for convergence of the
two grammars in areas where the two languages are quite the same. Their study of Spanish-English
bilinguals who were dominant in English shows that, in onset position, the bilinguals produced a /l/
variant divergent from, though close to the target in both English and Spanish. On the contrary, when
in rhyme position the target-like allophone 8 was produced.
Intriguingly, there is also evidence that language shift is driven by both the effects of language
contact (or transfer from the dominant language) and universal principles, available in the Universal
Grammar. Applebaum and Gordon (2013) suggest that the shift of labialized ejectives towards
voiceless unaspirated and voiced stops and the simplification of the heritage phonemic inventory, as
signified by the fading of contrast between postalveolar /ʃ, ʒ/ and alveopalatal /ɕ, ɕ’, ʑ/ coronal
fricatives in Turkish Circassian in diaspora are due to the universal tendency for ease of articulation.
Polinsky (2018) concludes that bilinguals use a strategy she calls “the good enough strategy”,
where
“speakers bring their knowledge of contrasts in one language to the other only
when such contrasts are useful, while minimizing less contrastive distinctions
and arriving at some kind of compromise in those instances where contrast is
not important” (Polinsky, 2018:141).
Indeed, research shows that heritage speakers tend to spotlight phonological differences between
their two languages, for example, heritage speakers of Polish in Canada (Łyskawa et al., 2016) show a
higher rate of final obstruent devoicing than monolingual Polish, monolingual English and firstgeneration Polish immigrants. Final obstruent devoicing is characteristic of Slavic languages (including
Polish) and the emphasis put on it by heritage speakers could be indicative of a tendency 9 to
overproduce differences between the dominant and the heritage language in heritage grammars
(Polinsky, 2018).
However, when those differences are not considered crucial, either because they are not
informative enough or because they come from a scalar, as opposed to a categorical binary
representation they are not brought into play (Polinsky, 2018). Heritage Russian speakers dominant in
Hebrew produced Russian words using the vowel reduction pattern of standard Russian (Asherov et
al., 2016). Nevertheless, when the heritage speakers produced Russian nonce-words, they used the
binary distinction phonological rule applied in the Hebrew vowel reduction pattern. This could suggest
that they have not acquired the Russian phonological rule for vowel reduction, and they have stored
the Russian words as learnt lexical items, actively applying only the Hebrew rule (Polinsky, 2018). It
8
9
the velarized allophone /ɫ/ in American English, which is different from the Spanish lateral approximant
cf. also Kupisch et al., 2014a for evidence for overapplying gemination in heritage Italian.
11
remains unknown, though, if Asherov and colleagues’ subjects showed signs of interference from
Russian in their Hebrew production (ibid).
Heritage speakers seem to be more conservative in their heritage language than native speakers
living in the country where said language has a majority status or, even, more conservative than firstgeneration immigrants. The studies of Thepboriruk (2015) for heritage Thai in Los Angeles and Kang
and Nagy (2012, 2016) for Seoul Korean in Toronto, have shown that the innovation and ongoing
language change observed in homeland are absent from heritage grammars. Thepboriruk (2015: 155158) concludes that Thai teens model their mothers’, rather than their peers’ speech, as the speech
of elders is considered to be more authentic and serves as a hallmark of their Thainess. Cultural
motivation is also present in Kiezdeutsch 10 (or hood German) (Jannedy and Weirich, 2014; Jannedy et
al., 2015) where “the raising and fronting of /ɔɪ/, a velar realization of /l/, tensing of final <-er>, dental
release of /t/, and most saliently, the realization of the palatal fricative /ç/ as [ɕ] or [ʃ]” (Jannedy et al.,
2015:1), which are present in both production and perception, are adopted even by German
monolingual youths, as a sign of community identity and belonging.
Phonetic/phonological conservatism is also apparent in the dominant language of heritage
speakers, irrespective of their home language. Polinsky (2018: 142-144) discusses an unpublished
study of hers, where she found that heritage speakers of various languages who were dominant in
English “produced significantly fewer unreleased stops than their age-matched counterparts”
(Polinsky, 2018:143). The productions were assessed by 4 native English listeners. The variety of home
languages of those heritage speakers excludes the possibility that this divergence is a result of transfer
from the home language. Polinsky (2018:144) suggests that heritage speakers feel the need to express
themselves in more clarity than the usual speaker, because they have frequent interactions with
nonnative speakers of English in their home setting and their communication is more successful when
word boundaries are clear, with no contractions or omissions. Thus, this divergence could be explained
as a result of their bilingual experience (Polinsky, 2018:144).
Research shows that, in the domains of stress and prosody, heritage speakers do not tend to have
a native-like ultimate attainment. Nevertheless, they still perform better than L2 learners of their
home language. Acquisition of suprasegmental features is considered to occur earlier than acquisition
of segmental features (cf. Peña et al., 2012 for a general discussion and Hua and Dodd, 2000; Hua,
2002 for discussions on Mandarin phonological acquisition). Chang and Yao (2016) compared the tonal
production of native Mandarin speakers, heritage speakers of Mandarin and late L2 learners, in an
acoustic study, concluding that heritage speakers produce suprasegmental contrasts that are
divergent from the contrasts produced by the native speakers and the L2 learners, falling closer to
either group at times. Similarly, Yang (2015) concludes that, unlike native Mandarin speakers, heritage
speakers do not depend on pitch contour to produce tonal patterns, while they also use a smaller pitch
range.
Investigating prosody and pitch contours, Colantoni et al. (2016) found that heritage speakers of
Mexican Spanish did not diverge significantly from long-term immigrants from Mexico. However, they
spotted some more differences in their reading task, than in their semi-spontaneous speech task, a
fact indicative of some effect due to formal educational background. Harris and Gries (2011), who
studied Spanish heritage speakers in the US, report some variation in vowel-length consistent with a
10
A youth dialect in urban areas of Germany, which are mainly populated by families of Turkish or Arabic descent
and German speakers of the same, low socioeconomic background.
12
stress-timed language (like English), as contrasted with a syllable-timed language (like Spanish),
suggesting that there is interference of English on the rhythm of heritage Spanish.
Chen et al. (2014) studied prosody in focus environments, in heritage speakers of Quanzhou
Southern Min who were dominant in Mandarin. Chen et al. (2014) found that the heritage speaker
group they studied produced focus consistent with their heritage language in Quanzhou Southern Min
(with no post-focus compression), as well as post-focus compression consistent with their dominant
language, when speaking it. Their findings suggest no transfer from either the subjects’ heritage or
their dominant language. Similarly, Pan (2007) found minimal transfer from the dominant language
in the same language pair. These findings suggest that, when it comes to focus, even in cases where
there is some interference from the dominant language, this interference is limited, at least for
Quanzhou Southern Min heritage speakers dominant in Mandarin. On the contrary, Van Rijswijk et al.
(2017) found some interference from heritage Turkish to the dominant language (Dutch), when they
investigated the prosody of focus, but they do not conclude whether this is an actual case of transfer
or just a cultural difference. However, Fenyvesi (2005) found that heritage speakers of Hungarian,
dominant in English use prosodic patterns which are consistent with English, in their heritage
Hungarian production. Altogether, findings in the prosody of focus are by far not conclusive, but they
could be indicative of a “prosodic motivation for ‘heritage accent’” (Polinsky, 2018: 152).
When native speakers listen to the oral production of heritage speakers, they tend to conclude that
their production is nonnative, even in cases where they judge excerpts with no pauses and no morphosyntactic errors.
Polinsky (2018: 118-121) discusses an unpublished study she conducted using the matched-guise
technique (Lambert et al., 1960), where 15 Russian native speakers in Moscow listened to sevensecond narratives of 7 heritage speakers, 11 native speakers living in the homeland, 5 highly fluent L2
learners of Russian (with English as their L1) and 8 Russian first-generation immigrants who had lived
in the US for at least 10 years. The raters had to decide if the person on the recording was born in
Russia and if they currently live in Russia and answer “yes”, “no” or “I don’t know”. Polinsky used
excerpts with no hesitations, long breaks or morpho-syntactic errors that could reveal the subjects’
nativeship. Examination of the data revealed that the recordings coming from heritage speakers were
deemed produced by speakers living abroad in 83.5% of the cases and produced by speakers who were
born abroad in 62.3% of the cases. Interestingly, in 85.7% of the cases of immigrants living in the US
for at least 10 years, the ratings were accurate, thus offering supporting evidence for attrition. Polinsky
(2018) suggests that the heritage speakers and the immigrants in her study were recognized as such
because of their divergent intonation and not because of segmental differences. She also thinks that
it is possible that the heritage speakers’ divergent prosody is due to the divergent prosody they heard
from the first-generation immigrants (baseline) who served as input source during heritage Russian
acquisition.
Kupisch et al. (2014a) had native speakers of German, French and Italian judge the pronunciation
(in naturalistic speech samples) of simultaneous bilingual heritage speakers raised in Germany, France
and Italy (German-French and German-Italian) of L2 learners, and of monolingual native speakers of
the respective languages. They found that speech samples of the majority language were judged to
be native, while speech samples containing the heritage language were judged to be nonnative. Still,
the heritage speakers’ groups performed better than the L2 learners’ groups, and the raters showed
uncertainty when judging heritage language speech samples. Also, Kupisch and colleagues found
correlation between perceived native accent and length of residence in the country where the
13
heritage language has a majority language status, during childhood, as the more years a subject spent
in the homeland, the less they are perceived to be nonnative.
In her dissertation, Bae (2015), states that her subjects’ prosody was native-like. However, the
subjects’ production was sometimes perceived as having a Korean-American accent, which she
attributes to the input they received as children growing up in Korean immigrant families in the US.
Moreover, raters in the study of Au et al. (2008) were highly consistent, perceiving the cohorts of adult
heritage re-learners of Spanish, L2 learners and childhood overhearers as nonnative. Again, the L2
learners were perceived as less native-like than the other three groups. In the study of Knightly et al.
(2003), Spanish overhearers performed better than Spanish L2 learners in producing Spanish
phonemes and narratives. However, Oh et al. (2003, 2010) and Au and Oh (2009) who included Korean
overhearers in their studies, reached a different conclusion as their data indicated that only heritage
speakers who had actually used the language had an advantage over L2 learners in production.
Polinsky (2018) claims that this can be due to the fact that Spanish overhearers were studied in
Southern California, where they grew up as a part of a Latino community, having increased
opportunities for language exposure, unlike Korean overhearers.
Yeni-Komshian et al. (2000) had native Korean speakers evaluate the presence of foreign accent in
recorded productions of Korean native speakers and Korean heritage speakers dominant in English, in
a sentence repetition task. Results indicated that “participants who arrived in the US before the age
of 12 years produced the Korean sentences with an “American” foreign accent” (Flege, 2007:364),
while Koreans who immigrated to the US after age 12 also demonstrated some signs of attrition, as
their perceived accent diverged from the monolingual native one.
Finally, there is also some evidence for loss of morphophonological rules (Vago, 1991). Vago (1991)
found evidence for attrition in a case study of a 36-year-old Hungarian heritage speaker, whose family
immigrated to Israel when she was at the age of 5;10. Vargo’s subject exhibited signs of attrition
(which Montrul (2016) interprets as incomplete acquisition), oversimplifying and overapplying glide
epenthesis between the stem and the possessive suffix in Hungarian nouns, failing to apply
phonological rules in the expected order and exhibiting rule loss in the case of 3 phonological rules
(vowel epenthesis or metathesis in /h/+liquid sequences, t-palatalization and assimilation).
2.3.2 Perception
In perception, it seems that bilingualism offers an advantage in perceiving both cross-language and
language-internal vowel contrasts.
Regarding segmental perception, research has shown that being exposed to the phonetic contrasts
of a language early on in life offers an advantage for discriminating those contrasts in adulthood (cf.
Werker and Tees, 1984; Tees and Werker, 1984, Werker, 1989). The studies of Werker (1989) and Tees
and Werker (1984) have shown that English-speaking overhearers of Hindi retained their ability to
discriminate between Hindi dental and retroflex contrasts which do not have a phonemic status in
English.
In the studies of Au et al. (2002) and Oh et al. (2003), it was shown that perception and VOT
measurements of heritage speakers of Spanish and Korean differed significantly from both the L2
learners’ groups (subjects who were first exposed to the language after childhood) and the native
speakers’ groups, performing more native-like than the L2 learners, but less native-like than the native
speakers. Kim (2012) also found differences in production-but not in perception, comparing the
14
perception and production of Spanish stops /p, t, k, b, d, g/ in 7 Spanish heritage speakers born in the
US and in Spanish native speakers.
Similarly, Lukyanchenko and Gor (2011) and Gor (2014) explored the perception of the hard versus
soft (non-palatalized versus palatalized) stop /t, t’/ and /p, p’/ language internal contrasts in Russian
heritage speakers living in the US, finding no difference in group results between highly proficient
heritage speakers and native speakers, in discrimination tasks including the target stops in word-final
and non-final positions. However, low-proficiency heritage speakers did not have a native-like
perception in any of the conditions, but all heritage speakers outperformed L2 learners in nonword
discrimination.
Chang (2016) studied heritage Korean speakers who were dominant in English, comparing them to
English and Korean monolingual groups. Investigating the bilingual advantage of heritage speakers,
Chang (2016) found that heritage speakers had a native-like perception of Korean contrasts in voiceless
stops and a better-than-native perception of English voiceless stops (which, unlike Korean voiceless
stops, are not obligatorily unreleased) in nonce-words. In the same study, in an English word
perception task, Korean heritage speakers performed better than English native speakers, suggesting
that there might, indeed, be a bilingual advantage in perception. Lee-Ellis (2012), who also studied
the perception of English stops in Korean heritage speakers, dominant in English, observed the same
advantage, finding no transfer from the heritage to the dominant language.
Gor (2014) investigated how low and high proficiency heritage speakers of Russian perform in
conditions of high and low babble noise, comparing those two groups to a group of L2 learners of
Russian. Gor used high and low predictability contexts and found that high proficiency heritage
speakers performed the same as native controls, outmatching highly proficient L2 learners. In contrast,
lower proficiency heritage speakers did not seem to have an advantage over L2 learners, suggesting
that higher proficiency helps take advantage of contextual cues in order to discern speech under
babble noise conditions.
Kim (2015) tested perception of lexical stress in 11 heritage speakers of Spanish, of Mexican
descent, in the US (sequential bilinguals, exposed to English after the age of 5), comparing them to a
group of 47 L2 learners of Spanish, who were native speakers of English, as well as to a control group
of 25 native speakers of Spanish, living in Mexico. The subjects participated in a minimal pairs
discrimination task. The words in the task were disyllabic Spanish words, differing only in the position
of lexical stress and presented in auditory form. Again, the heritage speaker group was more sensitive
to acoustic cues than the L2 learners, performing better than the L2 learners, but still worse than the
native speaker control group.
Yang (2015) 11, had L2 learners of Mandarin and Mandarin heritage speakers participate in a tone
identification task, where they had to discern and differentiate 4 Mandarin tones contained in
sentences, and correspond these sentences to interpretations. While the heritage speaker group
significantly outperformed that of L2 learners, heritage speakers “perceived a narrower pitch range
than [native speakers]” (Yang, 2015:109). The heritage speakers’ perception diverged from native
perception, indicating transfer from dominant English. However, the assumption that divergent tonal
patterns in heritage speakers are due to transfer from a (non-tonal) dominant language should be
further confirmed by relevant studies (Polinsky, 2018).
11
Yang (2015) studied both production and perception in heritage speakers of Mandarin (cf. 2.3.1).
15
Finally, Laleko and Polinsky (2016, 2017) investigated the differences in duration of the
contrastive/anaphoric topic marker -nun in Korean 12 as well as the placement of boundary tone, in
Korean native speakers, heritage speakers of Korean and Korean L2 learners. Once again, heritage
speakers perform worse than native speakers and their performance falls close to that of L2 learners
in anaphoric topics comprehension, while they have a better comprehension of contrastive topics,
which are more prosodically salient. All in all, heritage speakers seem to be sensitive only to “stronger,
more salient prosodic cues” (Polinsky, 2018: 161), and Polinsky (2018) suggests that “[t]he prosodic
defaults in heritage speakers may be more general than the defaults established by native speakers”
(Polinsky, 2018: 162), a suggestion calling for further testing.
Despite growing research in heritage phonetics and phonology, syllable and phonotactics in
heritage grammars remain vastly understudied. To the best of my knowledge, there is only one
relevant study, carried out by Shelton et al. (2017). Shelton and colleagues had 29 heritage speakers
of Spanish, who were dominant in English and 29 Spanish monolinguals participate in a pen-and-paper
Spanish word syllabification task, testing the syllabification of Spanish diphthongs. The researchers
found significant interference of English phonotactics in the heritage language, in all four categories
they studied.
The studies overviewed in the last two sections suggest that phonetics and phonology are subject
to “different developmental schedules for acquisition and loss” (Montrul, 2016:85), as they seem to
be affected in more subtle ways than the other domains. The research conducted so far indicates that
heritage speakers have separate mental representations for their two languages, as expected in
bilingualism. Even so, they tend to accentuate the differences in the two systems and de-emphasize
their congruities. Regarding perception, heritage speakers seem to have an advantage over
monolinguals in segmental perception and, despite some difficulties in the perception of tone, heritage
speakers tend to outperform L2 learners and have close to native or native-like performance. Still,
there is evidence for dominant language interference in segmental and suprasegmental production,
as well as in phonotactics, and a nonnative accent impression, reported by native speakers who judge
heritage language production.
2.4 Albanian as a heritage language in Greece
Albania remained a completely isolated state during the Enver Hoxha regime and, after the fall of
communism in the early 1990s, there was a migration flow from Albania, which resulted in 1/5 of its
population moving abroad (Carletto et al., 2006:768; Manos et al., 2017:29). Due to the country’s
adjacency to Greece, 52.7% (Mattheoudakis et al., 2017:2) of immigrants in Greece where of Albanian
origin (Manos et al. 2017:31), a percentage that translates into 480,804 individuals in the 2011 census
(Manos et al. 2017:29). This means that Albanians are the largest immigrant group in Greece today
and, at the same time, they form a 5% population proportion (Gogonas, 2009). A large population of
Albanian immigrants has lived in Greece for the last three decades, leading to a large number of
second and third generation immigrants of Albanian origin. The child immigrants of the early 1990s,
as well as children born to Albanian parents in Greece during that period are, today, adults who have
grown up in a bilingual setting, speaking Albanian at home and Greek at school and almost any other
12
Differences in duration and placement of boundary tone guide the interpretation of the -nun topic marker
as contrastive or anaphoric. (see Laleko and Polinsky (2016, 2017), Polinsky (2018), Jun (1996:101) and Lee
(2007) for discussions on the -nun topic marker in Korean and the prosodic properties linked to its
interpretations.
16
setting, thus being dominant in Greek, as Greek is the language of public life, while Albanian is the
language of private (i.e. family) life (Gkaintartzi, 2012: 385, 388).
Despite this ethnolinguistic group’s population in Greece, the Albanian language has failed to
remain alive both as a minority language and as a heritage language. This is due to low institutional
support, the educational policies in Greece, the immigrant group’s low status and exclusion and the
group’s high degree of integration and assimilation.
In contrast to the considerable demographical presence of people of Albanian origin, Albanian
associations are few, with very low participation (Gogonas, 2009). For example, only 18% of Albanians
living in Thessaloniki participated in such associations in 2006 (Hatziprokopiou, 2006). The primary
goal of Albanian associations is identity preservation and protection from racism and prejudice
(Gogonas, 2009), while they present an opportunity for the second generation to socialize in an
Albanian environment, getting in touch with their homeland’s language and culture. Nevertheless,
such associations do not provide mother tongue classes, with the associations based in Thessaloniki
and Volos being the only exceptions (Gogonas, 2009), failing thus to fill in the gap created by the Greek
Ministry of Education. These complementary schools are not subsidized or officially recognized by the
Greek state (Maligkoudi, 2009). Their students are mainly second-generation immigrants, born and/or
raised in Greece who have little or no competence in the heritage language (Manos et al., 2017).
The educational policy in Greece does not seem to support minority and heritage languages, as the
main provision made for immigrant students is teaching Greek as a second language in Reception or
Support classes (Gogonas, 2009; Dimakos and Tasiopoulou, 2003; Mitakidou et al., 2007) in order to
assimilate them successfully. In addition to that, the effort to establish 26 intercultural schools did not
pay off, as there was no successful integration of foreign students both linguistically and culturally
(Mitakidou et al., 2007), while there was no awareness-raising amongst the non-immigrant
background student population, since intercultural schools isolated and otherized their students
(Damanakis, 1997). A 1999 ministry decision (Φ1/22/Γ1/720-14/9/99) included the initiation of
heritage language and culture classes (Gaintartzi et al., 2014:3), a measure hardly enacted (Kiliari,
2005), for which the state blames the insufficient interest of immigrant parents on mother tongue
classes (Mitakidou et al., 2007; Gaintartzi et al., 2016; Androulakis et al., 2016).
Moreover, the Greek educational system seems to promote “ethnocentrism and conformity with
monolingual forms” (Androulakis et al., 2016:10) (see also Frangoudaki and Dragona, 1997; Katsikas
and Politou, 1999). Language ideologies in the Greek school suggest that Greek be the only language
of education (Gaintartzi et al., 2014) and underpins the language hierarchies in the EU, seeing that
English, French and German are the only three languages taught in public schools (Dendrinos and
Mitsikopoulou, 2004) and are always taught as second/foreign. Teachers tend to advise immigrant
parents to raise their children monolingual, in fear of the heritage language impeding school learning
and confusing the students (Gaintartzi et al., 2014). Gaintartzi et al. (2014) found that 48.2% of
teachers in their sample think that “heritage languages hinder the learning of the school language”
(Gaintartzi et al., 2014:65), while 52.5% of them believe that heritage language classes should be a
responsibility of the immigrant communities and not of the public school. Even if mother tongue
classes were to be offered at the public school setting, most of the teachers (79.2%) Gaintartzi et al.
(2014) interviewed suggested that they should be extracurricular.
17
On top of the invisibility of heritage languages at school, the stigmatization of the Albanian
language in Greek society forces young Albanians to dissociate from their discredited mother tongue
(Androulakis et al., 2016; Gogonas, 2009, Michail, 2010) and perceive the Albanian ethnolinguistic
vitality as very low (Gogonas, 2009). Besides, young second-generation Albanians communicate
almost exclusively in Greek with their peers -including their siblings (Gogonas, 2009; Androulakis et
al., 2016), and grow up immersed in a Greek speaking environment, which leads to language shift
(Gogonas, 2009; Chatzidaki, 2005; Chatzidaki and Xenikaki, 2012; Maligkoudi, 2010; Michail, 2010;
Tsokalidou, 2005).
At the same time, the heritage language use is reduced to the home and family domain (Gogonas,
2007), especially with older relatives, although Albanian heritage speakers in Greece seem to prefer
using their dominant language even at home (Chatzidaki and Xenikaki 2012). Their linguistic
competence in Albanian is lower than their competence in Greek, predominantly regarding their
literacy skills (Gogonas, 2010). However, Kiliari (2014) reports that primary and secondary school
children of Albanian background wish to develop their skills in their homeland language in order to
communicate with Albanian-speaking friends and family and because the heritage language has a
symbolic value for them, as it is their main link to their Albanian roots. Also, some think that having
skills in Albanian may be of use in the future (ibid). Archakis (2014) affirms that Albanian students
want to integrate into the Greek culture and society in which they are growing up and, at the same
time, they want to preserve their own cultural identity and pride.
Albanian parents wish for heritage language maintenance, in order to preserve and protect their
ethnolinguistic identity and their family ties with kin in the homeland, but also for the possibility of
future repatriation (Gkaintartzi et al., 2016). Nevertheless, they do not seem to invest in it and support
it systematically (Chatzidaki and Maligkoudi, 2013; Androulakis et al., 2016). However, 88.3% of the
Albanian parents Gkaintartzi et al. (2016) interviewed maintain that it is the Greek state school’s
responsibility to have mother tongue classes as part of the curriculum, a practice that would also
increase the Albanian language social status. They mainly reinforce the use of Greek not only to avoid
academic failure of their children due to confusing them (as many teachers suggest) (Mitakidou and
Daniilidou, 2007; Gogonas, 2007; Skourtou, 2002; Androulakis et al., 2016), but also because they
consider Greek to be indispensable for their future social, academic and work life and success
(Gkaintartzi, 2012; Androulakis et al., 2016). Apart from that, during the first decades of Albanian
immigrants’ presence in Greece, Albanian parents wanted their children to integrate and feared that
using Albanian was a reason for children to become marginalized and the target of racist attitudes
(Manos et al., 2017). Nowadays, the stigma seems to be fading and both parents and children
recognize the benefits of being bilingual (ibid). Still, links to the homeland are fading more and more
and children are not interested in Albanian music, literature and films or shows, which would enhance
their involvement in Albanian language and culture (ibid).
All in all, heritage speakers of Albanian in Greece seem to be dominant in the majority language,
having little or no skills in the heritage language. Language ideologies, insufficient language policies in
education and little interest for mother tongue classes, alongside racism and stigmatization of the
Albanian language and its speakers have led to a language shift towards Greek. However, Albanian
heritage speakers are aware of their double identity and students of Albanian origin construct
“multiple and complex identities” (Androulakis et al., 2016:14).
18
2.5 Conclusion
To conclude this chapter, heritage speakers are successive or simultaneous bilinguals whose home
or heritage language has the status of a minority language in their community. They are usually
dominant in their community’s majority language and can have different degrees of proficiency in the
heritage language, ranging from limited or no fluency to full native-like proficiency. Research on
heritage grammars suggests that there is evidence for incomplete acquisition or attrition in adult
heritage speakers, laying the blame on the low quality and quantity of input heritage speakers receive
in their home language as they grow up, especially after the age of 4-5, when they start formal
schooling and, consequently start to interact with peers and nonfamily using the majority language.
Incomplete acquisition or attrition is evident in the domain of morphosyntax. However, heritage
phonetics and phonology seem to be less affected, seeing that heritage speakers’ perception in their
heritage language is near-native or native-like, though divergent from native perception. Heritage
speakers tend to outperform L2 learners of the heritage language in perception tasks. Nonetheless,
studies show divergent production in the heritage language. Heritage speakers’ segmental and
suprasegmental production are different from native and they tend to spotlight the distinctive
contrasts in their two languages, while they play down unimportant contrasts, following a “goodenough strategy” (Polinsky, 2018). In addition, there is evidence for interference from the dominant
language in heritage segmental and tonal production and in heritage intonation and phonotactics.
Interestingly enough, the relevant literature indicates that heritage speakers tend to sound nonnative
to native speakers of their heritage language.
Migrant children and children born to migrant families are typical examples of heritage speakers.
In many cases, migrant families move to a community where their home language has a minority
status and a low prestige. First generation immigrants are forced to become bilingual and use the
community’s majority language in all settings but home. Their children grow up in a majority
language-speaking environment and become dominant in that language. The insufficient input they
receive in their home language, together with peer pressure and lack of motivation for heritage
language use, lead to language choices that favor the majority language, amplifying its dominance.
In Greece, Albanians form the largest and oldest immigrant group. Given that a great number of
this sociolinguistic group arrived in the early 1990s, there are many second and third generation
Albanians living in Greece today. Second generation Albanians fit the profile of heritage speakers. They
are adults, who grew up in a Greek-speaking setting, using Albanian only with family members. Many
Albanian families opted for the use of Greek, even at home, mainly to help their children integrate
and succeed at school, but also because of the low prestige their language has in Greek society. Greek
school language policies discourage use of heritage languages, favoring monolingualism. There is no
state support for mother tongue classes, and teachers advise parents against using Albanian at home.
At the same time, lack of motivation for language maintenance and lack of an Albanian community
that would work to this end, result in language shift.
19
3. Syllable and phonotactics
3.1 The syllable in phonology
3.1.1 Syllable and syllable structure 13
The syllable is a phonological suprasegmental constituent, a prosodic unit imperative for word
prosodic structure. It was first described in linguistic literature by Whitney (1874) and Saussure (1959)
(cf. Goldsmith, 2011) and, although it is not included in generative phonological theory 14 at first, it is
now valued, as phonological research has reached a consensus that the syllable should be used in
phonological analysis.
There is plenty of supporting evidence for syllable (cf. Blevins, 1995:206-209 for a detailed
discussion). The syllable constitutes the domain of phonological processes (e.g. epenthesis and
deletion) as well as the domain of phonotactics, therefore allowing for generalization. In addition, it is
the domain of morphophonological processes (such as reduplication and affixation). Nonetheless,
there is also psycholinguistic evidence for syllable being a unit. Preliterate children, as well as adult
native speakers of unwritten languages divide words into syllables when asked to break them into
units. Interestingly, most speakers of a given language divide a given word into the same number of
syllables. Also, there are types of writing systems that are based on syllables rather than phonemes
(syllabaries), such as the writing systems of Cherokee and Japanese (cf. Walker and Sarbaugh, 1993;
Cushman, 2011 among others, for the writing system of Cherokee, and Iwasaki, 2002; Tsujimura, 2013
among others, for the writing system of Japanese). Besides, the syllable is the unit on which language
games are based. However, some scholars seem to be skeptical about the existence of syllables, since
syllable structure varies across languages and syllabic units are rather abstract, having no phonetic
basis (Ladefoged and Maddieson, 1996).
Cross-linguistically, syllables have an internal structure containing 2 basic constituents (Kurylowicz,
1948; Hockett, 1955): onset and rhyme (cf. (1)). The onset is consonantal, with one or more
consonants, depending on language-specific phonotactics, and optional in some languages, for
example in Greek and Albanian. The rhyme is further divided into nucleus and coda. While the nucleus
is the cross-linguistically obligatory syllable sub-constituent and the only constituent of a syllable
where stress can be assigned, codas can be optional, depending on the syllable types a language
allows. Codas contain one or more, depending on phonotactics, consonants, whereas nuclei contain
vowels or in some languages, for example English, liquids or nasals (syllabic sonorants) 15. Coda-less
syllables are also called open (cf. (1a), while syllables with a coda are called closed (cf. (1b)). Open
syllables of the form CV (consonant-vowel) are universal and unmarked, as they are found across
languages and are the first to be acquired by infants (Jakobson, 1941). On the contrary, closed syllables
of the type VC are universally marked. Languages that have VC syllables in their grammar will also
have every other syllable type, namely CV, CVC and V 16 (Clements and Kayser, 1983: 28-29).
13
This section is based on Nespor (1996, Chapter 7.1), Kager (1999, Chapter 3), Zec (2007), Lass (1984, Chapter
10.3), Blevins (1995) and Kappa (in press, Chapter 4)
14
In Chomsky and Halle’s classic work The Sound Pattern of English (SPE) (Chomsky and Halle, 1968), which was
founding in generative phonology, there are not any phonological units described.
15
For example, the final syllable of the word syllable /sɪ.lə.bl/ has a consonantal nucleus.
16
In more detail, the syllable types found in natural languages are: V, CV, CVC, VC, CCV, CCVC, CVCC, VCC, CCVCC,
and CVCC, where two Cs stand for complex onsets and/or codas. The syllable types found in each language varies.
20
(1) Syllable types (open/closed)
Onset and Rhyme can consist of up to two timing units (Goldsmith, 1976; Clements and Kayser,
1983). Segments typically occupy one timing unit, but long vowels and diphthongs are associated with
two timing units that are situated in a branching nucleus (cf. (2)). In this case, there is no place available
for the coda consonant to occupy, therefore, it is attached directly to the syllable (or to the
phonological word) node as extrasyllabic (appendix) (cf. (3b)). The same thing happens when coda
consists of two or more consonants (cf (3c)). Complex codas are marked cross-linguistically (Bybee,
2001). Languages tend to allow the presence of clusters 17 in onset position, rather than in coda and
the restrictions for each phonological system tend to be stricter for word-internal syllables (cf. (4a-b)
for a representation of a simple and a complex onset). In the case of onset clusters that consist of three
or more consonants, the two last consonants of the cluster are situated in a branching onset,
occupying its two timing units. As there is no place in the onset for more than two consonants to be
attached, the remaining consonant(s) can be directly attached to the syllable node as extrasyllabic
(appendix) (cf. (4c) for a three-consonant onset cluster with an appendix consonant). If extrasyllabic
consonants are not attached to a higher level of prosodic structure they will be deleted through stray
erasure or syllabified in an extra syllable through vowel epenthesis (cf. Steriade, 1982; Clements and
Keyser, 1983; Borowsky, 1986; Itô, 1986, among others, for discussions on the treatment of
extrasyllabic consonants).
(2) Branching nuclei (Adapted from Kappa, 2013a)
17
When sequential consonants are part of the same syllable (tautosyllabic) they form consonant clusters. When
they are not part of the same syllable (heterosyllabic) they form consonant sequences (Gussmann, 2002:94).
21
(3) Consonant attached to coda position (a), consonant attached to syllable node due to branching
nucleus (b), three-consonant coda cluster (c) (Adapted from Kappa, 2013a)
(4) Simple (a) and complex (b) onset, and complex onset with an appendix consonant (c) (a. and
b. are adapted from Kappa, 2013a)
22
3.1.2 Syllabification and Phonotactics
Sonority is a universal constraint that restricts segments in onset and coda position and their
relative position inside the syllabic constituent. Segments are ordered in a sonority scale, originally
proposed by Selkirk (1982) (cf. (5), adapted from Zec, 2007: 178).
(5)
Selkirk (1982) advocates for sonority indices of segments (cf. (6) adapted from Kappa (2013: 57)). The
higher the index, the higher the segment’s sonority, thus all segments with a ≥3 sonority index can
occupy the nucleus position. Selkirk (1982) argues that syllable well-formedness is subject to the
Sonority Sequence Generalization, according to which the sonority in a syllable must peak at the
nucleus, increasing from onset to nucleus and decreasing from nucleus to coda. So, well-formed
consonant clusters should contain segments of increasing sonority when in onset and decreasing
sonority when in coda position.
(6)
Universal Sonority Indices (Selkirk, 1982) 18
Natural Class
Sonority index
vowels
6
glides
5
liquids
4
nasals
3
fricatives/ affricates
2
stops
1
Consonants found between vowels have to be parsed into syllables and, according to the Maximal
Onset Principle (Kahn, 1976:19) a consonant occupies the onset of the posterior syllable to satisfy
universal markedness constraints, conforming to the CV universally unmarked syllable type. So, onsets
18
In more detail, the sonority increases as follows:
/p,t.k/</b,d,g/</f,θ/</v, z, ð/</s/</m,n/</l/</r/</i,u/</e,o/</a/
19
Or, according to others, following the Minimal Onset Satisfaction principle (Steriade, 1982; Roca, 1994; Roca
and Johnson, 1999)
23
are formed by as much consonants as universal constraints (sonority) and language-specific
constraints for syllable formation (phonotactics) allow and only remaining consonants are syllabified
in the coda of the preceding syllable. In addition, a coda should be more sonorous than the following
onset, as dictated by the Syllable Contact Law (Murray and Vennemann, 1983; Vennemann, 1988).
Finally, Steriade (1982: 91 et seq.) suggests different sonority scales for different languages and
argues for the Minimal Sonority Distance principle, i.e. the minimal distance that consonant sequences
in onset or coda position should have on the sonority scale, in order for the cluster to be well-formed
and tautosyllabic 20. This distance may vary across languages.
3.2 Standard Albanian syllable and phonotactics 21
Standard Albanian (henceforth SA) have CV, V, CVC, VC syllables. The language allows clusters of
up to 4 consonants in onset position and up to 3 consonants in word-final coda position. So, the
maximal syllable pattern of SA is (C1)(C2)(C3)(C4)V(C1)(C2)(C3). The nucleus position is occupied by short
vowels and diphthongs (both falling and rising). Open syllables are the most common syllable type in
SA (Dodi (2004:133).
Simple codas in SA may contain any consonant in the SA phonetic system (cf. Appendix I). The
Sonority Scale proposed by Dodi (2004) for SA, as cited in Xhaferaj (2018: 230) is the following:
Voiceless stops (1) < Voiceless fricatives (2) <Voiced stops (3) < Voiced fricatives (4) < Nasals (5) <
Laterals (6) < Rhotics (7) < Glides (8) < High vowels (9) < Mid vowels (10) < Low vowels (11).
Cluster syllabification in SA is subject to universal principles (sonority, Maximal Onset Principle,
Syllable Contact Law) and to language-specific phonotactics. Therefore, the word sapllak /sapɬak/
‘cup’ is syllabified as /sa. pɬak/, because /pɬ/ is a well-formed cluster, syllabified in onset position,
abiding by the Maximal Onset Principle (Kappa, 2013: 63), while the word qendra /cendɾa/ ‘center’ is
syllabified as /cen.dɾa/ (Memushaj, 2010: 144) because the cluster /ndɾ/ is not well-formed and not
phonotactically legal in SA, unlike the well-formed and legal cluster /dɾ/ (Kappa, 2013: 63). On the
contrary, the cluster /ʃtɾ/ is legal in SA phonotactics, despite being ill-formed (/ʃ/ is more sonorous than
/t/, thus there is reversed sonority inside the cluster). So, the word spastroj (/spaʃtɾoi/) ‘to purge’
(Memushaj, 2010: 144) is syllabified as /spa.ʃtɾoi/, following the language’s phonotactics and the
Maximal Onset Principle (Kappa, 2013: 63). In addition to that, the word aksion (/aksion/) ‘act’ is
syllabified as /ak.sion/ (Memushaj, 2010: 145) due to SA phonotactics, in spite of the Syllable Contact
Law.
SA phonotactics allow a great number of consonant clusters, both word-internal and word-final
(Memushaj, 2010: 221-224). The following cluster types (7) are found word-initially in SA (adapted
from Kappa (2013b: 69-70):
20
When a consonant sequence of C1 and C2 does not form a well-formed cluster, C1 and C2 cannot form a complex
onset, so C1 will be attached to the syllable node as an appendix (cf. Kappa, in press:129-131). In this thesis,
reversed sonority and plateau sonority consonant sequences will be referred to as complex onsets and sequences
of one or more consonants at the right syllable margin will be referred to as complex codas, although only the
first consonant of the sequence can be syllabified in coda position and the rest will have to be attached to the
syllable node as appendices (cf. (3c), (4c) in this chapter).
21
This section is based on Kappa (2013:59-63, 69-73) unless stated otherwise.
24
(7) Word-initial two-consonant clusters
Cluster type
a. [stop+liquid]
b. [stop+nasal]
c. [stop+fricative 22]
d. [stop+stop]
e. [fricative+liquid]
f.
[fricative+nasal]
g. [fricative+stop]
h. [fricative+fricative]
Examples
pl/bl, pɬ/bɬ, kl/gl, kɬ/gɬ, pɾ/bɾ, tɾ/dɾ, kɾ, gɾ
*tl/*dl
tm
*pn/*bn, *kn/*gn, *tn/*dn, *dm
ps, ks 23, pj/bj, tj/dj (i.a.)
*bz, *gz (i.a.)
tk
fl/vl, fɾ/vɾ, θɾ
*xl/*ɣl, *xɾ/*ɣɾ, *ðl, *sl/*zl, *sɾ/*zɾ
θn, sn, sm/zm
*vm, *fn/*vn, *xn/*ɣn, *zn, *ɣm, *ðm
sp, st, sk, ʃt, ʃc, ʒb/ʒd, fc (i.a.)
sf/zv, ʃθ, fʃ, sj, zj, fj/vj, θj/ðj
i.
[nasal+nasal]
-
j.
[nasal+liquid]
ml, mɬ, mɾ
k. [nasal+homorganic stop]
nd, ɲg, ŋɟ, mb
There are also some two-consonant clusters which, though well-formed are illegal in SA
phonotactics due to the Obligatory Contour Principle (henceforth OCP, Leben 1973, 1978) which
prohibits two adjacent segments from having the same feature specifications. In SA, OCP forbids
homorganic consonants in clusters. Two-consonant illegal clusters in SA and their place of articulation
are shown in (8) (adapted from Kappa (2013b: 70):
22
“The manner in which stress falls upon the vowels in Albanian shows that it [the glide] should be treated as a
consonant and distinguished from diphthongs formed by the combinations of [i] with other vowels. The varying
length of vowels before the semi-vowel brings one to the same conclusion. Initially a fricative member of the
phoneme is used. Examples: ju ‘you’ [ju], atje ‘there’ [a'tje], prroje ‘streams’ ['prro:je], aj ‘he’ [aj], shkruejne ‘they
write’ [ʃkru: :jn]” Lowman (1932: 281; italics mine). This means that the glide is realized as fricative.
23
Only in Greek loanwords.
25
(8) Two-consonant illegal clusters in SA and their place of articulation
illegal clusters
*tn, dn, tl, ðn, ðl, sl, sɾ, zl
*pm, fm
*fn, θm
place of articulation
[coronal]
[labial]
[+front]
Also, there are two-consonant clusters that are not present word-initially in the phonological
system of SA due to ill-formedness (reversed sonority or sonority plateau). They are shown in (9)
(adapted from Kappa (2013b: 70):
(9) Ill-formed two consonant clusters (word-initial)
[fricative+stop] e.g. sp, st, sk
reversed sonority
[fricative+frivative] e.g. sf/zv, ʃθ
sonority plateau
[stop+stop] e.g. tk
sonority plateau
[nasal+homorganic stop] e.g. nd, ŋg, mb
reversed sonority
SA also allows three-consonant clusters in word-initial and word-internal positions. These cluster
types and some examples for each type are shown in (10) (adapted from Kappa (2013b: 71):
(10) Word-initial and word-internal three consonant clusters in SA
Cluster type
Examples
a. [nasal+homorganic stop+liquid]
ndɾ, ŋgɾ, mbɾ, mbl
b. [nasal+homorganic stop+fricative]
ndj, mbj
c. [s+stop+liquid]
stɾ/zbɾ, ʃtɾ, ʃpɾ, skl, ʒgɾ
d. [s+fricative+liquid]
ʃfɾ, ʃfl, ʒvl
e. [s+stop+fricative]
ʃpj, ʃtj
f. [stop+fricative+stop]
g. [stop+fricative+fricative]
kθj
h. [fricative+fricative+fricative]
zvj
In the cluster types (10a-e) there is reversed sonority between the first two consonants of the
cluster, namely between the nasal and the homorganic stop (10a, b), as the nasal is more sonorous,
and between the sibilant and the stop (10c, e) or the fricative (10d) where the sibilant is more
sonorous. The sonority rises between the second and the third consonants in those clusters, as stops
are less sonorous than liquids (10a, c,) and fricatives (10b, e) and fricatives are less sonorous than
liquids (10d). In (10f, g) there is rising sonority between the two first consonants and the sonority falls
between the fricative and the stop (10f), but there is plateau sonority between the two fricatives (10g).
There is, also, plateau sonority between the three consonants in (10h).
Finally, SA also allows four-consonant clusters in word-initial position. These cluster types and some
examples for each type are shown in (11) (adapted from Kappa (2013b: 71):
26
(11) Word-initial four-consonant clusters in SA
Cluster type
a. [affricate+nasal+homorganic stop+liquid]
b. [fricative+nasal+homorganic stop+liquid]
Examples
tʃndɾ, tʃmbɾ
zmbɾ, ʃndɾ
Clusters in (11) start with a fricative or affricate consonant, followed by a prenasalized stop and a
liquid (Kappa, 2013b: 71).
Word-finally, SA phonotactics allow two-consonant and three-consonant clusters in coda position.
The two-consonant cluster types and some examples for each type are shown in (12) (adapted from
Kappa (2013b: 72):
(12) Word-final two-consonant clusters in SA (coda position)
Cluster type
Examples
a. [liquid+stop]
lp, lt, lk, lc, ɾp, ɾt, ɾk, ɾc, ɾb
b. [liquid+fricative]
lf, ls, ɾf, ɾθ, ɾð
c. [liquid+nasal]
lm, ɬm, ɾm, ɾn
d. [nasal+stop]
mp, nt, ŋk, ŋc
e. [nasal+fricative]
ns, nz, nθ, mf, mθ, mʃ
f. [fricative+stop]
ht, fk, vg, ft, st, zg, ʃt
g. [fricative+fricative]
fʃ
h. [fricative+nasal]
zm
i. [stop+fricative]
pθ, ps, kθ
j. [stop+stop]
tk
k. [stop+nasal]
tm
In cluster types (12a-f) the sonority falls, meaning that clusters (12a-f) are well-formed. On the
contrary, clusters (12g-k) are ill-formed, as the sonority rises (12h, i, k) or there is a sonority plateau
(12g, j).
The three-consonant cluster types and some examples for each type are shown in (13) (adapted
from Kappa (2013b: 72-73):
(13) Word-final three-consonant clusters in SA (coda position)
Cluster type
Examples
a. [liquid+stop+fricative]
ɾgs
b. [nasal+stop+fricative]
ŋks, ŋkθ, ŋgθ
c. [nasal+stop+liquid]
mbɬ
d. [liquid+stop+stop]
ɾkt
e. [nasal+stop+stop]
ŋkt
f. [stop+fricative+stop]
kst, pʃt
In Clusters (13a-e) the sonority falls from C1 to C2 ([liquid+stop] in (13a, d) and [nasal+stop] in (13b,
c, e)), but rises from to C3 ([stop+fricative in (13a, b) and [stop+liquid] in (13c)], there is plateau sonority
between C2 to C3 ([stop+stop] in (13d, e)). In (13f), the sonority rises from the stop (C1) to the fricative
(C2) and falls from the fricative (C2) to the stop (C3).
27
3.3 Standard Modern Greek syllable and phonotactics 24
Standard Modern Greek (henceforth SMG) has CV, V, CVC and VC type syllables and allows clusters
consisting of up to three-consonants in onset position, but disallows branching codas 25 (Pagoni 1993).
It is an open-syllable language (Setatos, 1974:22), so words are syllabified in CV syllables unless there
are word-internal clusters. The maximum syllable type in SMG words is (C1)(C2)(C3)V(C) (e.g. σκνίπες
/sknipes/ ‘gnats’, which is syllabified as /skni.pes/) 26.
Although all consonants present in the phonemic inventory of SMG can be found in onset position
(both word-initially and word-internally) 27, the only consonants allowed in coda position are /s/ and
/n/ 28 in word-final syllables 29. In word-medial position, the only consonants allowed in codas are
coronal sonorants /n, l, r/.
SMG syllabification is subject to language-specific constraints, namely SMG phonotactics and
directionality 30 and to universal principles, namely the Maximal Onset Principle (cf. 3.1.2), the Syllable
Contact Law (cf. 3.1.2), the Structure Preservation constraint 31, as well as the locality condition 32 and
the Principle of Prosodic Licensing 33. Directionality in SMG is right-to-left (Kappa, 1995), meaning that
a syllable is formed starting from the rightest vowel and a maximal onset containing a legal consonant
cluster and continuing with the next syllable on its left. Codas are formed only after the maximal
possible onsets are formed. In addition, word-medial clusters are not only subject to phonological
principles, but they can also be dictated by morphology, for example the word εκτόνωση /ektonosi/
‘decompression’ is syllabified as /ek.’to.no.si/, despite [kt] being a legal cluster in SMG onsets and
despite of the Maximal Onset Principle, because of morphological boundaries (/ek-/ is a prefix). In such
cases, syllabification is informed by morphological criteria 34. However, the syllabification of /ektonosi/
as /e.’kto.no.si/, where phonological criteria prevail, is also acceptable for native speakers of SMG.
24
This section is based on Kappa (2013:63-68, 73-77) unless stated otherwise.
With the exception of some ancient Greek words still rarely used in formal registers and some Katharevousa
(a purist, artificial variety of Greek that was used in formal registers and is no longer in use, cf. Mackridge, 1985:611) words that allow [obstruent+s] codas which were once allowed at word-final position by Ancient Greek
phonotactics (e.g. βλαξ /vlaks/ ‘idiot’, βασιλεύς /vasilefs/ ‘king’) (Steriade, 1982; Itô, 1986). There is also the
exception of loanwords (e.g. τανκς /tanks/ ‘tanks’, φιλμ /film/ ‘film’) (cf. Kappa 2013a and Kappa, in press,
Chapter 4).
26
The maximum syllable type of SMG is (C1)(C2)(C3)V(C1)(C2)(C3), if loanwords are considered to be part of SMG
grammar (e.g. τανκς /taŋks/ ‘tanks’) (Setatos, 1974: 34).
27
Velars /k, g, x, γ/ in onset position are realized as [c, ɟ, ç, ʝ] before front vowels /i, e/.
28
/n/ is rarely found in word-final coda position, as it is often deleted (e.g. των ανθρώπων /tonanθropon/
[tonanθropoø] ‘of the humans’) or epenthesized with an /e/ (e.g. παίζουν /pezun/ [pezune] ‘they play’),
forming an open syllable (Joseph and Philippaki-Warburton, 1987: 236).
29
With the exception of words coming from Katharevousa (cf. footnote 25) which can end in /r/ (e.g. ύδωρ /iðor/
‘water’), loanwords (e.g. κλαμπ /klab/ ‘club’, τσεκ /tsek/ ‘check’) and onomatopoeic exclamations (e.g. αχ βαχ
/axvax/) (cf. Kappa 2013a and Kappa, in press, Chapter 4).
30
The direction of syllabification in a language (Itô, 1986: 10)
31
Lexical rules do not introduce distinctions not present in lexical entries. This means that lexical rules maintain
the underlying representation and any additional rules introducing marked specifications are applied in the
postlexical domain (Kiparsky, 1985)
32
The local domain in which a phonological rule can be applied (cf. Trask, 1996: 210)
33
A well-formedness principle which dictates that every phonological unit must be organized as part of a prosodic
structure (cf. Kenstowicz, 1994; Itô, 1986)
34
Additional examples of morphological criteria working together with phonotactic constraints in SMG are the
words ευστροφία /ef.stro.’fia/ ‘wittiness’, εκστρατεία /ek.stra.’tia/ ‘campaign’, έκπληξη /’ek.pli.ksi/ ‘surprise’
and εκδρομή /ek.ðro.’mi/ ‘excursion’. Apart from /ef/ and /ek/ prefixed words, SMG does not allow fourconsonant clusters in onset position, while it accepts the occurrence of reversed sonority onset-initially, but only
when the first cluster-member is a fricative.
25
28
Similarly, native speakers prefer the syllabification of εκροή /ekroi/ ‘’ as /e.kro.’i/, which abides by the
Maximal Onset Principle, in spite of the fact that /ek-/ is, again, a prefix.
According to Kappa (1995), the sonority of labial and coronal nasals, as well as the sonority of
voiceless fricatives and voiceless stops is distinct. Kappa (1995) revises the sonority scale proposed for
SMG by Malikouti-Drachman (1987), where voiceless stops had the same sonority as voiceless
fricatives and the sonority was the same across nasals. The Minimum Sonority Distance (cf. 3.1.2) for
SMG is ≥4, thus clusters found in onset position are of the type: [stop+nasal/liquid] or
[fricative+coronal nasal/liquid]. The Sonority Scale proposed by Kappa (1995) (cf. 3.1.2) for SMG is
shown in (14) 35.
(14) Sonority Scale for SMG (Kappa, 1995:138)
SMG phonotactics allow various consonant clusters, both word-medially and word-initially (cf.
Setatos, 1974; Kappa, 1995:147 et seq.). The following cluster types (15) are found word-initially in
SMG (adapted from Kappa (2013b:73):
(15) Word-initial two-consonant clusters in SMG
Cluster type
Examples
a. [stop+liquid]
pl/bl, kl/gl, pr/br, tr/dr, kr/gr
*tl/*dl
b. [stop+nasal]
pn, kn, tm
*tn/*dn, *bn, *gn, *dm
c. [stop+fricative]
ps, ks, pç (i.a.)
*bz, *gz, *ts/*dz
d. [stop+stop]
pt, kt
*bd, *gd, *tk, *tp
e. [fricative+liquid]
fl/vl, fr/vr, xl/ɣl, xr/ɣr, θr/ðr
*ðl, *sl 36/*zl, *sr/*zr
f.
θn, xn/ɣn, zm
*sn/*zn, *sm, *ɣm, *ðm, *vm/*fm
[fricative+nasal]
g. [fricative+stop]
sp, st, sk, ft, xt, zb
*vg, *vb, *θt, *ðd, *θk, *xp (i.a)
35
Malikouti-Drachman (1987) and Kappa (1995) do not include the voiced stops [b,d,g] in the Sonority Scale,
because they claim that [b,d,g] do not have phonemic value but are derived from an underlying [nasal+stop]
sequence.
36
/sl/ is an acceptable cluster in loanwords (e.g. Σλαβικός /slavikos/ ‘Slavic’)
29
h. [fricative+fricative]
γð, vð
*fx
i.
[s/z+fricative]
sf, sθ, sx, zɣ, zv
*fs
j.
[nasal+nasal]
mn
Clusters that violate the Minimum Sonority Distance principle (cf. 3.1.2), such as */sn, zn, sm, ɣm,
ðm, vm, zl, zr/ are not allowed word-initially (cf. 15e,f), however [voiceless stop+voiceless fricative]
clusters (but not [voiced stop+voiced fricative] clusters */bz, gz/), which comply with the OCP (cf. 3.2)
are present in onset position (cf. 15c), although they violate the Minimum Sonority Distance principle.
In addition, the clusters in (16) are disallowed by the OCP which, in SMG, does not allow homorganic
clusters in onset position. In addition to that, Drachman (1990) suggests that any articulatory feature
can only appear once in an onset cluster in SMG. This results in consonants with dissimilar place of
articulation in onset clusters. Reversed-sonority clusters, i.e. [fricative+stop] clusters in (15g) and
[s/z+fricative] clusters in (15i), as well as sonority-plateau clusters, i.e. [stop+stop] clusters in (15d),
[fricative+fricative] clusters in (15h) and [nasal+nasal] clusters in (15j) are present in SMG onsets wordinitially. It also has to be mentioned that, in SMG, [obstruent+obstruent] clusters must always agree
in voicing. According to Kappa (2013b) the absence of words starting with /km/ and /xm/ clusters is
just an accidental gap.
(16) Clusters not allowed in SMG onsets due to the OCP (adapted from Kappa, 2013b: 74)
Illegal clusters
place of articulation
* tn, tl 37, ðn, ðl, sl, sr
*pm, fm
*fn, fm, θm
[coronal]
[labial]
[+front]
In word-medial onset position, SMG allows the same cluster types as in word-initial position (17).
The clusters in bold are illegal in word-initial position and accepted in word-medial position, or viceversa.
(17) Word-medial two-consonant clusters in SMG
Cluster type
37
Examples
a. [stop+liquid]
pl/bl, kl/gl, pr/br, tr/dr, kr/gr, tl/dl
b. [stop+nasal]
pn, kn, tm, tn, bn, dm, km
*dn, *gn/*gm, *pm/*bm
c. [stop+fricative]
ps, ks, pç, bʝ, gz, kf, kθ, kx, kð, kv
*bz, *pf, *bv, *pθ, *bð, *px, *bɣ, *gv, *ts/*dz (i.a.)
But cf. the archaic words τλήμων /tlimon/ ‘forbearing’, τλημοσύνη /tlimosini/ ‘forbearance’.
30
d. [stop+stop]
pt, kt
*bd, *gd, *tk, *tp
e. [fricative+liquid]
fl/vl, θl, fr/vr, xl/ɣl, xr/ɣr, θr/ðr
*ðl, *sl/*zl, *sr/*zr
f.
θn, , xn/ɣn, zm, xm, fn, vn, ðn, zn, vm, ðm, ɣm, θm
*sn, *sm, *ɣm, *fm
[fricative+nasal]
g. [fricative+stop]
sp, st, sk, ft, xt, fp, xp
*vg, *vb, *θt, *ðd, *θk *zb
h. [fricative+fricative]
sf, sθ, sx, zv, fx, fs
*zɣ
i.
mn
[nasal+nasal]
The three-consonant cluster types found word-initially and word-medially in SMG and some
examples for each type are shown in (18) (adapted from Kappa (2013b: 76).
(18) Three-consonant clusters in SMG
Cluster type
a. [s+stop+liquid]
b. [s+stop+nasal]
c. [s+fricative+liquid]
d. [s+fricative+nasal]
e. [fricative+stop+liquid]
f. [fricative+fricative+liquid]
g. [fricative+fricative+nasal]
h. [stop+stop+liquid]
i. [stop+fricative/s+liquid]
j. [fricative+stop+fricative/s]
k. [fricative+fricative/s+stop]
l. [fricative+fricative+fricative/s]
Examples
str, spr, skl
skn, spn
sfr, sθr
sθm, sxn
ftr, xtr
fxr, fθr, vɣr, vɣl
vɣn, vɣm, vzm
ptr, ktr, kpl, kpr
kfr, kvr, kθr
ftç, fps
fxt, fsk, fst
sθç, fsç
Clusters (18a-d) start with a sibilant, followed by a well-formed two-consonant cluster of
[stop/fricative+nasal] or [stop/fricative+liquid]. Sibilants are more sonorous than stops and fricatives,
thus the two first members of clusters (18a-d) have reversed sonority. Clusters (18e-k) are the result
of affixation, as the prefixes /ef-/, /ek-/, /ðis-/, /pros-/ are affixed to roots starting with a well-formed
two-consonant cluster.
Four-consonant clusters are scarce in SMG. They are only found word-medially and are always the
result of affixation of the prefixes /ef-/ and /ek-/ to roots starting with three-consonant clusters which
are legal in SMG phonotactics. The four-consonant cluster types found word-medially in SMG and
some examples for each type are shown in (19) (adapted from Kappa (2013b: 77).
31
(19) Four-consonant clusters in SMG
Cluster type
examples
a. [stop+s+stop+liquid]
kstr
b. [f+s+stop+liquid]
fspl, fstr
3.4 Conclusion
To sum up, the syllable is a fundamental constituent of word prosodic structure. It is further
analyzed into onset and rhyme, while the rhyme’s structure can contain two constituents: nucleus and
coda. The nucleus is the only universally obligatory constituent in syllable structure, whereas onset and
coda can be absent, depending on a language’s phonotactics. Of all the possible syllable types (CV,
CVC, V, VC) the open syllable type CV is universal and unmarked. While nuclei typically contain a vowel,
word-initial consonants and consonants found between vowels, word-medially, are syllabified in onset
and coda position abiding by universal and language-specific principles and constraints.
The maximum syllable pattern in SA is (C1)(C2)(C3)(C4)V(C1)(C2)(C3) and the maximum syllable
pattern in SMG (for non-prefixed words of native origin) is (C1)(C2)(C3)V(C). All consonants in SA can be
syllabified in coda position word-finally, forming clusters of up to 3 consonants (there can also be
extra-syllabic consonants directly attached to the syllable node as appendices). On the contrary, SMG
word-final codas (in words of native origin) can only contain one of the two coronals [n] and [s] and
complex codas are not allowed. Word-medially, both languages allow only simple codas. SA codas
contain a nasal, a liquid, a fricative or a stop, while SMG word-medial codas can only contain one
coronal sonorant [n], [l] or [r]. In both languages, syllabification is subject to universal principles (the
Maximal Onset Principle, the Sonority Principle and the Syllable Contact Law) as well as to languagespecific constraints (phonotactics).
It should also be highlighted that, while in the Sonority Scale proposed for SA by Dodi (2004) /s/
and /z/ have the same sonority as all fricatives, being discerned only by voicing, in the Sonority Scale
proposed for SMG by Kappa (1995) /s/ and /z/ are more sonorous than /f, θ, x/ and /v, ð, ɣ/.
32
4. Experimental Design
This chapter treats the experimental part of this study in detail. There is a detailed presentation of
the goals and predictions of the experiment, the background of the participants, the stimuli used in
the experimental procedure and, finally, a detailed description of the experimental procedure and its
limitations.
4.1 Goals and predictions
The goal of this experiment is to explore cluster perception and production in heritage speakers.
To this end, the phonotactic knowledge of heritage speakers of Albanian who are dominant in SMG is
tested via a metalinguistic nonce-word syllabification task, as the heritage speakers’ syllabification
patterns reveal their knowledge of phonotactic constraints in the heritage language (cf. Chapter 3, for
a detailed discussion on syllable and phonotactics).
As the relevant research to this day has presented data indicative of dominant language
interference in the phonetics and phonology of heritage languages (cf. 2.3), the main hypothesis of
this study is that heritage speakers will exhibit patterns suggesting interference of their dominant
language phonotactics with the syllabification of their heritage language. To be more specific, SMG
phonotactic constraints are expected to be transferred to the phonological grammar of Albanian by
Albanian heritage speakers who are dominant in SMG. This means that consonants in SA clusters which
do not comply with SMG phonotactics (cf. 3.2, 3.4) are expected to be syllabified under separate
syllables, influenced by SMG phonotactics (cf. 3.3).
In addition to exploring the patterns of dominant language interference with the heritage language
syllable structure and phonotactics, this experiment aspires to bring forth data that will shed some
light on the constraints that guide syllabification in a language that is acquired in a natural setting
during the critical period, but its grammar is incompletely acquired or attrited later in life (cf. 2.2).
4.2 Participants
5.2.1 Experimental group
The participants in the experimental group were six (N=6, Female=4). Of these, four (N=4) were SA
heritage speakers, namely adult Albanian-Greek bilinguals, who were born in Albania and moved to
Greece during infancy (0;4-2;0). They fit the profile of second-generation immigrants (cf. 2.1), as they
moved to Greece in early childhood, before the acquisition of their L1 was completed and before
formal schooling started. Due to their age at the time of their arrival in Greece, their only interactions
in their L1 were those in the setting of their family and they fit the definition of simultaneous bilinguals
(cf. 2.2). The experimental group includes one (N=1) participant (Aggeliki) who arrived in Greece at
age 11 (11;0), after completing 6 years of formal schooling (elementary school) in Albania (child
immigrant, cf. 2.1). Her mother’s native language is SMG and her father’s native language is SA. She is
a simultaneous bilingual who acquired both languages in infancy, as they were spoken in her family
environment. Until the age of 11, she lived in a SA-speaking environment, with SA being the language
of community, school, institutions and media, thus receiving reduced input in SMG. At the age of 11,
her input in SA was significantly reduced in quantity and quality, as SMG became the language of her
environment and she only interacted in SA with one parent at home. Her profile fits the heritage
speaker definition, as she is an adult whose input in the heritage language became insufficient during
her prepubescent years and within the critical period, resulting in attrition effects in the heritage
language, and in SMG becoming the dominant language of the participant (cf. 2.2). One participant
(N=1) was born in Greece and raised a simultaneous bilingual, in a SMG-speaking environment, with
insufficient input in Albanian. He is a minor (14 years old), but was included in the experiment as he is
33
postpubescent and his language acquisition is, supposedly, complete at the end of the critical period
(cf. note 3) at least concerning phonology.
The language used by all participants in their everyday interactions is SMG. They stated that they
use SMG to communicate with friends and with their siblings. However, almost everyone (with the
exception of Vassiliki and Alex who have always used SMG at home) communicated in SA at home
while they were growing up. Most of them (with the exception of Thodoris) communicate in SMG with
their mother, but not with their father, with whom they continue interacting in SA as adults. All
participants have some relatives in Albania they often communicate with, using the heritage language.
Almost all participants are literate in the heritage language, with one exception (Alex). Two (N=2) of
them have had some mother tongue classes in a formal setting, one (N=1) of them went to school in
Albania, while three (N=3) of them where taught reading and writing at home by family members
(Florida was taught SA in both a formal and a home setting). All participants in the experimental group
grew up in Greece, and underwent formal schooling in the Greek educational system, completing (at
the minimum) Greek secondary education. They grew up in major Greek cities, with no notable
exposure to dialects, SMG being the ambient language and the language of media and institutions in
their environment.
The self-assessments of the participants’ heritage language proficiency varied. Three (N=3)
participants reported an average proficiency in the heritage language, one (N=1) reported advanced
proficiency, and two (N=2) participants reported low proficiency, with receptive skills limited to oral
language and rudimentary productive skills, fitting the description of an overhearer (cf. note 5).
Florida and Participant 1 are related (twins). Although they have shared family background, their
educational backgrounds differ, as Florida has had SA classes in a formal setting, while her sister has
not.
The experimental group’s participation in this experiment was voluntary.
The experimental group’s background information is presented in detail in (20)
(20) Participants’ background information (experimental group)
Gender
Florida
Participant
1
Vasiliki
Thodoris
Aggeliki
Alex
Female
Female
Female
Male
Female
Male
Age (years)
23
23
28
24
40
14
Education
university
vocational
high school
master’s
vocational
junior high
school
Athens
Athens
Denmark
Thessaloniki
Rethymno
Thessaloniki
Athens
Athens
Athens
Thessaloniki
Athens
Thessaloniki
NO
NO
NO
NO
NO
Yes
2;0
2;0
0;4
0;9
11;0
N/A
SA
SA
SA
SA
SMG
SA
Location
Place of
growing up
Born in
Greece
Age of arrival
in Greece
(years;months)
Mother's
native
language
34
Father's
native
language
SA
SA
SA
SA
SA
SA
Siblings
2
2
1
1
1
1
SMG
SMG
SMG
SMG
SMG
SMG
SA
SA
SMG
SA
SA
SMG
SMG
SMG
SMG
SMG
SMG
SMG
SMG
SMG
SMG
SMG
SMG
SMG
SA
SA
SMG
SA
SA
SMG
YES
YES
YES (2-3)
YES (a few)
YES
YES
YES
NO
NO
YES
primary school
in Albania
NO
YES
YES
YES
NO
NO
NO
Average
Average
Overhearer
Advanced
Average
Overhearer
Language
used to
communicate
with mother
(usually)
Language
used to
communicate
with father
(usually)
Language
used to
communicate
with siblings
(usually)
Language
used to
communicate
with friends
Language
used at home
while growing
up
Relatives in
Albania
(communication
in SA)
Studied SA in
a formal
setting
Studied SA in
an informal
setting (e.g.
with a parent)
Reported
proficiency in
Albanian
Notes
Participant
1's twin
Florida's twin
5.2.2 Control group
The participants in the control group are two (N=2, Female=2). They are the baseline: first
generation immigrants who are presumably the input source for heritage speakers (cf. 2.3, note 7).
They are adults who grew up in Albania and immigrated to Greece in early adulthood (at the age of
19). They have lived in Greece for 19-25 years and they use mainly Greek in their everyday
35
interactions. Both baseline participants have two children with whom they communicate in Greek
(Alma reports some code-switching, sometimes using SA too with her children). They both report that
their youngest children have low proficiency in SA, describing the proficiency level of overhearers,
while their eldest have a better level of receptive and productive skills. Evelina’s eldest can speak in
SA, but her proficiency is low, while Alma’s eldest has an advanced proficiency in SA, which she ascribes
to longer visits to the homeland and, consequently, longest exposure to the language. At the time of
the experiment, both baseline participants had undergone some formal teaching in Greek language.
Alma had Greek classes organized by a Non-Governmental Organization (Arsis), for one semester,
while Evelina had 3 weeks of private tutoring with a SMG teacher, at the time of the experiment.
Evelina reports heavy attrition, forgetting words and not being able to pronounce SA phonemes the
way she would like to. She also reported people commenting on her foreign accent during her visits
to Albania. Alma (who lives closer to the border and visits family in Albania at least once a year) also
reports little attrition. She affirms she forgets words, but her pronunciation is perceived as native by
people in her homeland and does not report any comments on her accent.
The control group’s participation in this experiment was voluntary.
The control group’s background information is presented in detail in (21)
(21) Participants’ background information (control group)
Evelina
Alma
Age (years)
44
38
Age of arrival (years)
19
19
Place of residence
Rethymno
Kozani
25
19
SMG
SMG
2 adult
2 children (10 y.o. and 8 y.o.)
Total time in Greece
(years)
Dominant language
(frequency)
Children
Language used with
children
Children's reported SA
proficiency
SMG
SMG and SA (code-switching)
very low (eldest), overhearer
(youngest)
10 y.o. advanced (longer stays in Albania,
longer exposure)/ 8 y.o. very low
Education
high school (Albania)
8-year compulsory education (Albania)
SMG classes
3 weeks (private tutoring)
1 semester (Arsis-NGO)
Reported attrition
(phonetics/phonology)
heavy
little
Notes
people in Albania comment on her
accent (perceived as foreign)
forgetting, but her reported level is good and
her accent is native
4.3 Experimental stimuli
The present research focuses on three-consonant clusters in word-internal position in SA. The
clusters studied are not accepted in SMG. In order to test this study’s hypothesis, five types of wordmedial three-consonant clusters 38 were used: [nasal + stop + liquid], [nasal + stop + fricative], [stop +
38
As there are only two timing units in a syllable onset, in a C1C2C3 consonant sequence, C1 is attached to the
syllable node as appendix (cf. 3.1), thus it cannot be part of the cluster. C2 and C3 form an onset cluster when
there is rising sonority between them. For short, when referring to clusters in this thesis, all consonant sequences
under the same syllable will be included.
36
fricative + fricative], [fricative + fricative + fricative] and [stop + fricative + stop] 39. There were 11
clusters used in total, as shown in (22). Twelve (N=12) nonce-words were created for each cluster,
resulting in 132 stimuli in total. The nonce-words for each cluster were the same as the nonce-words
for each other cluster, with the sole difference of the cluster in the second syllable onset.
(22) Consonant clusters used in this experiment, grouped by cluster type
a. [mbl]
1. [nasal + stop + liquid]
b. [mpl]
c. [ndj]
d. [mbj]
2. [nasal + stop + fricative]
e. [kθj]
3. [stop + fricative + fricative]
4. [fricative + fricative + fricative]
5. [stop + fricative + stop]
f.
[zvj]
g.
h.
i.
j.
k.
[bst]
[bsk]
[kst]
[ksk]
[ksp]
All clusters were word-medial and all nonce-words were disyllabic. Half nonce-words were stressed
in the first syllable (in this case, the penultimate) (N=66) (cf. (23a)) and half nonce-words were stressed
in the final syllable (N=66), in order to control for stress effects (cf. (23b, c)). The stress in SA words
typically falls on the two final syllables and its position is subject to vowel quality and to syllable weight.
Typically, closed final syllables are stressed, unless they contain a schwa [ə] (e.g. [ar.’mik] ‘enemy’, but
[‘a.fər] ‘near’) . Open final syllables are also stressed, provided that they contain a high or a low vowel
(e.g. [cer.‘ʃi] ‘cherry’). If they contain a mid-vowel [e, o, ə], the stress falls on the penultimate (e.g.
[‘ba.bo] ‘midwife’) (Trommer and Grimm, 2004) 40. For this reason, the nonce-words created for this
experiment end in a closed syllable, or an open syllable containing [a] or [i] when they are stressed in
the final syllable. Nonce-words that are stressed in the penultimate have an open final syllable
containing [e] or [o].
32F
The non-target syllable always started with a voiceless stop onset (thus being less marked and less
perceptually strong than the target syllable) and was always an open syllable, to avoid additional
cognitive load and effects due to extra consonants in the sequence (cf. (23)). The target syllable always
had a complex (three-consonant) onset, which was the cluster under investigation and was either open
39
[k] and [c] are allophonic in SMG but have phonemic value in SA. In SMG, [k] becomes [c] before front vowels
[i] and [e], while in SA [k] becomes [c] only before [e] and, still, the palatalization is morphology-dependent (cf.
Lengeris, 2013: 36-37; Nikolou, 2013:82-83). In an effort to avoid effects that are due to this language pair’s
differences in phonetic systems and palatalization rules, [bsk] and [ksk] clusters always become [bsc] and [ksc]
before [i] or [e] in this experiment, as the sequences [ce], [ci] are accepted by both languages.
40
There are numerous exceptions, but Trommer and Grimm (2004) base their algorithm and their consequent
analysis on statistical frequency (cf. Trommer and Grimm, 2004: 20).
37
(23a, c) or closed (23b). Closed syllables always had simple codas. The segments used in the nonceword formation were only segments that SMG and SA share, to avoid any additional effects. A
comprehensive table of all the nonce-words used in the experiment can be found in Appendix II.
(23) Examples of nonce-words used in the experiment:
a. pikthje [‘pi.kθje]
b. tundjep [tu.’ndjep]
c. pambla [pa.’mbla]
The three-consonant clusters used in the experiment, the two-consonant clusters that can result
from their syllabification as heterosyllabic sequences, in two consequent syllables, and whether they
are legal in SMG and SA onsets and codas are shown in (24).
(24) Clusters SMG and SA
Word-initial
SA
Cluster
Word-medial
SA
Word-final
SA
Word-initial
SMG
Word-medial
SMG
Word-final
SMG
NO
NO
NO
nasal + stop + liquid
YES
YES
[mb]
YES
YES
[bl]
YES
YES
NO
[mp]
[pl]
[mbl]
two
member
clusters
that can
result
[mpl]
two
member
clusters
that can
result
NO
(but: [mbɬ])
NO
NO
(but:
prenasalization)
(but:
prenasalization)
NO
NO
YES
YES
NO
YES
NO
NO
NO
NO
YES
NO
YES
NO
NO
NO
YES
YES
NO
YES
YES
NO
NO
NO
NO
NO
NO
(but:
prenasalization)
(but:
prenasalization)
YES
nasal + stop + fricative
YES
YES
NO
[nd]
YES
YES
YES
[dj]
YES
YES
NO
NO
NO
NO
YES
YES
NO
NO
NO
NO
[ndj]
two
member
clusters
that can
result
[mbj]
[n]
38
two
member
clusters
that can
result
[mb]
YES
YES
[bj]
YES
YES
YES
NO
NO
(but:
prenasalization)
(but:
prenasalization)
NO
YES
YES
NO
NO
stop + fricative + fricative
YES
YES
NO
NO
NO
NO
[kθ]
YES
YES
YES
NO
YES
NO
[θj]
YES
NO
NO
[θç]
[θç]
NO
[kθj]
two
member
clusters
that can
result
fricative+ fricative + fricative
YES
YES
NO
NO
NO
NO
[zv]
YES
YES
NO
YES
YES
NO
[vj]
YES
YES
NO
[vʝ]
[vʝ]
NO
[zvj]
two
member
clusters
that can
result
stop + fricative + stop
[bst]
NO
YES
NO
NO
NO
NO
[bsk]
NO
YES
NO
NO
NO
NO
[kst]
NO
YES
YES
NO
NO
NO
[ksk]
NO
YES
NO
NO
NO
NO
[ksp]
NO
YES
NO
NO
NO
NO
[bs]
NO
YES
NO
NO
NO
[s]
[ks]
(only in Greek
loanwords)
YES
YES
YES
YES
[st]
YES
NO
YES
YES
YES
YES
NO
YES
YES
(also [sc])
(but [sc])
NO
(also [sc]:
allophonic)
(also [sc]:
allophonic)
[s]
YES
YES
NO
YES
YES
[s]
YES
two
member
clusters
that can
result
[sk]
[sp]
YES
(loanwords)
[s]
39
4.4 Experimental procedure
During the experiment, the participant saw the stimuli appearing one-by-one on display, while hearing
a recording of the same stimuli list. The stimuli appeared at the center of the display, in white background.
The pace was chosen by the participant, by pressing a button (the right arrow) on their keyboard. Each
nonce-word was played only once and the recording stopped after each stimulus and resumed when the
participant pressed the button to continue to the next stimulus. The participant had to syllabify each
nonce-word into its two syllables orally, according to their linguistic intuition. The syllabification had to be
done immediately after the stimulus was presented and there was no time for the participants to reflect
on their answer. The syllabification provided by the participant was recorded. Background information on
each participant’s background (cf. 4.2) was collected through a questionnaire after the experiment.
In order to eliminate fatigue effects, the stimuli were divided into two groups with sixty-six (N=66)
nonce-words each. Thus, each participant had to syllabify sixty-six (N=66) nonce-words. The nonce-words
were presented in a distinct, pseudorandomized order for each participant, while no successive noncewords containing the same cluster were presented to any of the participants, to avoid any effects from
the successive presentation of a cluster.
The experiment took place at each participant’s home and the data were recorded using Audacity®
(Audacity Team, 2012) and transcribed by the researcher. The transcribed data are presented in
Appendix III and discussed in Chapter 5.
4.5 Limitations
This experiment had two major limitations that impeded a more sophisticated experimental
procedure. First, there were time limitations. As this research is part of a master’s thesis, the
experiment should be completed within a limited timeframe. This limitation, together with the small
volunteer pool and the absence of SA monolingual volunteers (that would, ideally, record the
experimental stimuli and participate in the experimental procedure as a second control group with no
signs of attrition), resulted in the experimental setup described in 4.4, which was the best feasible
setup in this context.
40
5. Discussing the data
This experiment tested the syllabification of three-consonant clusters in disyllabic nonce-words.
This means that the participants heard and read nonce-words containing consonant sequences of
three consonants and had to decide which of those consonants are part of the same syllable. There
are three main possible outcomes:
a) the three consonants can be perceived as a cluster, which means that all three consonants are
either syllabified under σ1, with C1 in the coda and C2 and C3 being attached to the syllable
node as appendices, or syllabified under σ2, with C2 and C3 forming a complex onset 41 and C1
being attached to the syllable node as an appendix.
b) C1 and C2 can be perceived as cluster, syllabified in a complex coda under σ142, while C3 is
syllabified in the (simple) onset of σ2. A complex coda can be present in SA (but not in SMG)
syllables.
c) C2 and C3 can be perceived as cluster, syllabified in a complex onset under σ2 (but, cf. note 41),
while C1 is syllabified in the (simple) coda of σ1. This is in line with SMG syllable structure, as
SMG disallows complex codas, but allows complex onsets (cf. 3.3). However, not all clusters
formed by C2 and C3 of the three-consonant clusters in the stimuli are allowed in SMG onsets
(cf. (24)).
Apart from these three outcomes, a cluster can be simplified by deletion, epenthesis or fusion. Of
these processes, only fusion was present in the dataset presented here, and it was a rare outcome.
The data in this experiment reveal a great variability, both between groups and within each group.
Even more interestingly, there is great individual variability. In this chapter, we discuss the
syllabification patterns for each cluster type, within an Optimality Theory approach, proposing a
constraint ranking for each pattern. At the end of this chapter, we argue for the interaction of multiple
parallel grammars in every subject’s phonological grammar. A comprehensive table of all the
experimental data can be found in Appendix III.
5.1 The data for each cluster type
5.1.1 [nasal+stop+liquid] clusters.
Two clusters of the type [nasal+stop+liquid] were tested. The three-consonant clusters of this type
consisted of a labial nasal [m] (C1), a voiced or an unvoiced homorganic stop [b/p] (C2) and a lateral
liquid [l] (C3). The nasal in the clusters of this type is impressionistically considered to be a labial nasal
and not part of a prenasalised stop. The data were not analyzed phonetically in order to confirm this
assumption.
The data show that [nasal+stop+liquid] sequences are not perceived as a cluster 43 for the vast
majority of heritage speakers. [mbl] is syllabified under σ2 (cf. 25a-b) in 30.56% of times, on average,
and [mpl] is syllabified under σ2 only by Aggeliki (100% of her data). On the other hand, the baseline
syllabifies the three consonants under σ2 (66.67% on average for [mbl] and 58.33% on average for
[mpl]). Interestingly enough, although Alma always (100%) syllabifies the three consonants under σ2,
Evelina only does so for 33.33% of the stimuli for [mbl] and 16.67% for [mpl].
41
When the cluster is ill-formed, C2 will also be an appendix.
In this case, C2 will be an appendix
43
Here, when the three consonants are syllabified as cluster, I assume that the first consonant of the cluster (C1)
is attached to the syllable node (appendix), while the remaining two consonants (C1 and C2) are attached to a
branching onset (cf. 3.1.1).
42
41
The heritage speaker group prefers to syllabify the nasal in coda position and the [stop+liquid] in
the onset of σ2, as a cluster (cf. 25c-d). This is true for 61.11% of the [mbl] stimuli and 69.44% of the
[mpl] stimuli. This is the most frequent syllabification for all heritage speakers: 66.67% for Florida,
participant 1, Vasiliki, and Thodoris, and 100% for Alex in [mbl] stimuli and 66.67% for Florida and
participant 1, 83.33% for Vasiliki and 100% for Thodoris and Alex in [mpl] stimuli, while Aggeliki always
(100%) syllabifies the three consonants as a cluster under σ2, in both cases ([mbl] and [mpl]). In
addition to that, Evelina, who is part of the baseline group, opts for a syllabification of the nasal in the
coda of the first syllable for 16.67% of the stimuli for [mbl] and 33.33% for [mpl].
Two heritage speakers (Florida and participant 1, who are twins) sometimes fuse the nasal and the
stop (cf. 25e). In the case of [m] and [b], the two phonemes have the same place of articulation (labial),
the same voicing (voiced) and they differ in manner of articulation ([m] is nasal and [b] is a stop). After
the fusion, the phoneme in the output is a voiced labial obstruent [b], thus keeping the manner of
articulation of [b]. In the case of [m] and [p], the two phonemes have the same place of articulation
(labial), but different voicing ([m] is voiced, while [p] is voiceless) and manner of articulation ([m] is
nasal and [p] is a stop). After the fusion, the phoneme in the output is a voiced labial obstruent [b],
hence keeping the manner of articulation of [b] and the voicing of [m].
The one control participant (Evelina) syllabifies [mb] in coda position (25f) in 50% of the data for
[mbl]. [mb] (but not [mp]) are legal coda clusters in SA (but not in SMG) (cf. (24)).
Finally, there is voicing of [p] in [mpl] clusters (cf. 25h) in 2.78% of the heritage group data (only in
Vasiliki’s data) and in 25% of the baseline group data (only in Evelina’s data). Voicing is attested only
in nonce-words that are stressed in penultimate. However, the data are not enough to consider this
to be an effect of stress.
Apart from voicing, no effect of stress position or environment was found in this dataset.
(25) a. [ko.mblo] (Alma)
b. [ci.mplek] (Aggeliki)
c. [pam.blo] (Thodoris)
d. [kom.plo] (Thodoris)
e. [ko.blo] (Florida)
f. [komb.lo], [pemp.lik] (Evelina)
h. /tumple/ [tu.mble] (Evelina)
The percentages of each syllabification pattern for each participant and the average for each
pattern are shown in (26) (for [mbl] cluster) and (29) (for [mpl] cluster) for heritage speakers and in
(27) (for [mbl] cluster) and (30) (for [mpl] cluster) for the baseline. The average for the totality of
participants is shown in (28) (for [mbl] cluster) and (31) (for [mpl] cluster). All percentages are rounded
down to the second decimal place.
42
(26) Heritage speakers’ syllabification patterns for [mbl] (%)
Syllabification
Florida
Partic. 1
Vasiliki
Thodoris
Aggeliki
Alex
[.mbl]
[.bl]
[m.bl]
16. 67
16. 67
66. 67
0
33.33
66.67
33. 33
0
66.67
33.33
0
66.67
100
0
0
0
0
100
Average
heritage
30.56
8.33
61.11
[mb.l]
0
0
0
0
0
0
0
(27) Baseline’s syllabification patterns for [mbl] (%)
Syllabification
Evelina
Alma
[.mbl]
[.bl]
[m.bl]
[mb.l]
33.33
0
16.67
50
100
0
0
0
Average
Baseline
66.67
0
8.33
25
(28) Total syllabification patterns for [mbl] (%)
Average
39.58
6.25
47.92
4.16
Syllabification
[.mbl]
[.bl]
[m.bl]
[mb.l]
(29) Heritage speakers’ syllabification patterns for [mpl] (%)
Average
heritage
Syllabification
Florida
Partic. 1
Vasiliki
Thodoris
Aggeliki
Alex
[.mpl]
[m.pl]
0
66.67
0
66.67
0
83.33
0
100
100
0
0
100
[.bl]
33.33
33.33
0
0
0
0
16.67
69.44
11.11
[m.bl]
[.mbl]
0
0
0
0
16.67
0
0
0
0
0
0
0
2.78
0
43
(30) Baseline’s syllabification patterns for [mpl] (%)
Syllabification
Evelina
Alma
Average Baseline
[.mpl]
[m.pl]
[.bl]
[m.bl]
16.67
33.33
0
0
100
0
0
0
58.33
16.67
0
0
[.mbl]
50
0
25
(31) Total syllabification patterns for [mpl] (%)
[.bl]
Average
27.08
56.25
8.33
[m.bl]
2.08
[.mbl]
6.25
Syllabification
[.mpl]
[m.pl]
The three-consonant clusters [mbl] and [mpl] are ill-formed. The sonority falls from C1 to C2, rising
again from the stop to the liquid. While the Maximal Onset Principle calls for syllabifying the three
consonants in onset position, the Sonority Sequencing Principle dictates that the nasal be syllabified
as coda in the first syllable, while the second syllable should start with a two-consonant cluster
[stop+liquid] in the onset, that is well-formed, as the sonority rises. Abiding by the Sonority Sequencing
Principle can also lead to syllabification of the nasal and the stop in coda position ([mb.l]). In this case,
the sonority falls in the coda and the cluster is well-formed. However, this option is more marked, as
it leads to a complex coda. Fusion also results in a well-formed cluster ([bl]) in onset position and to a
coda-less, hence unmarked, first syllable. As [mpl] sequences are extremely marked 44, the voiceless
obstruent should be assimilated (voicing). Fusion could also be a strategy to avoid [nasal+voiceless
obstruent] sequences. Cf. (24) for the clusters accepted in word-initial, word-medial, and word-final
position in SA and SMG.
There are no [mpl] or [mbl] clusters in SMG, neither word-initially nor word-medially. There are
[mpl] and [mbl] word-initial clusters in SA. According to Papafilis (2003) there are 11 entries that start
with [mbl] and only 1 entry that starts with [mpl], which is a different orthography for a [mbl]-starting
word 45. The clusters [bl], [pl] are legal in both SMG and SA onsets. In SA, there are 114 entries starting
44
“Nasal leak [airflow through the nose at the time the obstruent begins, that is due to the preceding articulation
of the nasal] and velum raising cooperate to facilitate continuous voicing throughout a nasal-obstruent cluster”
(Kager, 1999:61). So, due to articulatory mechanisms, post-nasal voicing is widespread cross-linguistically (Kager,
1999:61) (cf. Huffman, 1993; Hayes and Stivers, 1995 among others)
45
mblak [mblak] ‘to grow old’ and mplak [mplak] ‘to grow old’
44
with [mb], 46 entries starting with [bl] and 136 entries starting with [pl] (Papafilis, 2003). If the
syllabification patterns are driven by cluster legality in word margins, then [bl] and [pl] clusters are
good candidates for onset position. Indeed, about half of the nonce-words (47.92% of [mbl] stimuli
and 56.25% of [mpl] stimuli, in the merged dataset, cf. (28) and (31)) were syllabified with a nasal in
coda position and a [pl] or [bl] cluster in the onset. In the experimental group, the heritage speakers
opt for [bl] onsets in 72.22% of their data (8.33% due to fusion, 61.11% due to syllabification of the
nasal in coda position and 2.78% due to syllabification of the nasal in coda position and concurrent
voicing of the voiceless stop). In the control group, the baseline opts for [bl] clusters in 16.66% of their
data (8.33% of which is due to syllabification of the nasal in coda, and 8.33% is due to syllabification
of the nasal in coda position and concurrent voicing of the voiceless stop, but only in data coming from
Evelina) and for [pl] clusters in 16.67% of their data (again, only due to syllabification of the nasal in
coda, and only in data coming from Evelina).
5.1.2 [nasal+stop+fricative]
Two clusters of the type [nasal+stop+fricative] were tested. The three-consonant clusters of this
type consisted of a nasal [m/n] (C1), a homorganic stop [b/d] (C2) and a voiced fricative [j] (C3) (cf. note
22). The nasal in the clusters of this type is impressionistically considered to be a labial nasal and not
part of a prenasalised stop. The data were not analyzed phonetically, in order to confirm this
assumption.
The data show that [nasal+stop+fricative] sequences are not perceived as cluster 46 by the vast
majority of heritage speakers. [ndj] is syllabified under σ2 (cf. 32a) in 30.55% of stimuli, on average
(50% by Thodoris, 100% by Aggeliki, 33.33% by Alex and zero times by everyone else). At the same
time, [mbj] is syllabified under σ2 in 16.67% of the stimuli (50% by Thodoris, 100% by Aggeliki and
16.67% by Alex). The baseline mostly perceives [mbj] as a cluster, syllabifying the three consonants
under σ2 in 66.67% of the data on average. However, [ndj] is syllabified as a cluster in only 16.67% of
the baseline data on average. So, heritage speakers accept the [ndj] cluster more than the [mbj]
cluster, while the exact opposite happens in the control group. Although Alma always (100%)
syllabifies the three consonants of [mbj] under σ2, she only syllabifies [ndj] as a cluster in 33.3% of her
data. At the same time, Evelina never syllabifies [ndj] as a cluster, but she opts for a syllabification of
[mbj] under σ2 in 16.67% of her data.
Like in the [nasal+stop+liquid] clusters, the heritage speaker group prefers to syllabify the nasal in
the coda of σ1 and the [stop+fricative] in the onset of σ2, as a cluster (cf. 32b-c). This is true for 66.67%
of the [ndj] stimuli and 69.44% of the [mbj] stimuli. This is the most frequent syllabification for most
heritage speakers: 100% for Florida, participant 1, and Vasiliki, and 66.67% for Alex in [ndj] stimuli
(Thodoris and Aggeliki prefer the syllabification of [ndj] as a three-consonant cluster under σ2) and
100% for Florida and participant 1, 83.33% for Vasiliki, 50% for Thodoris and 83.33% for Alex in [mbj]
stimuli, while Aggeliki always (100%) syllabifies the three consonants under σ2. Additionally, Evelina,
who is part of the baseline group, opts for a complex coda with a [nasal+stop] cluster in the coda of σ1
(cf. 32d) (83.33% for [ndj] stimuli, but 50% for [mbj] stimuli). Two heritage speakers also syllabify the
nasal and the stop in a complex coda, in σ1 (Thodoris, in 16.67% of his data for [ndj] and Vasiliki, in
16.67% of her data for [mbj]). Interestingly enough, Alma, a control participant, syllabifies the nasal
46
Here, when the three consonants are syllabified as cluster, I assume that the first consonant of the cluster (C1)
is attached to the syllable node (appendix), while the remaining two consonants (C1 and C2) are attached to a
branching onset (cf. 3.1.1).
45
in coda position in 66.7% of her data for [ndj], but always (100%) syllabifies [mbj] as a cluster, under
σ2.
No effect of stress position or environment was found in this dataset.
(32) a. [ce.mbje] (Aggeliki)
b. [kon.djo] (Vasiliki)
c. [cem.bje] (Alex)
d. [tund.je] (Evelina)
The percentages of each syllabification pattern for each participant and the average for each
pattern are shown in (33) (for [ndj] cluster) and (36) (for [mbj] cluster) for heritage speakers and in
(34) (for [ndj] cluster) and (37) (for [mbj] cluster) for the baseline. The average for the totality of
participants is shown in (35) (for [ndj] cluster) and (38) (for [mbj] cluster). All percentages are rounded
down to the second decimal place.
(33) Heritage speakers’ syllabification patterns for [ndj] (%)
Syllabification
Florida
Partic. 1
Vasiliki
Thodoris
Aggeliki
Alex
[.ndj]
[n.dj]
[nd.j]
0
100
0
0
100
0
0
100
0
50
33.33
16.67
100
0
0
33.33
66.67
0
Average
heritage
30.55
66.67
2.78
(34) Baseline’s syllabification patterns for [ndj] (%)
Syllabification
Evelina
Alma
Average
Baseline
[.ndj]
[n.dj]
[nd.j]
0
16.67
83.33
33.3
66.7
0
16.67
41.67
41.67
(35) Total syllabification patterns for [ndj] (%)
Syllabification
[.ndj]
[n.dj]
[nd.j]
Average
27.08
60.42
12.5
(36) Heritage speakers’ syllabification patterns for [mbj] (%)
Syllabification
Florida
Partic. 1
Vasiliki
[.mbj]
[m.bj]
[mb.j]
0
100
0
0
100
0
0
83.33
16.67
Thodoris Aggeliki
50
50
0
100
0
0
Alex
16.67
83.33
0
Average
heritage
16.67
69.44
2.78
46
(37) Baseline’s syllabification patterns for [mbj] (%)
Syllabification
Evelina
Alma
Average
Baseline
[.mbj]
33.33
100
66.67
[m.bj]
16.67
0
8.33
[mb.j]
50
0
25
(38) Total syllabification patterns for [mbj] (%)
Syllabification
Average
[.mbj]
37.50
[m.bj]
54.17
[mb.j]
8.33
Like the clusters of the type [nasal+stop+liquid] (cf. 5.1.1), the three-consonant clusters [ndj] and
[mbj] are ill-formed. The sonority falls from C1 to C2, rising again from the stop to the fricative. While
the Maximal Onset Principle calls for syllabifying the three consonants in onset position, the Sonority
Sequencing Principle dictates that the nasal be syllabified as coda in the first syllable, while the second
syllable should start with a two-consonant cluster [stop+fricative] in the onset, that is well-formed, as
the sonority rises. Abiding by the Sonority Sequencing Principle can also lead to syllabification of the
nasal and the stop in coda position ([nd.j] or [mb.j]). In this case, the sonority falls in the coda and the
cluster is well-formed. However, this option is more marked, as it leads to a complex coda. Fusion also
results in a well-formed cluster ([bj]) in the onset of σ2 and to a coda-less, hence unmarked, first
syllable. Both [bj] and [dj] are legal onsets in SA (cf. (7)).
There are no [ndj] or [mbj] clusters in SMG, neither word-initially nor word-medially, but there are
[ndj] and [mbj] word-initial clusters in SA. Cf. (24) for the clusters accepted in word-initial, wordmedial, and word-final position in SA and SMG. According to Papafilis (2003), there are 13 entries that
start with [ndj] and 6 entries that start with [mbj] in SA. Heritage speakers in this study accepted [ndj]
clusters nearly twice as much as [mbj] clusters (30.55% and 16.67% respectively). The cluster [dj] is
legal both in word-initial and in word-medial position in SA, but is not accepted in SMG. The cluster
[bj] is accepted in both SMG and SA onsets, both word-initially and word-medially. In SA, there are 148
entries starting with [nd], 53 entries starting with [dj], 6 entries starting with [bj] and 114 entries
starting with [mb] (Papafilis, 2003). This means that, if the syllabification patterns are driven by cluster
legality in word margins, then [dj] and [bj] clusters are good candidates for onset position. If
syllabification is driven by dominant language interference, then heritage speakers should avoid [dj]
onsets, that are not accepted in SMG. The latter is not the case, as heritage speakers produced [dj]
onsets in 66.67% of their relevant data on average. At the same time, less proficient heritage speakers
(Florida, Participant 1 and Vasiliki) produced [nd] onsets in 100% of their relevant data.
47
5.1.3 [stop+fricative+fricative]
One cluster of the type [stop+fricative+fricative] was tested, namely a cluster consisting of a
voiceless stop [k] (C1), a voiceless fricative [θ] (C2), and a voiced fricative [j] (C3) (cf. note 22).
The data show that [stop+fricative+fricative] sequences are not perceived as a cluster 47 for the vast
majority of heritage speakers. [kθj] is syllabified as a cluster under σ2 (cf. 39a) in 22.22% of stimuli, on
average (33.33% by Thodoris, and 100% by Aggeliki). The baseline mostly perceives [kθj] as a cluster,
syllabifying the three consonants under σ2 in 41.67% of the data on average. This result comes only
from Alma (83.33% of her data for this cluster), as Evelina never produces this syllabification.
Most heritage speakers (72.23% on average in the experimental group’s data) syllabified the stop
in the coda of σ1, forming a complex onset in σ2. Of this percentage, only the 16.67% represents a [θj]
onset cluster (cf. 39b), while 55.56% of the experimental group’s data for [stop+fricative+fricative]
represent an onset with a [θç] cluster in σ2 (100% in Florida’s data, 50% in Thodoris’ and Alex’s data,
and 66.67% in Vasiliki’s and participant 1’s data) (cf.39c). Similarly, Evelina, who is a baseline
participant chooses [θç] clusters in onset position in 50% of her relevant data, but also [kθç] cluster
onsets (cf. 39d) in 16.67% of her relevant data. Alma, a control group participant, produces a [k] coda
in σ1 and a [θj] cluster in the onset of σ2 only once (16.67% of her syllabifications for this cluster), but
never produces a [θç] cluster. These data show that participants, especially heritage speakers, do
assimilate voicing in their cluster productions, which is obligatory in both their languages (SA and
SMG), but this only happens in about half the data and this is not always the case for most participants.
No effect of stress position or environment was found in this dataset.
(39) a. [pi.kθje] (Aggeliki)
b. [pek.θjik] (Participant 1)
c. [tuk.θçe] (Florida)
d. [tu.kθçep] (Evelina)
The percentages of each syllabification pattern for each participant and the average for each
pattern are shown in (40) for heritage speakers and in (41) for the baseline. The average for the totality
of participants is shown in (42). All percentages are rounded down to the second decimal place.
(40) Heritage speakers’ syllabification patterns for [kθj] (%)
Syllabification
Florida
Partic. 1
Vasiliki
Thodoris
Aggeliki
Alex
Average
heritage
[.kθj]
[.kθç]
[k.θç]
[k.θj]
0
0
100
0
0
16.67
66.67
16.67
0
16.67
66.67
16.67
33.33
0
50
16.67
100
0
0
0
0
0
50
50
22.22
5.56
55.56
16.67
47
Here, when the three consonants are syllabified as cluster, I assume that the first two consonants of the cluster
(C1 and C2) are attached to the syllable node (appendices) under σ2, while the remaining consonant (C1) is
attached to the onset (cf. 3.1.1).
48
(41) Baseline’s syllabification patterns for [kθj] (%)
Syllabification
Evelina
Alma
Average
Baseline
[.kθj]
0
83.33
41.67
[.kθç]
16.67
0
8.33
[k.θç]
50
0
25
[k.θj]
33.33
16.67
25
(42) Total syllabification patterns for [kθj] (%)
Syllabification
Average
[.kθj]
27.08
[.kθç]
6.25
[k.θç]
47.92
[k.θj]
18.75
The three-consonant cluster [kθj] is not well-formed. There is rising sonority from the stop to the
fricative but, according to the sonority scale proposed for SMG (cf. 3.3), [j] is more sonorous than [θ]
for SMG, but there is a sonority plateau in SA (cf. 3.2). When the stop is syllabified as coda, the second
syllable starts with a two-consonant cluster that is well-formed in SMG and accepted in SA (cf. (7)).
However, in the data where voice assimilation is present, there is plateau sonority between C2 and C3
in both languages, so the second syllable starts with a two-consonant cluster that is ill-formed in both
SMG and SA., but this cluster ([θç]) is accepted in SMG. Cf. (24) for the clusters accepted in word-initial,
word-medial, and word-final position in SA and SMG.
The cluster [kθj], is allowed in both word-initial and in word-medial position in SA. The cluster [θj]
is legal in SA onsets word-initially, but not word-medially, and it is allowed, only after voicing
assimilation ([θç]) in SMG onsets. According to Papafilis (2003), there are 11 entries starting with [kθj]
in SA, though they all are one word’s derivatives. There are 5 entries starting with [kθ] (which is also
legal in SMG, word-medially), though they all are one word’s derivatives, and 20 entries starting with
[θj], though they all are one word’s derivatives. There are no words starting with or containing a [θç]
cluster, as there is no [x] or [ç] phoneme in SA. However, /j/ is usually assimilated and produced as [ç],
as voice assimilation in the domain of word is a common phonological process in both SA and SMG (cf.
Nikolou, 2013). As the clusters mentioned above seem to be rare in SA, at least word-initially, word
frequency effects cannot be ruled out.
5.1.4 [fricative+fricative+fricative]
One cluster of the type [fricative+fricative+fricative] was tested, namely a cluster consisting of
three voiced fricatives: a sibilant [z] (C1), a labial [v] (C2), and [j] (C3) (cf. note 22).
49
The data show that [fricative+fricative+fricative] sequences are not perceived as a cluster 48 for the
vast majority of heritage speakers and for all control participants. [zvj] is syllabified as a cluster under
σ2 (cf. 43a) in 30.56% of stimuli, on average, by heritage speakers (33.33% by Thodoris, 100% by
Aggeliki, 16.67 by Florida, participant 1 and Vasiliki and not at all by Alex). The average percentage is
even lower in the control group (16.67%), as both participants produced this syllabification in 16.67%
of their data.
The vast majority of the data coming from both groups show a preference towards syllabifying the
sibilant as a simple coda in σ1 and the cluster [vj] as a complex onset in σ2 (cf.43b). The average
percentage is 69.44% for the experimental group, where everyone but Aggeliki demonstrated a clear
preference to this pattern. Meanwhile, the average percentage in the control group is 83.33% where,
again, everyone preferred this syllabification pattern.
No effect of stress position or environment was found in this dataset.
(43) a. [ce.zvje] (Aggeliki)
b. [koz.vjo] (Florida)
The percentages of each syllabification pattern for each participant and the average for each
pattern are shown in (44) for heritage speakers and in (45) for the baseline. The average for the totality
of participants is shown in (46). All percentages are rounded down to the second decimal place.
(44) Heritage speakers’ syllabification patterns for [zvj] (%)
Syllabification
Florida
Partic. 1
Vasiliki
Thodoris
Aggeliki
Alex
Average
heritage
[.zvj]
16.67
16.67
0
30.56
83.33
83.33
33.33
66.67
100
[z.vj]
16.67
83.33
0
100
69.44
(45) Baseline’s syllabification patterns for [zvj] (%)
Syllabification
Evelina
Alma
Average
Baseline
[.zvj]
16.67
16.67
16.67
[z.vj]
83.33
83.33
83.33
(46) Total syllabification patterns for [zvj] (%)
Syllabification
Average
[.zvj]
27.08
[z.vj]
72.92
48
Here, when the three consonants are syllabified as cluster, I assume that the first two consonants of the cluster
(C1 and C2) are attached to the syllable node (appendices) under σ2, while the remaining consonant (C1) is
attached to the onset (cf. 3.1.1).
50
While the Maximal Onset Principle calls for syllabifying the three consonants in onset position, the
three-consonant cluster [zvj] is not well-formed. There is sonority plateau between C1 and C2 and
between C2 and C3 in SA. Following the Sonority Scale proposed for SMG (Kappa, 1995) (cf. 3.3, 3.4),
/z/ is more sonorous than the following fricative, which means that there is reversed sonority clusterinitially, followed by sonority plateau. When the first consonant of the cluster is syllabified as coda,
the second syllable starts with a two-consonant cluster that is accepted in SMG and SA onsets (cf. (7)).
Cf. (24) for the clusters accepted in word-initial, word-medial, and word-final position in SA and SMG.
The cluster [zvj] is not allowed in SMG but it is allowed in both word-initial and word-medial onsets
in SA. However, according to Papafilis (2003), there is only 1 entry starting with [zvj] in SA (zvjerdh
/zvjerð/ ‘to wean’ or, metaphorically, ‘to alienate’). The cluster [vj], is allowed in both word-initial and
in word-medial onset position in SA and SMG. There are 73 entries starting with [vj] in a SA dictionary
(Papafilis, 2003). The cluster [zv] is also accepted in both SMG and SA, and there are 29 entries in a SA
dictionary (Papafilis, 2003). This implies that word frequency effects could result in syllabification of
[zvj] in two different syllables, as heritage speakers might have never heard the word /zvjerð/.
5.1.5 [stop+fricative+stop]
Five clusters of the type [stop+fricative+stop] were tested. They all consisted of a labial voiced stop
[b] or a velar voiceless stop [k] (C1), a voiceless sibilant [s] (C2) and a voiceless stop [t], [k], or [p] (C3).
The data show that [stop+fricative+stop] sequences are not perceived as a cluster 49 for the vast
majority of heritage speakers and for all control participants. [bst] is syllabified as a cluster under σ2
(cf. 47a) in 19.44% of stimuli, on average, by heritage speakers (only 16.67% by Alex, but 100% by
Aggeliki). The average percentage is even lower in the control group (16.67%), a result which comes
only from Alma (in 33.33% of her relevant data). The cluster [bsk] is syllabified as a cluster under σ2
(cf. 47b) in 22.22% of the experimental group’s data (66.67% by Aggeliki, and 16.67% by Florida,
participant 1, Thodoris and Alex), and in 8.33% of the control group’s data. Again, the result comes
only from Alma (16.67%). The cluster [kst] was syllabified under σ2 (cf. 47c) in 27.78% of the
experimental group’s data and in 16.67% of the control group’s data (33.33% by Alma and zero by
Evelina). The cluster [ksk] is syllabified as a cluster under σ2 (cf. 47d) in the 8.33% of the experimental
group’s data and not at all at the control group’s data. Finally, the cluster [ksp] is syllabified under σ2
(cf. 47e) in 8.33% of the experimental group’s data (only by Aggeliki, in half her relevant data (50%))
and in 8.33% of the control group’s data.
The vast majority of the data, coming from both groups, show a preference towards syllabifying
the first stop as a simple coda in σ1 and the [s+stop] clusters [st], [sk] and [sp] under σ2 (cf.47f-j). The
sequence [st] was syllabified as a cluster under σ2 by heritage speakers, in 50% of the relevant data,
where it was part of a [kst] cluster and in 75% when it was part of a [bst] cluster. [sk] was syllabified
as an onset cluster in 63.89% of the relevant data, where it was part of a [ksk] cluster and in 72.22%
when it was part of a [bsk] cluster. Finally, [sp] was syllabified as cluster under σ2 in 66.67% of the
relevant data, where it was part of a [ksp] cluster. In the control group, [st] was syllabified as cluster
under σ2, in 58.33% of the relevant data, where it was part of a [kst] cluster and in 50% when it was
part of a [bst] cluster. [sk] was syllabified as an onset cluster in 58.33% of the relevant data, both when
49
Here, when the three consonants are syllabified as cluster, I assume that the first and the second consonant
of the cluster (C1 and C2) are attached to the syllable node (appendices) under σ2, while C3 is attached to the
onset (cf. 3.1.1).
51
it was part of a [ksk] cluster and when it was part of a [bsk] cluster. Finally, [sp] was syllabified as an
onset cluster in 41.67% of the relevant data, where it was part of a [ksp] cluster.
Intriguingly, there were several data where there was a complex coda in the first syllable and a
simple onset in the second syllable (cf. 47k-o). In the experimental group, [bs] was a coda in σ1 in 5.56%
of the relevant data (both as part of [bst] and as part of [bsk]). Even more interestingly, the results do
not come from the same participants in both instances. Florida and Alex syllabify the cluster [bst] as
[bs.t] in 16.67% of their relevant data (that is, only once), while Aggeliki syllabifies the cluster [bsk] as
[bs.k] in 33.33% of her relevant data. The cluster [ks] was a coda in 22.22% of the relevant data, as
part of the cluster [kst], in 27.78% of the relevant data when it was part of [ksk] and in 25% of the
relevant data, as part of [ksp]. In the control group, Alma never produces complex codas, while Evelina
produces [bs] in coda position in 66.67% of her relevant data (both when [bs] is part of [bst] and when
it is part of [bsk]. Evelina also syllabifies [ks] in coda position: in 50% of her relevant data when it is
part of [kst], in 83.33% when it is part of [ksk] and in 100% when it is part of [ksp].
No effect of stress position or environment was found in this dataset.
(47) a. [pe.bstik] (Alma)
b. [pa.bska] (Aggeliki)
c. [te.ksti] (Aggeliki)
d. [pi.ksce] (Aggeliki)
e. [to.kspo] (Aggeliki)
f. [kob.sto] (Florida)
g. [kab.ska] (Vasiliki)
h. [tuk.step] (Thodoris)
i. [tuk.sce] (Evelina)
j. [cek.spe] (Alex)
k. [pebs.tik] (Evelina)
l. [cebs.ce] (Aggeliki)
m. [peks.tik] (Evelina)
n. [ceks.ce] (Alex)
o. [toks.po] (Alex)
The percentages of each syllabification pattern for each participant and the average for each
pattern are shown in (48) for heritage speakers and in (49) for the baseline. The average for the totality
of participants is shown in (50). All percentages are rounded down to the second decimal place.
52
(48) Heritage speakers’ syllabification patterns for [stop+fricative+stop] clusters (%)
Syllabification
Florida
Partic. 1
Vasiliki
Thodoris
Aggeliki
Alex
Average
heritage
[.bst]
[b.st]
[bs.t]
[.bsk]
[b.sk]
[bs.k]
[.kst]
[k.st]
[ks.t]
[.ksk]
[k.sk]
[ks.k]
[.ksp]
[k.sp]
[ks.p]
0
83.33
16.67
16.67
83.33
0
33.33
66.67
0
16.67
83.33
0
0
100
0
0
100
0
16.67
83.33
0
33.33
50
16.67
0
100
0
0
83.33
16.67
0
100
0
0
100
0
0
83.33
16.67
0
83.33
16.67
0
100
0
0
100
0
16.67
83.33
0
16.67
83.33
0
0
100
0
0
83.33
16.67
100
0
0
66.67
0
33.33
83.33
0
16.67
33.33
0
66.67
50
0
50
16.67
66.67
16.67
16.67
83.33
0
0
16.67
83.33
0
16.67
83.33
0
33.33
66.67
19.44
75
5.56
22.22
72.22
5.56
27.78
50
22.22
8.33
63.89
27.78
8.33
66.67
25
(49) Baseline’s syllabification patterns for [stop+fricative+stop] clusters (%)
Syllabification
Evelina
Alma
Average
Baseline
[.bst]
0
33.33
16.67
[b.st]
33.33
66.67
50
[bs.t]
66.67
0
33.33
[.bsk]
0
16.67
8.33
[b.sk]
33.33
83.33
58.33
[bs.k]
66.67
0
33.33
[.kst]
0
33.33
16.67
[k.st]
50
66.67
58.33
[ks.t]
50
0
25
[.ksk]
0
0
0
[k.sk]
16.67
100
58.33
[ks.k]
83.33
0
41.67
[.ksp]
0
16.67
8.33
53
[k.sp]
0
83.33
41.67
[ks.p]
100
0
50
(50) Total syllabification patterns for [stop+fricative+stop] clusters (%)
Syllabification
Average
[.bst]
18.75
[b.st]
68.75
[bs.t]
12.5
[.bsk]
18.75
[b.sk]
66.67
[bs.k]
12.5
[.kst]
25
[k.st]
52.08
[ks.t]
22.97
[.ksk]
6.25
[k.sk]
62.5
[ks.k]
31.25
[.ksp]
8.33
[k.sp]
60.42
[ks.p]
31.25
The Maximal Onset Principle calls for syllabifying the three consonants in onset position, but the
three-consonant clusters of the type [stop+fricative+stop] are not well-formed. There is rising sonority
from the first stop to the sibilant, but reversed sonority between the sibilant and the stop that follows.
As the sibilant is in the middle of the three-consonant cluster, it cannot form a well-formed cluster
even if one of the marginal consonants is not part of the cluster. This means that, if the sibilant forms
a complex coda with the first stop of the [stop+fricative+stop] cluster (C1), this coda will have rising
sonority and will, thus, be ill-formed. Additionally, if the sibilant forms a complex onset with the second
stop of the [stop+fricative+stop] cluster (C3), this onset will have reversed sonority and thus, be illformed. However, in SMG reversed sonority is accepted in onsets, when the first consonant is a
fricative (cf. 3.3)
The clusters [bst], [bsk], [kst], [ksk], and [ksp] are accepted only in word-medial position in SA. The
cluster [bs] is accepted in SA onsets only word-medially. The clusters [ks] and [sp] are accepted clusters
in both word-initial and word-medial position in both SA and SMG, while [st] and [sk] are accepted
54
only in word-initial position in SA and accepted in any onset in SMG. Cf. (24) for the clusters accepted
in word-initial, word-medial, and word-final position in SA and SMG. According to Papafilis (2003),
there are only 2 entries starting with [ks] in SA (both are Greek loanwords). There are 26 entries
starting with [sk], 77 entries starting with [sp], and 138 entries starting with [st]. [ks] and [st] are also
accepted in word-final coda position in SA. So, frequency effects can be an explanation for the
syllabifications in this dataset.
Concluding, there was no effect of cluster legality or cluster acceptability attested in the dataset
studied in section 5.1. The only effect that may be present is that of word-frequency.
5.2 Participants and their data
This section describes the syllabification patterns that each participant used in the experiment. The
number of times each heritage speaker used each of the patterns that are described in 5.1 is shown
in detail in (51) and the relevant percentages in (52), while the same information for the control group
is presented in (53)-(54). The same information is presented in graphical form in figures 1-3 for
heritage speakers, and in figures 4-6 for the baseline. In figures 7-9 there is a contrastive presentation
of the average for each group.
Nearly all heritage speakers in this study perceived the clusters under investigation as
heterosyllabic consonant sequences, in the vast majority of their syllabifications. In a total of 66 stimuli
for each participant, 51 (77.27%) of them, on average, were perceived as heterosyllabic consonant
sequences by heritage speakers. Five participants in the experimental group (Florida, Participant 1,
Vasiliki, Thodoris and Alex) perceived most of the clusters as heterosyllabic consonant sequences. Out
of 66 stimuli in total, Florida followed a syllabification of a heterosyllabic consonant sequence, with a
complex (two-consonant) onset in σ2 in 56 (i.e. in 93.33% of the syllabifications of the three consonants
as a heterosyllabic sequence in her data), Participant 1 in 55 (i.e. 90.16%), Vasiliki in 59 (i.e. 95.16%),
Thodoris in 50 (i.e. 96.15%) and Alex in 46 (i.e. 75.41%). On the contrary, Aggeliki, who completed
elementary school in Albania (which means that she studied SA in a formal setting for 6 years) never
produced this pattern and most of her syllabifications treated the three-consonant cluster as a cluster
under σ2 (56 out of 66 syllabifications, or 84.85% of her total data).
No heritage speaker syllabified the clusters in the experiment as a cluster under σ1. Complex codas,
with the two first members of the cluster (C1 and C2) in coda position and the third member (C3) in the
onset of σ2, are rare in the syllabifications of Florida (only 1, or 1.67% of her syllabifications of the
three consonants as a heterosyllabic sequence), Participant 1 (2, or 3.28% of her syllabifications of the
three consonants as a heterosyllabic sequence), Vasiliki (3, or 4.84%) and Thodoris (2, or 3.85%).
Aggeliki and Alex produced complex codas somewhat more often. Aggeliki produced complex codas
10 times (i.e. in 100% of her syllabifications of the three consonants as a heterosyllabic sequence) and
Alex 15 times (i.e. in 24.59% of the syllabifications of the three consonants as a heterosyllabic sequence
in his data). All complex codas in the experimental group data are of the type [stop+fricative], namely
[bs] and [ks].
Fusion was not present in the data of every cluster type. In fact, two-consonant clusters which are
a result of fusion were observed only in [labial nasal+homorganic stop] sequences, in the clusters [mbl]
and [mpl]. The stimuli containing these two clusters were 12 for each participant (6 containing [mbl]
and 6 containing [mpl]. Fusion was mainly attested in Florida’s (3 times, or in 4.54% of the clusters
where fusion was attested) and participant 1’s data (4 times, or in 6.06% of the clusters where fusion
was attested). There was no fusion in Vasiliki’s, Aggeliki’s, Thodoris’, or Alex’s data.
55
Voice assimilation was observed between [nasal+homorganic voiceless stop] in [mpl] clusters
(voicing of the voiceless stop) and between [voiceless fricative+voiced fricative] in [kθj] clusters. The
stimuli containing these two clusters were 6 for each cluster, for each participant (12 in total) and,
intriguingly, although voice assimilation is obligatory in both SMG and SA, voice assimilation was not
frequent in the data coming from the experimental group. Voice assimilation was present 6 times in
Florida’s and in Vasiliki’s data (i.e. in 50% of the stimuli containing the clusters in question), 5 times in
participant 1’s data (or in 41.67%) 3 times in Thodoris’ and Alex’s data (or 25%) and not at all in
Aggeliki’s data. In Florida’s, participant 1’s, Thodoris’ and Alex’s data, assimilation was present only in
data containing the [kθj] cluster. In Vasiliki’s data, assimilation was present 5 times in data containing
the [kθj] cluster and one time in data containing the [mpl] cluster. This means that there was only one
instance of voicing assimilation in [nasal+voiceless stop] clusters in the data of the experimental group.
The scarcity of assimilation can be due to an effect of the stimuli, i.e. because of the fact that the
stimuli were nonce-words, participants tried to repeat the exact segments they heard and read,
ignoring any phonological rules and processes and focusing on the task.
(51) Number of times each heritage speaker (experimental group) used each strategy to syllabify the clusters in the
experiment and number of occurrences of fusion and/or assimilation
Pattern
Florida
Partic. 1
Vasiliki
Thodoris
Aggeliki
Alex
Cluster under σ2
Heterosyllabic
Heterosyllabic
complex onset
Heterosyllabic
complex coda
Fusion
Assimilation
6
60
5
61
4
62
14
52
56
10
5
61
Average
heritage
15
51
56
55
59
50
0
46
44.33
1
2
3
2
10
15
5.5
3
6
4
5
0
6
0
3
0
0
0
3
1.17
3.83
(52) Percentages for syllabification patterns and of the occurrences of fusion and/or assimilation for each heritage
speaker (experimental group)
Pattern
Florida
Partic. 1
Vasiliki
Thodoris
Aggeliki
Alex
Average
heritage
Cluster under σ2
9.09
7.57
6.06
21.21
84.85
7.58
22.73
Heterosyllabic
90.90
92.42
93.93
78.79
15.15
92.42
77.27
90.16
95.16
96.15
3.28
4.84
3.85
4.54
6.06
1.51
0
0
0
50
41.67
50
25
0
25
Heterosyllabic
complex onset
Heterosyllabic
complex coda
Fusion
Assimilation
93.33
1.67
0
100
75.41
24.59
75.03
23.04
2.02
31.94
56
Figure 1. Patterns % of syllabification of the experimental stimuli by heritage speakers (experimental group)
Figure 2. Patterns (% of the heterosyllabic data) of heterosyllabic syllabification of the experimental stimuli by
heritage speakers (experimental group)
Figure 3. Occurrence of fusion and assimilation in the data of heritage speakers (experimental group),
percentage of the relevant data, where fusion and/or assimilation could be applied
57
There was great diversity in the control group. Alma, who reports little or no attrition syllabified
the clusters under σ2 in about half the stimuli (32 out of 66, or 48.48% of the total). In 51.51% of her
data, Alma syllabified the clusters as heterosyllabic consonant sequences, always with a simple coda
in σ1, followed by a complex onset in σ2. The difference with Evelina is striking. Evelina, who reports
heavy attrition in the domain of phonetics-phonology, syllabified clusters under σ2 only 10 times
(15.15%), syllabifying the first consonant in coda position in σ1 and a two-consonant cluster in σ2 onset
23 times (in 41.07% of the heterosyllabic syllabifications in her data). Complex codas are present in
half of Evelina’s data (33 times, i.e. in 58.93% of her syllabifications of the clusters as heterosyllabic
consonant sequences), but not at all present in Alma’s data. Fusion was not present in the control
group’s dataset.
Voice assimilation was observed between [nasal+homorganic voiceless stop] in [mpl] clusters
(voicing of the voiceless stop) and between [voiceless fricative+voiced fricative] in [kθj] clusters. Voice
assimilation was present only in the data coming from Evelina (7 times, or in 58.33% of the stimuli
containing the clusters in question). Assimilation was present in Evelina’s data 4 times in [kθj] stimuli
and 3 times in [mpl] stimuli. There was no assimilation in Alma’s data. Again, this is probably due to
an effect of the experiment and does not reflect the subjects’ phonological grammar.
(53) Number of times each baseline participant (control group) used each strategy to syllabify the clusters in the
experiment and number of occurrences of fusion and/or assimilation
Average
Baseline
Pattern
Evelina
Alma
Cluster under σ2
10
32
21
Heterosyllabic
56
34
45
23
34
28.5
33
0
Fusion
0
0
0
Assimilation
7
0
3.5
Heterosyllabic
complex onset
Heterosyllabic
complex coda
16.5
(54) Percentages for syllabification patterns for each baseline participant (control group) and of the occurrences
of fusion and/or assimilation
Average
Baseline
Pattern
Evelina
Alma
Cluster under σ2
15.15
48.48
31.81
Heterosyllabic
84.85
51.51
68.18
100
70.54
Heterosyllabic
complex onset
Heterosyllabic
complex coda
Fusion
Assimilation
41.07
58.93
0
58.33
0
29.47
0
0
0
29.17
58
Figure 4. Patterns % of syllabification of the experimental stimuli by the baseline (control group)
Figure 5. Patterns (% of the heterosyllabic data) of heterosyllabic syllabification of the experimental stimuli by
the baseline (control group)
Figure 6. Occurrence of fusion and assimilation in the data of the baseline (control group), percentage of the
relevant data, where fusion and/or assimilation could be applied
59
Figure 7. Average syllabification patterns of the baseline (control group) and the heritage speakers (experimental
group) %
Figure 8. Average patterns of heterosyllabic syllabification in the baseline (control group) (experimental group)
and in heritage speakers %
100
80
60
40
20
0
Average Baseline
Heterosyllabic complex onset
Average heritage
Heterosyllabic complex coda
Figure 9. Average occurrence of fusion and assimilation in the data of the heritage speakers (experimental group)
and the baseline (control group), percentage of the relevant data, where fusion and/or assimilation could be
applied
60
5.3 Discussion of the data in the framework of Optimality Theory (OT)
5.3.1 Some basics of Optimality Theory 50
Optimality Theory (henceforth: OT) (Prince and Smolensky, 1993; McCarthy and Prince, 1993a, b)
is a framework, development of Generative Grammar, that allows for cross-linguistic variation. It is
based on the idea that there are constraint conflicts, whose resolution is reflected on the surface form,
or representation. The surface representation (output) is the best possible -or optimal- surface form,
as it is the representation that induces the least serious violations on a list of ranked and violable
constraints. The constraints are universal, but their ranking is language-specific. The ranking means
that the higher-ranked constraint has to be satisfied before all others. Every outcome will definitely
violate some constraints in any grammar, but it is less possible for forms that violate higher-ranked
constraints to surface. Markedness (or well-formedness) constraints favor universally unmarked
structures, while faithfulness constraints preserve the structure and the lexical contrasts, meaning that
they call for congruence with the lexical input. So, markedness constraints apply pressure towards
unmarked structures, and are in clash (and counterbalance) with faithfulness constraints that preserve
the language contrasts which would not exist if all structures were unmarked. All candidate outputs
are simultaneously evaluated and the optimal output is chosen, according to a grammar’s constraint
ranking.
Tableaus, like the one in (55a) are used to graphically represent the constraint ranking and the
evaluation of the candidate structures. On the top of the first column, there is the input, or underlying
form in //. The rest of the column is filled with the candidate output forms in [ ]. The constraints are
on the top of each column, the higher-ranked first. The symbol (*) here, does not mark
ungrammaticality, but it denotes a constraint violation. One (*) is used for every violation. This means
that, if a candidate form violates a constraint more than once, there will be more than one (*) symbols.
The symbol (!) marks a fatal violation, that excludes the candidate, while the symbol (☞) marks the
optimal candidate, with the least serious violations, that will surface as output.
When the ranking between two or more constraints is irrelevant for the selection of the optimal
output form, the lines in the tableau are dashed. In (55), the constraint ranking is the following:
(55) constraint 1 >> constraint 2, constraint 3>> constraint 4
(55a) Example of an OT tableau
/input/
constraint 1
constraint 2
[candidate 1]
☞ [candidate 2]
[candidate 3]
constraint 3
constraint 4
*!
*
*!
The constraints that are relevant for the evaluation of the output candidates in the present study are
described below.
Well-formedness constraints or Markedness constraints:
*COMPLEXONS: Onsets are simple, clusters are not allowed in onsets 51
50
This section is based on Kager (1999) and McCarthy (2004)
This constraint is violated when there is more than one consonant at the left syllable margin, even if they do
not form a well-formed cluster under onset.
51
61
*COMPLEXCOD: Codas are simple, clusters are not allowed in codas (appendices are not allowed at the
right syllable margin)
ONSET: Syllables must start with a consonant in onset position
NO-CODA: Syllables must be open, with no consonant in coda position. One violation for every closed
syllable
*NC̥: Sequences of nasal plus voiceless obstruent are not allowed 52
AGREE[voice]OBST: Obstruents in clusters should agree in voicing
*APPENDIX-LEFT: A consonant at the left syllable margin must be immediately dominated by onset 53,
i.e. appendices are not allowed at the left edge of a syllable 54.
Faithfulness Constraints
CONTIG-IO: (CONTIGUITY) No epenthesis or deletion
UNIFORMITY-IO: No element in the output can have multiple correspondents in the input 55
IDENT-IO(ObsVce): Correspondent obstruents are identical in their specification for voice (no changes
in the voicing of obstruents) 56
IDENT-IO(voice): The specification for the feature [voice] of a segment in the input has to be preserved
in its correspondent in the output.
The resolution of conflict between Coda and Onset constraints results in whether compliance with or
violation of the Maximal Onset Principle (cf. 3.1.2).
5.3.2 Data discussion
There is great variability in the syllabification patterns in this dataset. There are three syllabification
patterns that account for the data in this study (cf. 5.1; 5.2; APPENDIX III), namely:
a. syllabification of the cluster under the second syllable (σ2), with no coda in the first syllable (σ1) and
a three-consonant sequence under σ2. In this case, the first consonant (C1) is considered to be
attached to the syllable node of σ2, as extrasyllabic (appendix), while the second consonant (C2)
may also be an appendix, when it does not form a well-formed cluster with the third consonant
(C3). C2 is attached to the syllable node as appendix in the cases of [stop+fricative+fricative],
[fricative+fricative+fricative] and [stop+fricative+stop] clusters (cf. 3.1.1).
b. syllabification of the cluster as a heterosyllabic consonant sequence, with C1 syllabified as a simple
coda in the first syllable (σ1) and a two-consonant cluster under the second syllable (σ2). In this
case, C2 may be an appendix, when it does not form a well-formed cluster with the third consonant
(C3). C2 is attached to the syllable node as appendix in the cases of [stop+fricative+fricative],
[fricative+fricative+fricative] and [stop+fricative+stop] clusters (cf. 3.1.1).
52
This constraint is violated by [mp] sequences in [mpl] clusters in the data of this study
cf. Goad and Rose, 2004:131
54
This constraint is violated by every appendix at the left syllable margin, in the case of this study: by C1 in threeconsonant sequences and by C2 when there is reversed sonority or sonority plateau between C2 and C3
55
This constraint is violated by fusion, which is present in the data of this study
56
Kager (1999:70)
53
62
c. syllabification of the cluster as a heterosyllabic consonant sequence, with C1 and C2 under the first
syllable (σ1) and a simple onset in the second syllable (σ2).
There are three distinct constraint rankings ((56)-(58)) that result in these three patterns surfacing.
The syllabification of the consonant clusters of this experiment under σ2, where C1 is attached to
the syllable node as an appendix, is a result of the constraint ranking in (56), which favors open, CV
syllables, while *COMPLEXONS and *APPENDIX-LEFT, which are violated by three-consonant clusters,
have a low ranking.
In (56a), there is an example with a nonce-word from the experimental stimuli that ends with an
open syllable, where C2 and C3 form a well-formed cluster in the onset of σ2, while C1 is an appendix,
attached to the syllable node of σ2.
In (56b), there is an example with a nonce-word from the experimental stimuli that ends with an
open syllable, where C2 and C3 form an ill-formed cluster (reversed sonority), so C3 is syllabified in the
onset of σ2 and C1 and C2 are appendices, attached to the syllable node of σ2.
In (56c), there is an example of a nonce-word that ends with a closed syllable, where C2 and C3 form
a well-formed cluster in the onset of σ2, while C1 is an appendix, attached to the syllable node of σ2.
In (56d) there is an example of a nonce-word that ends with a closed syllable where C2 and C3 form
an ill-formed cluster (reversed sonority), so C3 is syllabified in the onset of σ2 and C1 and C2 are
appendices, attached to the syllable node of σ2.
(56) CONTIG-IO, ONSET, NO-CODA, *COMPLEXCOD >>*COMPLEXONS >> *APPENDIX-LEFT
(56a)
/kondjo/
☞ [ko.ndjo]
[kon.djo]
CONTIG-IO
ONSET
*COMPLEXCOD
*!
[kond.jo]
[kondj.o]
[ko.djo]
NO-CODA
*!
*COMPEXONS
*
*APP-LEFT
*
*
*!
*
*
*
*!
*
(56b)
/pikspe/
CONTIG-IO
ONSET
NO-CODA
*COMPLEXCOD
☞ [pi.kspe]
[pik.spe]
*!
[piks.pe]
*!
*
*
*
[piksp.e]
[pi.spe]
*!
*!
*COMPEXONS
*APP-LEFT
*
**
*
*
*
*
63
(56c)
/pemblik/
CONTIG-IO
ONSET
NO-CODA
*COMPLEXCOD
*COMPEXONS
*APP-LEFT
*
☞ [pe.mblik]
*
*
[pem.blik]
**!
*
[pemb.lik]
**!
*
**
*
[pembl.ik]
[pe.blik]
*!
*!
*
*
(56d)
/cikstek/
CONTIG-IO
ONSET
NO-CODA
*COMPLEXCOD
*COMPEXONS
*APP-LEFT
☞ [ci.kstek]
*
*
**
[cik.stek]
**!
*
*
[ciks.tek]
**!
*
**
*
*
*
[cikst.ek]
[ci.stek]
*!
*!
*
The syllabification that results in a simple coda in σ1 and a complex onset in σ2 is a result of the
constraint ranking in (57), which favors syllables with a simple coda and syllables with an onset, while
*COMPLEXONS which is violated by complex onsets, and NO-CODA that is violated by codas have a low
ranking.
In (57a), there is an example with a nonce-word from the experimental stimuli that ends with an
open syllable, where C2 and C3 form a well-formed cluster in the onset of σ2, while C1 is an appendix,
attached to the syllable node of σ2.
In (57b), there is an example with a nonce-word from the experimental stimuli that ends with an
open syllable, where C2 and C3 form an ill-formed cluster (reversed sonority), so C3 is syllabified in the
onset of σ2 and C1 and C2 are appendices, attached to the syllable node of σ2.
In (57c), there is an example of a nonce-word that ends with a closed syllable, where C2 and C3 form
a well-formed cluster in the onset of σ2, while C1 is an appendix, attached to the syllable node of σ2.
In (57d) there is an example of a nonce-word that ends with a closed syllable where C2 and C3 form
an ill-formed cluster (reversed sonority), so C3 is syllabified in the onset of σ2 and C1 and C2 are
appendices, attached to the syllable node of σ2.
(57) CONTIG-IO, ONSET, *COMPLEXCOD>>*APPENDIX-LEFT >> *COMPLEXONS, NO-CODA
64
(57a)
/tundje/
CONTIG-IO
ONSET
*COMPLEXCOD
[tu.ndje]
*APP-LEFT
*COMPEXONS
*!
*
☞ [tun.dje]
*
[tund.je]
[tundj.e]
[tu.dje]
*!
NO-CODA
*
*!
*
*
*
*!
*
(57b)
*APP-LEFT
*COMPEXONS
[ko.ksko]
**!
*
☞ [kok.sko]
*
*
/koksko/
CONTIG-IO
ONSET
[koks.ko]
[koksk.o]
[ko.sko]
*!
*COMPLEXCOD
NO-CODA
*
*!
*
*
*
*!
*
*
*APP-LEFT
*COMPEXONS
NO-CODA
*!
*
*
*
**
(57c)
/pendjik/
CONTIG-IO
ONSET
*COMPLEXCOD
[pe.ndjik]
☞ [pen.djik]
[pend.jik]
[pendj.ik]
[pe.djik]
*!
*!
**
*
**
*!
*
*
*APP-LEFT
*COMPEXONS
NO-CODA
[pe.kspik]
**!
*
*
☞ [pek.spik]
*
*
**
(57d)
/pekspik/
CONTIG-IO
ONSET
[peks.pik]
[peksp.ik]
[pe.spik]
*!
*!
*COMPLEXCOD
*!
**
*
**
*
*
*
65
The syllabification that results in a complex coda in σ1 and simple onset in σ2 is the result of the
constraint ranking in (58), which favors syllables with a simple onset, even if this leads to a complex
coda (which is more marked than a complex onset (cf. 3.1.1)) where NO-CODA and *COMPLEXCOD
which are violated by codas have a low ranking.
In (58a), there is an example with a nonce-word from the experimental stimuli that ends with an
open syllable, where C2 and C3 form a well-formed cluster in the onset of σ2, while C1 is an appendix,
attached to the syllable node of σ2.
In (58b), there is an example with a nonce-word from the experimental stimuli that ends with an
open syllable, where C2 and C3 form an ill-formed cluster (reversed sonority), so C3 is syllabified in the
onset of σ2 and C1 and C2 are appendices, attached to the syllable node of σ2.
In (58c), there is an example of a nonce-word that ends with a closed syllable, where C2 and C3 form
a well-formed cluster in the onset of σ2, while C1 is an appendix, attached to the syllable node of σ2.
In (58d) there is an example of a nonce-word that ends with a closed syllable where C2 and C3 form
an ill-formed cluster (reversed sonority), so C3 is syllabified in the onset of σ2 and C1 and C2 are
appendices, attached to the syllable node of σ2.
(58) CONTIG-IO, ONSET, *COMPLEXONS, *APPENDIX-LEFT >> NO-CODA, *COMPLEXCOD
(58a)
*COMPEXONS
*APP-LEFT
[ko.mblo]
*!
*
[kom.blo]
*!
/komblo/
CONTIG-IO
ONSET
[ko.blo]
*!
*!
*COMPLEXCOD
*
☞ [komb.lo]
[kombl.o]
NO-CODA
*
*
*
*
NO-CODA
*COMPLEXCOD
*
(58b)
*COMPEXONS
*APP-LEFT
[tu.kste]
*!
**
[tuk.ste]
*!
*
/tukste/
CONTIG-IO
ONSET
☞ [tuks.te]
[tukst.e]
[tu.ste]
*!
*!
*
*
*
*
*
*
*
66
(58c)
*COMPEXONS
*APP-LEFT
NO-CODA
[pe.mbjik]
*!
*
*
[pem.bjik]
*!
/pembjik/
CONTIG-IO
ONSET
**
☞ [pemb.jik]
[pembj.ik]
[pe.bjik]
*!
*!
*COMPLEXCOD
*
**
*
**
*
*
(58d)
*COMPEXONS
*APP-LEFT
NO-CODA
[tu.kspep]
*!
**
*
[tuk.spep]
*!
*
**
/tukspep/
CONTIG-IO
ONSET
☞ [tuks.pep]
[tuksp.ep]
[tu. spep]
*!
*!
*
*
*COMPLEXCOD
**
*
**
*
*
In addition to the above, two phonological processes are present in this dataset (cf. 5.1; 5.2),
namely:
a. fusion of two consonants as a strategy to avoid [nasal+voiceless obstruent] clusters [mpl]
which violate *NC̥, and as a strategy to simplify the three-consonant cluster [mbl].
b. Voicing assimilation between [m] and [p] in [mpl] clusters, again, to avoid violation of *NC̥,
and between [θ] and [j] in [kθj] clusters, as voicing assimilation in obstruents is obligatory in
both SA and SMG (cf. 5.1.3).
There are seven distinct constraint rankings that result in these two processes (cf.59-65).
The syllabification that results in fusion in [mpl] clusters in the experiment is the result of the
constraint ranking in (59), where *NC̥ is highly ranked and UNIFORMITY-IO that is violated by fusion
has a low ranking. In (59a), there is an example with a nonce-word from the experimental stimuli that
ends with an open syllable. There was no fusion attested in stimuli that contain the [mpl] cluster and
end in a closed syllable (cf. 5.1.1).
(59) *NC̥, ONSET, CONTIG-IO >> NO-CODA >> *APPENDIX-LEFT >> COMPEXONS >> UNIFORMITY-IO
67
(59a)
/tumple/
*NC̥
[tu.mple]
*!
[tum.ple]
*!
[tumpl.e]
*!
ONSET
CONTIG-IO
NO-CODA
*APP-LEFT
*COMPEXONS
*
*
*
*
*
*
[tu.mble]
*!
[tum.ble]
*!
[tumb.le]
*!
[tumbl.e]
UNIFORMITY-IO
*!
*
*
*
☞ [tu.ble]
*
*
[tu.ple]
*!
*
The syllabification that results in fusion in [mbl] clusters of the type [nasal+voiced stop+liquid] is
the result of the constraint ranking in (60), where CONTIG-IO is highly ranked and *COMPEXONS is
ranked higher than UNIFORMITY-IO, which is violated by fusion. In (60a), there is an example with a
nonce-word from the experimental stimuli that ends with an open syllable. In (60b), there is an
example with a nonce-word that ends with a closed syllable.
(60) ONSET, CONTIG-IO >>*COMPLEXCOD, *APPENDIX-LEFT, NO-CODA >> *COMPEXONS, UNIFORMITY-IO
(60a)
/komblo/
ONSET
CONTIG-IO
*COMPLEXCOD
[ko.mblo]
*!
*
*
*!
*
*!
UNIFORMITY-IO
*
*
☞ [ko.blo]
[kob.lo]
*COMPEXONS
*
*!
[komb.lo]
[ko.mlo]
NO-CODA
*!
[kom.blo]
[kombl.o]
*APP-LEFT
*
*
*!
68
(60b)
/tumblep/
ONSET
CONTIG-IO
*COMPLEXCOD
*APP-LEFT
NO-CODA
*COMPEXONS
*!
*
*
**!
*
[tu.mblep]
[tum.blep]
[tumbl.ep]
*!
[tumb.lep]
[tu.mlep]
*
**
*!
**
*!
*
*
☞ [tu.blep]
*
*
[tub.lep]
**!
UNIFORMITY-IO
*
Fusion in [mbl] and [mpl] clusters of the type [nasal+stop+liquid] cannot happen when faithfulness
constraints are ranked higher than well-formedness constraints, as in (61). Constraints that are
violated by syllable structure are not relevant here and can be at any ranking after UNIFORMITY-IO
and CONTIGUITY-IO, depending on the syllabification of the output (cf. (56)-(58) for the different
rankings of *COMPLEXCOD, NO-CODA, *COMPEXONS, *APPENDIX-LEFT, and ONSET for each
syllabification pattern). As syllabification is not relevant here, the output candidates in the tableaus
are not syllabified.
(61) UNIFORMITY-IO, CONTIG-IO
(61a)
/komblo/
UNIFORMITY-IO
CONTIG-IO
☞ [ko.mblo]
[komlo]
[koblo]
*!
!*
Voicing in [mpl] clusters is the result of the constraint ranking in (62), where *NC̥, UNIFORMITY-IO
(which is violated by fusion), and CONTIGUITY-IO (which is violated by epenthesis or deletion) are
highly ranked. In (62a), there is an example with a nonce-word from the experimental stimuli for the
cluster [mpl] that ends with an open syllable. Constraints that are violated by syllable structure are not
relevant here and can be at any ranking after IDENT-IO(voice), depending on the output (cf. (56)-(58)
for the different rankings of *COMPLEXCOD, NO-CODA, *COMPEXONS, *APPENDIX-LEFT, and ONSET for
each syllabification pattern). As syllabification is not relevant here, the output candidates in the
tableaus are not syllabified.
(62) *NC̥ >> UNIFORMITY-IO, CONTIG-IO >> IDENT-IO(voice)
69
(62a)
/pamplo/
*NC̥
[pamplo]
*!
UNIFORMITY-IO
CONTIG-IO
[paplo]
*!
[pamlo]
*!
[pablo]
IDENT-IO(voice)
*!
*
☞ [pamblo]
*
When assimilation is not applied in [nasal+voiceless obstruent] clusters, it is a result of the
constraint ranking shown in (63), where *NC̥ is ranked lower than faithfulness constraints. Constraints
that are violated by syllable structure are not relevant here and can be at any ranking after *NC̥,
depending on the output (cf. (56)-(58) for the different rankings of *COMPLEXCOD, NO-CODA,
*COMPEXONS, *APPENDIX-LEFT, and ONSET for each syllabification pattern). As syllabification is not
relevant here, the output candidates in the tableaus are not syllabified.
(63) CONTIG-IO, UNIFORMITY-IO, IDENT-IO(voice) >> *NC̥
(63a)
/pamplo/
CONTIG-IO
UNIFORMITY-IO
IDENT-IO(voice)
☞ [pamplo]
*
[paplo]
*!
[pamlo]
*!
[pablo]
*NC̥
*!
[pamblo]
*
*!
Voicing in [kθj] clusters is the result of the constraint ranking in (64), where AGREE[voice]OBST is
ranked higher than faithfulness constraints. Constraints that are violated by syllable structure are not
relevant here and can be at any ranking after IDENT-IO(ObsVce), depending on the syllabification of
output (cf. (56)-(58) for the different rankings of *COMPLEXCOD, NO-CODA, *COMPEXONS, *APPENDIXLEFT, and ONSET for each syllabification pattern). As syllabification is not relevant here, the output
candidates in the tableaus are not syllabified.
(64) AGREE[voice]OBST >> CONTIG-IO >> IDENT-IO(ObsVce)
70
(64a)
/tukθjep/
AGREE[voice]OBST
[tukθjep]
*!
CONTIG-IO
☞ [tukθçep]
IDENT-IO(ObsVce)
*
[tujep]
**!
[tukθep]
*!
[tugðjep]
**!
When voicing assimilation is not applied in [voiceless fricative+voiced fricative] clusters, it is a result
of the constraint ranking shown in (65), where AGREE[voice]OBST is ranked lower than faithfulness
constraints. Constraints that are violated by syllable structure are not relevant here and can be at any
ranking after AGREE[voice]OBST, depending on the syllabification of output. (cf. (56)-(58) for the
different rankings of *COMPLEXCOD, NO-CODA, *COMPEXONS, *APPENDIX-LEFT, and ONSET for each
syllabification pattern). As syllabification is not relevant here, the output candidates in the tableaus
are not syllabified.
(65) IDENT-IO(voice), CONTIG-IO >> AGREE[voice]OBST >> *COMPLEXCOD, NO-CODA, *COMPEXONS, ONSET
(65a)
/tukθjep/
IDENT-IO(voice)
CONTIG-IO
☞ [tukθjep]
*
[tukθçep]
*!
[tujep]
**!
[tukθep]
*!
[tugðjep]
AGREE[voice]OBST
**!
5.4 Interpretation of the results
According to the data presented in detail in this chapter, the vast majority of the heritage speakers
who participated in this study (with the sole exception of Aggeliki) usually complies with grammar (57)
CONTIG-IO, ONSET, *COMPLEXCOD>>*APPENDIX-LEFT >> *COMPLEXONS, NO-CODA, syllabifying threeconsonant clusters as heterosyllabic consonant sequences, with a simple coda in the first syllable and
a complex onset in the second syllable (e.g. [tun.dje], Vasiliki). However, at the same time, they also
comply with two additional, peripheral grammars, at a low statistical frequency: the one in (56)
CONTIG-IO, ONSET, NO-CODA, *COMPLEXCOD >>*COMPLEXONS >> *APPENDIX-LEFT, which dictates
open syllables (e.g. [ko.zvjo], Thodoris) and the one in (58) CONTIG-IO, ONSET, *COMPLEXONS,
*APPENDIX-LEFT >> NO-CODA, *COMPLEXCOD, which results in complex codas (e.g. [piks.ce], Alex).
Additionally, they are not settled as to whether use phonological processes (fusion and assimilation),
71
using the grammars in (59)-(65) indecisively and interchangeably 57 (e.g. [tu.ble], participant 1;
[tuk.θçep], Vasiliki, but [kom.plo], participant 1; [pe.kθjik], Vasiliki). Aggeliki, who moved to Greece at
the age of 11, syllabifies the clusters in the experimental stimuli as a cluster under σ2 (e.g. [pa.ndja]) in
the vast majority of her data, thus usually complying with the grammar in (56), as she produces an
open σ1. Nevertheless, she also uses another peripheral grammar (i.e. (58)) at a considerably lower
frequency (e.g. [koks.po]). The fact that there is no fusion and assimilation in her data indicates that
she applies the grammars described in (61), (63) and (65) where faithfulness constraints are highly
ranked (e.g. [pi.mple], [pi.kθje])(but, cf. note 57).
This variability was also present in the data coming from the baseline. Evelina, the one that reports
phonological attrition in her self-assessment, mostly complies with the grammar in (57) CONTIG-IO,
ONSET, *COMPLEXCOD >> *APPENDIX-LEFT >> *COMPLEXONS, NO-CODA, preferring heterosyllabic
syllabifications with a simple coda followed by a complex onset (e.g. [kon.djo]). Nevertheless, her
heterosyllabic syllabifications vary, as she produces either simple (grammar in (57)), or complex codas
(which result from the grammar in (58)) (e.g. [tund.je]) at similar frequencies. Nonetheless, a
statistically infrequent grammar is also observed here, as she sometimes syllabifies three-consonant
clusters as a cluster under σ2, abiding by the grammar described in (56) (e.g. [ko.zvjo]). Assimilation
(which is an implementation of the grammars in (62) and (64)) is present in half of Evelina’s data, an
outcome that might imply indecisiveness, as she also uses the grammars in (63) and (65) at an equal
frequency (e.g. [kok.θçο] but [tuk.θje]) (but cf. note 57). Alma, who reports only slight attrition, also
uses the grammars (56) and (57) interchangeably (e.g. [pe.kstik], but [tuk.step]) , at about the same
frequency, but she never produces complex codas, assimilation or fusion, thus never implements the
grammars in (58), (59), (60), (62) and (64).
The great within-subject variability attested in the subjects of the experimental group (heritage
speakers), as well as the impressive variability in the data of the control group (both within-subject
and between-subjects), call for an analysis that describes them and accounts for them. The approach
of the Multiple Parallel Grammars model (Kiparsky, 1993; Anttila, 2002a, 2002b; Anttila and Cho, 1998;
Revithiadou and Tzakosta 2004a, 2004b; Tzakosta, 2004, among others) argues for multiple parallel
grammars that are distinct from each other and are activated simultaneously, forming distinct
developmental paths which children, but also L2 learners, follow during language development.
According to the Multiple Parallel Grammars model, there are three stages in phonological acquisition:
At first, unmarked structures prevail, and markedness constraints overcome faithfulness constraints.
In the second stage, there is great variation and the child has access to every possible grammar
including the adult native speaker grammar. Children can follow any possible developmental path in
order to achieve the target grammar (e.g. Revithiadou and Tzakosta 2004b; Tzakosta 2004:224, 226).
The parallel grammars are driven by the UG, target-language typology and positive evidence (Tzakosta
2004, 2006). Grammars that are typologically closer to the target grammar are present in every
developmental stage and statistically frequent, while peripheral grammars which produce very
marked or very unmarked outputs are statistically less frequent and are, eventually, dropped
(Revithiadou & Tzakosta 2004a). In the final stage, children settle on the target grammar (the native
adult grammar), where faithfulness constraints are ranked higher than markedness constraints
(Tzakosta 2004, 2006). In the course of L2 acquisition, the speaker uses parallel grammars whose
typology is present in L1, in L2, as well as in the UG (Tzakosta, 2007:102). Given the poor quality and
57
Though this could be an effect of the experiment and not reflect their actual intuitions about phonological
processes, as fusion and assimilation result in unmarked structures and assimilation is obligatory in both their
languages.
72
quantity of input during the course of heritage language acquisition and the lack of positive evidence
that would help reach the native adult grammar (Albanian, in this case), variability in heritage
phonology can be due to multiple parallel grammars that are still active. Thus, incomplete acquisition
can mean that the speaker is “stuck” in the second developmental stage. This is in line with Montrul’s
(2008, 2016) claim that heritage grammars resemble grammars in early stages of L1 and L2
development (cf. 2.2).
In the case of the present study, there is great variability in the syllabification of simultaneous and
sequential bilinguals. The two participants of the control group grew up monolingual in a SA speaking
environment, but moved to Greece as young adults and acquired Greek as their L2 in an immersion
environment. Having lived in Greece for decades, interacting in SMG in the society they live in but,
also, at home with their children who were born in Greece and have always been dominant in SMG,
they have started manifesting signs of attrition. Alma, who visits her homeland quite often, reports
slight attrition which she describes as reduced fluency and “forgetting words”. Her data show little
variability. She accepts half of the three-consonant clusters of SA in the experiment and syllabifies
them as such. She does not accept only those clusters that are not accepted word-initially in SA
([stop+fricative+stop] clusters) and the cluster [zvj], which is extremely rare word-initially in her L1
(found only in the word zvjerdh /zvjerð/ ‘to wean’, ‘to alienate’) 58. This means that the slight variability
found in Alma’s data can be due to frequency effects and/or possible prescriptive rules of
syllabification. For instance, prescriptive rules of SMG (Triantafyllidis, 1990:20) and SA (as described in
Xhaferaj, 2018:232) dictate that a consonant sequence is syllabified under the same syllable only if
there is a word in the language starting with at least the first two consonants of said sequence. Thus,
it is possible that Alma complies with only one grammar for SA, the one she acquired when she
completed the phonological acquisition of her L1, and that her phonology remains intact.
Evelina, who has lived in Greece for 25 years and reports heavy attrition, even in the domains of
phonetics and phonology, rarely perceives the clusters used in the experiment as such. Even when she
does, there is no uniformity, i.e. she does not syllabify a cluster type under the same syllable and a
different type as heterosyllabic. She can syllabify a given cluster under the same syllable, as
heterosyllabic with a simple coda and complex onset, or as heterosyllabic with a complex coda and a
simple onset. Just like Alma, she never accepts clusters that do not occur word-initially in her L1 and
she syllabifies [zvj], which is rare in SA, under the same syllable only once. However, the variability in
her data implies some attrition in her phonological grammar and some employment of multiple
parallel grammars, at least for phonotactics.
In the experimental group, Aggeliki, who was raised bilingual (SA-SMG) in a SA-speaking
environment and was dominant in SA before moving to Greece at age 11 and becoming dominant in
SMG, shows very little variability. She perceives the vast majority of the three-consonant SA clusters
as clusters and syllabifies only some [stop+fricative+stop] clusters as heterosyllabic consonant
sequences, always with a complex coda and a simple onset. Although she produces codas that are
disallowed by SA phonotactics ([bs]), this is extremely rare (only two times). The rest of her complex
codas are allowed in SA ([ks]). Although Aggeliki reports heavy attrition in SA, her data suggest little
to no attrition, at least in the domain of phonology. It seems that her phonological acquisition was
complete before she left Albania as a child and that her phonological grammar is intact, at least for
58
The clusters [zv] and [ks] (clusters formed by the two first members of the clusters Alma does not accept) are
also rare word-initially in SA, while [bs] is not accepted word-initially (cf. 5.1.4 and 5.1.5)
73
phonotactics. However, this can also be a result of prescriptive rules she learnt at the Albanian school
and still remembers, even if she has forgotten a large part of the grammar and vocabulary of SA.
The five remaining heritage speakers in the experimental group display considerable variability. In
the vast majority of the data, they do not perceive the SA clusters in the experiment as clusters under
the same syllable. When they do, it is sporadic. They also sporadically form complex codas and they
sometimes use fusion as a strategy to simplify the cluster or avoid *NC̥ violations. The variability
attested in these five subjects implies the implementation of multiple parallel grammars in their SA
phonology, at least for phonotactics. I propose that the insufficient quantity and quality of input these
heritage speakers received throughout the course of the heritage language acquisition resulted in
incomplete acquisition, which translates into the use of multiple parallel grammars at an age past the
critical period and the period of phonological acquisition (cf. 2.2 for an extensive discussion on
incomplete acquisition and the role of input in it). The phonological acquisition of the heritage
language seems to be fossilized in a developmental stage where multiple co-grammars are activated,
as the heritage speakers have never reached the adult native speaker proficiency level, hence the adult
native speaker grammar has never prevailed over the other, peripheral grammars. Additionally, the
poor quality and quantity of input the heritage speakers received during the acquisition period, results
in poor vocabulary which can, in turn, result in incomplete acquisition of heritage language
phonotactics, as many clusters are part of words the heritage speakers may have never heard. Of
course, the syllabification of SA three-consonant clusters that are not allowed in SMG as heterosyllabic
consonant sequences could also imply some dominant language interference in the phonotactics of
the heritage language, but this cannot be confirmed by the results of this study. Finally, the nature of
the clusters used in this study can definitely play a part in the results. This means that, syllabifying the
word-medial consonant sequence as a cluster would mean that C2 and C3 would form a complex onset
under σ2 while C1 would be attached to the syllable node as appendix. This is a very complex and
marked structure, so it would be natural for participants to avoid it, especially if they do not recognize
it as a consonant cluster due to insufficient input.
74
6. General discussion and concluding remarks-implications for future research
This thesis investigated the metalinguistic intuitions of heritage speakers of Standard Albanian in
Greece, regarding the phonotactics of their heritage language. As heritage speakers tend to be
imbalanced bilinguals who grew up in a community where their one language (in the case of the
present study, Standard Modern Greek) is widely used in everyday life, media and institutions, while
their other language (in the case of the present study, Standard Albanian) is restrained in the domain
of home and family life. The language of their community becomes their dominant language while
they demonstrate reduced proficiency in the language of their home, or heritage language (cf. 2.1).
Even in cases when the heritage speaker is a sequential bilingual, who acquired the minority language
as a first language, which is especially true for child immigrants, who are born in a country where their
home language is widely used and, later in life but within the critical period (i.e. before puberty), move
to a community where their first language has a minority status, the majority language tends to
become their dominant language. Due to the reduced input they receive in the heritage language, the
acquisition of this language is not typical and incomplete acquisition is the usual outcome. Moreover,
even if the heritage language is completely acquired, it may be subject to attrition later in life, as
heritage speakers tend to interact in the minority language less and less as they grow up and start
interacting mainly with people outside their family (cf. 2.2).
Research on heritage linguistics has indicated signs of incomplete acquisition and/or attrition
especially in morphosyntax, but also in the domains of semantics, pragmatics, discourse and
vocabulary (cf. Montrul, 2016; Polinsky, 2018 for an extensive literature overview and discussion on
the findings concerning the above-mentioned domains). Although phonetics and phonology seem to
be rather unaffected by incomplete acquisition or attrition, the poor quality and quantity of input that
heritage speakers receive seems to be affecting these domains as well. Heritage speakers’ phonetic
and phonological attainment lies on a continuum, somewhere in between L2 learners’ proficiency and
the proficiency of native speakers. Although heritage speakers tend to demonstrate an advantage in
perception over L2 learners, their perception is not native-like, while their segmental and
suprasegmental production diverges from native. Apart from the fact that native speakers tend to
perceive a foreign accent in heritage speakers’ utterances, as relevant experiments have shown so far,
there are signs of interference from the dominant language not only in production of segments and
tones, but also in intonation and phonotactics. However, the field of heritage phonology is far from
being sufficiently studied and all the above indications have to be further confirmed.
The present study has focused on the phonotactic representations of heritage speakers of Standard
Albanian, who grew up in Greece, hence Standard Modern Greek is their dominant language. As
Albanians are the largest immigrant group in Greece today and, as a large population of them arrived
in Greece in the early 1990s, there is a great number of Albanian heritage speakers living in the country
nowadays. However, Albanian is far from being preserved in Greece, due to the speakers’ desire to
integrate in the community, but also due to the low status the language has, as an effect of racism
and prejudice. The lack of institutional support, the ethnocentric language ideologies of the Greek
school, which promote monolingualism and the ignorance of school teachers, who advise parents
against using their native language at home only add to this fact. For this reason, heritage speakers of
Albanian in Greece tend to have low proficiency in their heritage language (cf. 2.4).
This study used an experimental procedure in order to test phonotactic representations in heritage
speakers of Albanian. Taking advantage of the phonotactic differences between Standard Albanian
and Standard Modern Greek (cf. chapter 3), the experimental stimuli were created using threeconsonant clusters of Standard Albanian that are accepted word-medially in onset position, but are
75
not accepted in Standard Modern Greek (cf. 4.3). The participants of this study, five heritage speakers
of Albanian who were born in Greece or moved to the country during infancy and a child immigrant
who moved to Greece at the age of 11, were asked to syllabify 66 disyllabic nonce-words that
contained the clusters under investigation in word-medial position. Two Albanian immigrants, who
have lived in Greece for 19-25 years were also tested, as a control group. The aim of this experiment
was to check for dominant language interference effects in the heritage speaker’s syllabification
patterns, as well as to investigate the patterns they produce and the grammar that drives these
patterns (cf. chapter 4).
The data of this experiment show a really interesting variability in both the heritage speakers and
the Albanian first generation immigrants (cf. 5.1, 5.2). This great variability advocates for incomplete
acquisition effects in the heritage speakers, but also for attrition in the case of the immigrants and the
child immigrant. In this thesis, I proposed that the results indicate the use of multiple parallel
grammars, when it comes to heritage language phonotactics. Heritage speakers tend to not accept
three-consonant clusters that are not accepted in their dominant language and syllabify them as
heterosyllabic. However, they sporadically put additional peripheral grammars in use. The same
phenomenon was observed in the data of the first generation immigrants who were the control group
of this study, though to a lesser extent. The fact that the heritage speakers in this study employ
multiple parallel grammars in phonotactics implies that the acquisition of their heritage language
phonology remains in a developmental stage where multiple parallel grammars are used, a
developmental stage attested in child first language acquisition as well as in adult second/foreign
language acquisition. Thus, heritage speakers in this study seem to never attain the native adult stage,
where peripheral grammars are left behind, a fact suggestive of incomplete phonological acquisition,
at least for phonotactics (cf. 5.4).
However, it is hard to conclude whether the different patterns that are used are a result of
dominant language interference, cluster frequency in the heritage language, or input frequency. As
the input in the heritage language is poor in quality and quantity, it is rational to assume that heritage
speakers may have never had the chance to hear and/or use clusters that are rarely found in the
heritage language. However, this study’s results call for corroboration. Future research should focus
on testing the motivation behind the patterns in heritage phonotactics. Additional experiments should
be conducted with larger groups of heritage speakers, ideally grouped by proficiency level in the
heritage language, as well as larger samples of the baseline. These syllabification patterns should also
be tested against syllabification patterns produced by monolingual native speakers of the heritage
language (i.e. people who live in the country where the language has a majority status and have not
lived abroad for an extended period of time, which could cause attrition effects). Finally, it would be
interesting to find out what the syllabification patterns would be in stimuli consisting of real-words.
Heritage phonology, and heritage phonotactics in particular, are fields of research that remain widely
unexplored. Future research in this field would bring about results that would shed light not only on
heritage language acquisition mechanisms, but on language acquisition mechanisms in general.
76
References
Albirini, Abdulkafi and Elabbas Benmammoun (2014). Aspects of second language transfer in the oral
production of Egyptian and Palestinian heritage speakers. International Journal of Bilingualism
18 (3), 244-273.
Allen, Shanley (2007). The future of Inuktitut in the face of majority languages: Bilingualism or
language shift?. Applied Psycholinguistics, 28(3), 515-536.
Allen, Shanley, Martha Crago, and Diane Pesco (2006). The effect of majority language exposure on
minority language skills: The case of Inuktitut. International Journal of Bilingual Education and
Bilingualism, 9(5), 578-596.
Alvord, Scott M., and Brandon M. Rogers (2014). Miami-Cuban Spanish vowels in
contact. Sociolinguistic Studies, 8(1), 139.
Amengual, Mark (2016). Acoustic correlates of the Spanish tap-trill contrast: Heritage and L2 Spanish
speakers. Heritage Language Journal, 13(2), 88-112.
Androulakis, George, Anastasia Gkaintartzi, Roula Kitsiou & Sofia Tsioli (2016). Parents-Schools’
communication and Albanian as a heritage language in Greece in Trifonas, Peter Pericles and
Themistoklis Aravossitas (eds.) Handbook of Research and Practice in Heritage language
Education. Springer International Publishing. 521-538.
Anttila, Arto (2002a). Variation and Phonological Theory. In Chambers, Jack K., Peter Trudgill, and
Natalie Schilling-Estes (eds.) The handbook of language variation and change. Oxford:
Blackwell. 206-243.
Anttila, Arto (2002b). Morphologically Conditioned Phonological Alternations. NLLT, 20(1), 1-42.
Anttila, Arto and Young-mee Yu Cho (1998). Variation and Change in Optimality Theory. Lingua 104.
31-56.
Applebaum, A., & Gordon, M. (2011). A comparative phonetic study of the Circassian languages.
In Chundra Cathcart, Shinae Kang, and Clare S. Sandy (eds.) Proceedings of the 37th Annual
Meeting of the Berkeley Linguistics Society: Special Session on Languages of the Caucasus, 37(2),
3-17.
Archakis, Argiris (2014). Immigrant voices in students’ essay texts: Between assimilation and pride.
Discourse & Society 25(3):297–314.
Asherov, Daniel, Alon Fishman, and Evan-Gary Cohen (2016). Vowel reduction in Israeli heritage
Russian. Heritage Language Journal, 13(2), 113-133.
Au, Terry Kit-Fong, Janet S. Oh, Leah M. Knightly, Sun-Ah Jun, and Laura F. Romo (2008). Salvaging a
childhood language. Journal of memory and language, 58(4), 998-1011.
Au, Terry Kit-Fong, Leah M. Knightly, Sun-ah Jun, and Janet S. Oh (2002). Overhearing a language
during childhood. Psychological Science 13, 238 – 243.
77
Au, Terry Kit-Fong., and Janet S. Oh (2009). Korean as a heritage language. In Ping Li (ed.), Handbook
of East Asian Psycholinguistics, Part III: Korean Psycholinguistics. Cambridge: Cambridge
University Press. 268–275.
Audacity Team (2012) Audacity®. Version 2.3.0. Audio editor and recorder. Available from:
http://audacityteam.org/ (Accessed 26/05/2019).
Bae, Sun-Hee (2015). The Syntax-Phonology Interface in Native and Near-native Korean. Ph.D.
dissertation, Harvard University.
Baker, Wendy, and Pavel Trofimovich (2005). Interaction of native-and second-language vowel
system(s) in early and late bilinguals. Language and speech, 48(1), 1-27.
Barlow, Jessica A. (2014). Age of acquisition and allophony in Spanish-English bilinguals. Frontiers in
psychology, 5, 288.
Barlow, Jessica A., Paige E. Branson, and Ignatius S. Nip (2013). Phonetic equivalence in the acquisition
of /l/ by Spanish–English bilingual children. Bilingualism: Language and Cognition, 16(1), 68-85.
Benmamoun, Elabbas, Silvina Montrul & Maria Polinsky (2013). Defining an “Ideal” Heritage speaker:
Theoretical and Methodological Challenges. Theoretical Linguistics 39:259-294.
Bevington, Gary Loyd (1970). Albanian Phonology. PhD dissertation, Amherst: University of
Massachusetts. Distributed 1974, Munich Albanien-Institut.
Blevins, Juliette (1995). The syllable in phonological theory. In Goldsmith (ed.) The handbook of
phonological theory. Oxford: Blackwell. 206-244.
Bley-Vroman, Robert (1989). What is the logical problem of foreign language learning? In Susan M.
Gass
&
Jacquelyn
Schachter
(Eds.), Linguistic
Perspectives
on
Second
Language
Acquisition (Cambridge Applied Linguistics). Cambridge: Cambridge University Press. 41-68.
Borowsky, Toni (1986). Topics in English phonology. PhD dissertation. Amherst: University of
Massachusetts.
Bullock, Barbara E., and Chip Gerfen (2004). Phonological convergence in a contracting language
variety. Bilingualism: Language and cognition, 7(2), 95-104.
Bybee, Joan L. (2001). Phonology and Language Use. Cambridge: Cambridge University Press.
Bylund, Emanuel (2009). Maturational constraints and first language attrition. Language
learning, 59(3), 687-715.
Carletto, Calogero, Benjamin Davis, Marco Stampini & Alberto Zezza (2006). A Country on the Move:
International Migration in Post-Communist Albania. International Migration Review 40: 767–
785.
Chang, Charles B. (2016). Bilingual perceptual benefits of experience with a heritage
language. Bilingualism: Language and Cognition, 19(4), 791-809.
78
Chang, Charles B., and Yao Yao (2016). Toward an understanding of heritage prosody: Acoustic and
perceptual properties of tone produced by heritage, native, and second language speakers of
Mandarin. Heritage Language Journal.
Chang, Charles B., Yao Yao, Erin F. Haynes, and Russell Rhodes (2011). Production of phonetic and
phonological contrast by heritage speakers of Mandarin. The Journal of the Acoustical Society
of America, 129(6), 3964-3980.
Chatzidaki, Aspasia (2005). Models of bilingual behaviour in Albanian pupils families: Empirical
research findings. Epistimes Ayogis, Thematic Issue: Bilingual Pupils in Greek Schools: Teaching
Approaches and Theoretical Issues, 3: 79–102.
Chatzidaki, Aspasia and Ioanna Xenikaki (2012). Language choice among Albanian immigrant
adolescents in Greece: The effect of the interlocutor’s generation. Menon 1 (1), 4-16.
Chatzidaki, Aspasia, and Christina Maligkoudi (2013). Family language policies among Albanian
immigrants in Greece. International Journal of Bilingual Education and Bilingualism, 16(6), 675689.
Chen, Ying, Yi Xu, and Susan Guion-Anderson (2014). Prosodic realization of focus in bilingual
production of Southern Min and Mandarin. Phonetica 71: 249–70.
Chomsky, Noam (1981). Lectures on Government and binding. Dordrecht: Foris.
Chomsky, Noam, and Morris Halle (1968). The sound pattern of English. New York, NY: Harper and
Row.
Clements, Nick, and Samuel Kayser (1983). CV Phonology: A Generative Theory of the Syllable.
Linguistic Inquiry Monograph No. 9. Cambridge, MA: MIT Press.
Colantini, Laura, Alejandro Cuza, and Natalia Mazzaro (2016). Task-related effects in the prosody of
Spanish heritage speakers and long-term immigrants. In Meghan E. Armstrong, Nicholas
Henriksen, Maria del Mar Vanrell (eds.) Intonational grammar in Ibero-Romance: Approaches
across linguistic subfields, Amsterdam/Philadephia: John Benjamins Publishing Company: 3-24.
Cook, Vivian (ed.) (2003). Effects of the second language on the first. Clevedon, Buffalo, Toronto,
Sydney: Multilingual Matters.
Crain, Steven and Rosalind Thornton (1998). Investigations in Universal Grammar: A guide to
experiments on the acquisition of syntax and semantics. Cambridge, MA: MIT Press.
Cushman, Ellen (2011). The Cherokee syllabary: A writing system in its own right. Written
Communication, 28(3), 255-281.
Damanakis, Michalis (1997). I ekpedefsi ton palinostoundon ke alodapon mathiton stin Ellada.
Diapolitismiki prosegisi [The education of repatriated and foreign pupils in Greece. An
intercultural approach]. Athens: Gutenberg.
De Bot, Kees (2004). The multilingual lexicon: Modelling selection and control. International Journal
of Multilingualism, 1(1), 17-32.
79
De Houwer, Annick (1995). L’alternance codique intra-phrastique dans le discours de jeunes bilingues
[Intra-sentential codeswitching in the discourse of young bilinguals]. Acquisition et interaction
en langue étrangère 6, 39-64.
De Houwer, Annick (2009). Bilingual first language acquisition. Bristol: Multilingual Matters.
DeGraff, Michel (Ed.) (1999). Language creation and language change: creolization, diachrony, and
development. Cambridge, MA: MIT Press.
Dendrinos, Bessie, and Bessie Mitsikopoulou (Eds.) (2004). Policies of linguistic pluralism and the
teaching of languages in Europe. Athens: Metaixmio, and the National and Kapodistrian
University of Athens.
Dimakos, Ioannis, and Katerina Tasiopoulou (2003). Attitudes towards migrants: What do Greek
students think about their immigrant classmates? Intercultural Education, 14, 307–316.
Dodi, Anastas (2004). Fonetika dhe fonologjia e gjuhës shqipe [The Phonetics and Phonology of the
Albanian language]. Tirana: The Albanian Academy of Sciences.
Drachman, Gaberell (1990). Onset clusters in Greek. In Mascaro, Joan and Marina Nespor (eds.),
Grammar in Progress. Glow Essays for Henk v. Riemsdijk. Dordrecht: Foris. 113-123.
Ellis, Nick C. (2006). Selective attention and transfer phenomena in L2 acquisition: Contingency, cue
competition, salience, interference, overshadowing, blocking, and perceptual learning. Applied
Linguistics, 27, 164-94.
Ellis, Rod (1989). Understanding second language acquisition. Vol. 31. Oxford university press.
Fenyvesi, Anna (2005). Hungarian in the United States. In Anna Fenyvesi (ed.) Hungarian language
contact outside Hungary: Studies on Hungarian as a Minority Language. Amsterdam: John
Benjamins. 265-318.
Flege, James E. (2007). Language contact in bilingualism: phonetic system interactions. Laboratory
Phonology 9, 354–381.
Frangoudaki, Anna, and Thalia Dragona (Eds.) (1997). ‘Ti in’ i patrida mas?’ ethnokentrismos stin
ekpedefsi [What is our country? Ethnocentrism in education]. Athens: Alexandria.
Friedman, Victor A. (1986). Albanian grammar. 466-468.
Gkaintartzi, Anastasia (2012). Issues of bilingualism in pre-school and early primary school children:
social and educational dimensions [in Greek]. PhD dissertation, Aristotle University of
Thessaloniki.
Gkaintartzi, Anastasia, Angeliki Kiliari & Roula Tsokalidou (2014). ‘Invisible’ bilingualism- ‘invisible’
language ideologies: Greek teachers’ attitudes towards immigrant pupils’ heritage languages.
International Journal of Bilingual Education and Bilingualism, 18, 1–15.
Gkaintartzi, Anastasia, Angeliki Kiliari & Roula Tsokalidou (2016). Heritage language maintenance and
education in the Greek sociolinguistic context: Albanian immigrant parents’ views. Cogent
Education, 3, 1–17.
80
Gkaintartzi, Anastasia, Evi Markou & Roula Tsokalidou (2012). Communication with immigrant
parents: when is it feasible and effective?. Polydromo: periodical for bilingualism and
multiculturalism in education and society, 5. 8-16.
Goad, Heather, and Yvan Rose (2004). Input elaboration, head faithfulness and evidence for
representation in the acquisition of left-edge clusters in West Germanic. In René Kager, Joe
Pater, and Wim Zonneveld (eds.) Constraints in phonological acquisition. Cambridge University
Press. 109-157.
Godson, Linda (2003). Phonetics of language attrition: Vowel production and articulatory setting in the
speech of Western Armenian heritage speakers. Ph.D. dissertation, University of California San
Diego.
Godson, Linda (2004). Vowel production in the speech of Western Armenian heritage speakers.
Heritage Language Journal 2, 1-22.
Gogonas, Nikos (2007). Ethnolinguistic vitality and language maintenance in second generation
immigrants. A study of Albanian and Egyptian pupils in Athens. Unpublished DPhil dissertation,
University of Sussex.
Gogonas, Nikos (2009). Language shift in second generation Albanian immigrants in Greece. Journal
of Multilingual and Multicultural Development 30, 95–110.
Gogonas, Nikos (2010). Bilingualism and multiculturalism in Greek education. Newcastle upon Tyne:
Cambridge Scholars.
Gogonas, Nikos, and Domna Michail (2015). Ethnolinguistic vitality, language use and social
integration amongst Albanian immigrants in Greece. Journal of Multilingual and Multicultural
Development, 36(2), 198-211.
Goldsmith, John A. (1976). Autosegmental phonology. Doctoral Dissertation, MIT, Cambridge MA.
(published (1979) by: Indiana University Linguistic Club).
Goldsmith, John A. (2011). The syllable. In John A. Goldsmith, Jason Riggle and Alan C. L. Yu (eds.) The
Handbook of Phonological Theory. Cambridge, MA: Blackwell. 164-196.
Gor, Kira (2014). Raspberry, not a car: context predictability and a phonological advantage in early and
late learners’ processing of speech in noise. Frontiers in psychology, 5, 1449.
Grosjean, François (1989). Neurolinguists, beware! The bilingual is not two monolinguals in one
person. Brain and language, 36(1), 3-15.
Grosjean, François (1998). Studying bilinguals: Methodological and conceptual issues. Bilingualism:
Language and cognition, 1(2), 131-149.
Grosjean, François (2001). The bilingual’s language modes. In Janet Nicol (ed.), One mind, two
languages: Bilingual language processing. Oxford: Blackwell.
Guion, Susan G. (2003). The vowel systems of Quichua-Spanish bilinguals. Phonetica, 60(2), 98-128.
Gussmann, Edmund (2002). Phonology: analysis and theory. Cambridge University Press.
81
Harris, Michael J., and Stefan Th. Gries (2011). Measures of speech rhythm and the role of corpusbased word frequency: a multifactorial comparison of Spanish (-English) speakers. International
Journal of English Studies, 11(2), 1-22.
Hatziprokopiou, Panos (2006). Globalization, migration and socio-economic change in contemporary
Greece. Processes of social incorporation of Albanian and Bulgarian immigrants in Thessaloniki.
Hayes, Bruce and Tanya Stivers (1995). A phonetic account of postnasal voicing. MS, University of
California, Los Angeles.
Henriksen, Nicholas C. (2015). Acoustic analysis of the rhotic contrast in Chicagoland Spanish: An
intergenerational study. Linguistic Approaches to Bilingualism, 5(3), 285-321.
Hockett, Charles Francis (1955). A manual of phonology. Waverly Press.
Holmes, Janet (1992). An introduction to sociolinguistics. New York: Longman.
Hua, Zhu (2002). Phonological Development in Specific Contexts: Studies of Chinese-Speaking Children.
Bristol: Multilingual Matters.
Hua, Zhu, and Barbara Dodd (2000). The phonological acquisition of Putonghua (modern standard
Chinese). Journal of child language, 27(1), 3-42.
Huffman, Marie K. (1993). Phonetic patterns of nasalization and implications for feature specification.
In Marie K. Huffman and Rena A. Krakow (eds.) Phonetics and phonology V: velars, nasalization
and the velum. San Diego: Academic press. 303-327.
Itô, Junko (1986). Syllable theory in prosodic phonology. Ph.D. dissertation, University of
Massachusetts.
Iwasaki, Shoichi (2002). Japanese. Amsterdam/Philadephia: John Benjamins.
Jakobson, Roman (1941). Kindersprache, Aphasie und allgemeine Lautgesetze. Uppsala: Uppsala
Universitets Aarskrift. (published in English (1968) Child language, aphasia and phonological
universals. The Hague and Paris: Mouton Publishers.
Jannedy, Stefanie, and Melanie Weirich (2014). Sound change in an urban setting: Category instability
of the palatal fricative in Berlin. Laboratory Phonology, 5(1), 91-122.
Jannedy, Stefanie, Melanie Weirich, Louisa Helmeke (2015). Acoustic analyses of differences in [ç] and
[ʃ] productions in Hood German. In Maria Wolters, Judy Livingstone, Bernie Beattie, Rachel
Smith, Mike MacMahon, Jane Stuart-Smith and Jim Scobbie (eds.) Proceedings of the XVIII
ICPhS, Glasgow. Glasgow: University of Glasgow Press.
Jessner, Ulrike (2003). The nature of cross-linguistic interaction in the multilingual system. In Cenoz,
Jasone, Britta Hufeisen, and Ulrike Jessner (eds.) The multilingual lexicon. Dordrecht: Springer.
45-55.
Jia, Gisela, & Doris Aaronson (2003). A longitudinal study of Chinese children and adolescents learning
English in the United States. Applied Psycholinguistics, 24(1), 131-161.
82
Jia, Ruiting, and Johanne Paradis (2014). The Use of Referring Expressions in Narratives by Mandarin
Heritage Language Children and the Role of Language Environment Factors in Predicting
Individual Differences. Bilingualism: Language and Cognition, 18(4), 737–752.
Joseph, Brian D., and Irene Philippaki-Warburton (1987). Modern Greek. London: Croom Helm.
Jun, Sun-Ah (1996). The Phonetics and Phonology of Korean: Intonational Phonology and Prosodic
Structure. New York: Garland.
Kager, René (1999). Optimality Theory. Cambridge, UK: Cambridge University Press.
Kahn, Daniel (1976). Syllable-based Generalizations in English Phonology. Indiana University Linguistic
Club (IULC).
Kang, Yoonjung, and Naomi Nagy (2012). VOT merger in Heritage Korean in Toronto. In Paula Caxaj
(ed.) Proceedings of the 2012 Annual Conference of the Canadian Linguistic Association,
Toronto: Canadian Linguistic Association. 1–15.
Kang, Yoonjung, and Naomi Nagy (2016). VOT merger in heritage Korean in Toronto. Language
Variation and Change, 28(2), 249-272.
Kappa Ioanna (1995). Silbenphonologie im Deutschen und Neugriechischen [Syllable phonology in
German and Modern Greek]. Ph.D. dissertation (unpublished) University of Salzburg.
Kappa, Ioanna (2013a). Syllaviki domi [Syllable structure]. In Revithiadou, Anthi & Vasileios
Spiropoulos (eds.) Contrastive study of Albanian and Greek grammatical structures.
[Antiparavoliki meleti grammatikon domon Alvanikis-Ellinikis]. Aristotle University Thessaloniki.
50-68.
Kappa, Ioanna (2013b). Fonotaktikoi periorismoi [Phonotactics]. In Revithiadou, Anthi & Vasileios
Spiropoulos (eds.). Antiparavoliki meleti grammatikon domon Alvanikis-Ellinikis [Contrastive
study of Albanian and Greek grammatical structures]. Aristotle University Thessaloniki. 69-78.
Kappa, Ioanna (in press). Fonologia [Phonology].
Katsikas, Christos & Eva Politou (1999). Ektos taksis to diaforetiko? Tsigani, mionotiki, palinostoundes
ke alodapi mathites stin eliniki ekpedeusi [Is difference out of the classroom? Roma, minority,
repatriated and immigrant students in Greek education]. Athens: Gutenberg.
Keel, William D. (2015). Noun Phrase Case Shift in Volga German Varieties on the Great Plains of
Kansas. In Richard B. Page, and Michael T. Putnam (eds.) Moribund Germanic heritage
languages in North America: theoretical perspectives and empirical findings. Leiden,
Netherlands: Brill. 133-152.
Kenstowicz, Michael J. (1994). Phonology in generative grammar. Cambridge, MA: Blackwell.
Kiliari, Angeliki (2005). Polyglosia kai glosiki ekpedeusi, mia kinonioglosiki prosegisi [Multilingualism
and language education, a sociolinguistic approach]. Thessaloniki: Vanias.
Kiliari, Angeliki (2014). Mathites alvanikis katagogis kai alvaniki glossa: Stasis, dexiotites, epithimies
[Pupils of Albanian origin and Albanian language: Attitudes, skills, wishes]. In Anastasia
83
Gkaintartzi, Stavros Kamaroudis, Valbona Hystuna, & Maria Viskadouraki (Eds.), Proceedings of
international conference 2nd crossroad of languages & cultures. Thessaloniki: Faculty of
Education, Aristotle University of Thessaloniki.
Kim, Ji Young (2012). Discrepancy between the perception and production of stop consonants by
Spanish heritage speakers in the United States. Paper presented at CASPSLaP [Current
Approaches to Spanish and Portuguese Second Language Phonology], Columbia, South Carolina.
Kim, Ji Young (2015). Perception and Production of Spanish Lexical Stress by Spanish Heritage Speakers
and English L2 Learners of Spanish. In Erik W. Willis, Pedro Martín Butragueño, and Esther
Herrera Zendejas (eds.) Selected Proceedings of the 6th Conference on Laboratory Approaches
to Romance Phonology. Sommerville, MA: Cascadilla Proceedings Project. 106-128.
Kiparsky, Paul (1985). Some consequences of lexical phonology. Phonology, 2(1), 85-138.
Kiparsky, Paul (1993). Blocking in nonderived environments. In Sharon Hargus and Ellen M. Kaisse
(eds.) Studies in lexical phonology. Academic Press. 277-313.
Knightly, Leah M., Sun-Ah Jun, Janet S. Oh, and Terry Kit-Fong Au (2003). Production benefits of
childhood overhearing. The Journal of the Acoustical Society of America, 114(1), 465-474.
Kondo-Brown, Kimi (Ed.) (2006). Heritage language development: Focus on East Asian
immigrants (Vol. 32). Amsterdam/Philadephia: John Benjamins Publishing.
Köpke, Barbara, Monica S. Schmid, Merel Keijzer, and Susan Dostert (Eds.) (2007). Language attrition:
Theoretical perspectives. Amsterdam/Philadephia: John Benjamins Publishing.
Kupisch, Tanja, Alyona Belikova, Öner Özçelik, Ilse Stangen & Lydia White (2017) Restrictions on
definiteness in the grammars of German-Turkish heritage speakers. Linguistic Approaches to
Bilingualism 7(1). 1–32.
Kupisch, Tanja, Dagmar Barton, Katja Hailer, I. Stangen, Tatjana Lein, and Joost van de Weijer (2014a).
Foreign accent in adult simultaneous bilinguals. Heritage Language Journal, 11(2), 123-150.
Kupisch, Tanja, Deniz Akpınar, and Antje Stöhr (2013). Gender assignment and gender agreement in
adult bilinguals and second language learners of French. Linguistic Approaches to Bilingualism,
3(2). 150-179.
Kupisch, Tanja, Tatjana Lein, Dagmar Barton, Dawn Judith Schröder, Ilse Stangen, and Antje Stoehr
(2014b). Acquisition outcomes across domains in adult simultaneous bilinguals with French as
weaker and stronger language. Journal of French Language Studies, 24(3), 347-376.
Kurylowicz, Jerzy (1948). Contribution à la théorie de la syllable [Contribution in syllable
theory]. Biuletin Poskiego Towarszistwa Jezyko-Znawaczego, 8, 80-113.
Ladefoged, Peter, and Ian Maddieson (1996). The sounds of the world's languages (Vol. 1012). Oxford:
Blackwell.
84
Laleko, Oksana, and Maria Polinsky (2016). Between syntax and discourse: topic and case marking in
heritage speakers and L2 learners of Japanese and Korean. Linguistic Approaches to
Bilingualism, 6(4), 396-439.
Laleko, Oksana, and Maria Polinsky (2017). Silence is difficult: On missing elements in bilingual
grammars. Zeitschrift für Sprachwissenschaft, 36(1), 135-163.
Lambert, Wallace E., Richard C. Hodgson, Robert C. Gardner, and Samuel Fillenbaum (1960).
Evaluational reactions to spoken languages. The Journal of Abnormal and Social
Psychology, 60(1), 44.
Lambert, Wallace E. (1977). The effects of bilingualism on the individual: cognitive and sociocultural
consequences. In Peter Hornby (ed.), Bilingualism: psychological, social and educational
implications. New York: Academic Press. 15–27.
Lass, Roger (1984). Phonology: An introduction to basic concepts. Cambridge University Press.
Leben, William R. (1973). Suprasegmental phonology. Doctoral dissertation, Cambridge, MA.: MIT.
Leben, William R. (1978). The representation of tone. In Victoria A. Fromkin (ed.), Tone. Academic
Press. 177-219.
Lee, Chungmin (2007). Contrastive (predicate) topic, intonation, and scalar meanings. In Chungmin
Lee, Matthew Gordon, and Daniel Büring (eds.), Topic and Focus: Cross-Linguistic Perspectives
on Meaning and Intonation, Dordrecht: Springer. 151–75.
Lee-Ellis, Sunyoung (2012). Looking into Bilingualism through the heritage speaker's mind. Doctoral
dissertation. University of Maryland
Lengeris, Angelos (2013). Phonetics [Fonitiki]. In Revithiadou, Anthi & Vasileios Spiropoulos (eds.)
Antiparavoliki meleti grammatikon domon Alvanikis-Ellinikis [Contrastive study of Albanian and
Greek grammatical structures]. Aristotle University Thessaloniki. 1-49.
Lenneberg, Eric H. (1967). The Biological Foundations of Language. Hospital Practice, 2:12. 59-67.
Louden, Mark L., and B. Richard Page (2005). Stable bilingualism and phonological (non) convergence
in Pennsylvania German. In James Cohen, Kara T. McAlister, Kellie Rolstad, Jeff MacSwan (eds.)
ISB4: Proceedings of the 4th International Symposium on Bilingualism. Somerville, Mass.:
Cascadilla. 1384-1392.
Lowman, Guy Sumner (1932). The Phonetics of Albanian. Language, 8(4), 271-293.
Lukyanchenko, Anna, and Kyra Gor (2011). Perceptual correlates of phonological representations in
heritage speakers and L2 learners. In Nick Danis, Kate Mesh, and Hyunsuk Sung (eds.)
Proceedings of the 35th Annual Boston University Conference on Language Development (Vol.
2). Somerville, MA: Cascadilla Press. 414-426.
Łyskawa, Paulina, Ruth Maddeaux, Emilia Melara, and Naomi Nagy (2016). Heritage speakers follow
all the rules. Language contact and convergence in Polish devoicing. Heritage Language
Journal, 13, 219-244.
85
Mackridge, P. (1985). The Modern Greek language: A descriptive analysis of standard Modern Greek.
Oxford: Oxford University Press.
Maligkoudi, Christina (2009). Tmimata didaskalias tis Alvanikis stin Ellada. Prospathia mias protis
katagrafis [Preliminary findings on Albanian mother-tongue courses in Greece]. Epistimes
Ayogis, 1:91–106.
Maligkoudi, Christina (2010). The language education of Albanian students in Greece: Government
policies and family practices. Ph.D. dissertation, University of Crete.
Malikouti-Drachman, Angeliki (1987). Syllables in Modern Greek. In Wolfgang U. Dressler, Hans C.
Luschützky, Oskar E. Pfeiffer, John R. Rennison (eds.) Phonologica 1984. Cambridge: Cambridge
University Press. 181-187.
Manos, Ioannis, Dora Papadopoulou, Vassiliki Makrygianni & Konstantinos Kolovos (2017).
Communities in Greece, Meletontas ptyches tis alvanikis metanasteusis stin Ellada.
Thessaloniki: CDRSEE.
Mattheoudakis, Marina, Aspasia Chatzidaki & Christina Maligkoudi (2017). Heritage language classes
and bilingual competence: the case of Albanian immigrant children in Greece. International
Journal of Bilingual Education and Bilingualism: 1–17.
Mayr, Robert, Jonathan Morris, Ineke Mennen, and Daniel Williams (2017). Disentangling the effects
of long-term language contact and individual bilingualism: The case of monophthongs in Welsh
and English. International Journal of Bilingualism, 21(3), 245-267.
McCarthy, John. J. & Allan Prince (1993a). Generalized alignment. In Yearbook of morphology 1993.
Dordrecht: Springer. 79-153.
McCarthy, John. J. & Allan Prince (1993b). Prosodic morphology: Constraint interaction and
Satisfaction. Linguistics Department Faculty Publication Series. 14.
McCarthy, John. J. (ed.) (2004). Optimality theory in phonology: a reader. Malden, Oxford and Carlton:
Blackwell Publishing.
McWhorter, John (2007). Language interrupted: Signs of non-native acquisition in standard language
grammars. Oxford University Press.
Meisel, Jürgen (2001). The simultaneous acquisition of two first languages. In Cenoz Jasone and Fred
Genesee (eds.), Trends in bilingual acquisition. Amsterdam: John Benjamins. 11-41.
Meisel, Jürgen (2004). The bilingual child. In William C. Ritchie, and Tej Bhatia (eds.) The handbook of
bilingualism, Malden, Oxford and Carlton: Blackwell Publishing Ltd. 91-113.
Meisel, Jürgen (2007). The Weaker Language in Early Child Bilingualism: Acquiring a First Language as
a Second Language? Applied Psycholinguistics 28 (3), 495-514.
Meisel, Jürgen (2011). Bilingual language acquisition and theories of diachronic change: Bilingualism
as cause and effect of grammatical change. Bilingualism: Language and Cognition, 14(2), 121145.
86
Meisel, Jürgen (ed.) (1994). Bilingual First Language Acquisition-French and German Grammatical
Development. Amsterdam: John Benjamins.
Memushaj, Rami (2010). Fonetika e Shqipes Standarde. [Standard Albanian phonetics]. Tirana: Toena.
Michail, Domna (2010). Boundary crossing and boundary maintenance among the Albanian
immigrants and the Greek community in Argos Orestiko-Kastoria. In Evangelos Avdikos (Ed.),
Boundaries and Folk culture in the Balkans. Athens: Pedio.
Mitakidou, Soula, & Evgenia Daniilidou (2007). Didaskalia ke mathisi tis elinikis os defteris glossas:
Apopsis ekpedeftikon [The teaching and learning of Greek as a second language: Teachers’
views]. In Konstantinos Dinas & Anna Chatzipanayotidi (Eds.), Proceedings of the international
conference on the Greek language as a second/foreign language. Research, teaching and
learning. Thessaloniki: University Studio Press.
Mitakidou, Soula, Evangelia Tressou & Evgenia Daniilidou (2007). Cross-cultural education. A challenge
or a problem? International Journal of Educational Policy, Research & Practice:
Reconceptualizing Childhood Studies, 8:67–81.
Montrul, Silvina (2002). Incomplete acquisition and attrition of Spanish tense/aspect distinctions in
adult bilinguals. Bilingualism: Language and cognition, 5(1), 39-68.
Montrul, Silvina (2008). Incomplete acquisition in bilingualism: Re-examining the age factor (Vol. 39).
Amsterdam/Philadephia: John Benjamins Publishing.
Montrul, Silvina (2010). Current issues in heritage language acquisition. Annual Review of Applied
Linguistics 30: 3-23.
Montrul, Silvina (2011). Multiple interfaces and incomplete acquisition. Lingua, 121(4), 591-604.
Montrul, Silvina (2012). Is the heritage language like a second language?. Eurosla Yearbook, 12(1), 129.
Montrul, Silvina and Tania Ionin (2010). Transfer effects in the interpretation of definite articles by
Spanish heritage speakers. Bilingualism: Language and Cognition 13:449-473.
Montrul, Silvina (2016). The acquisition of heritage languages. Cambridge: Cambridge University
Press.
Murray, Robert W., and Theo Vennemann (1983). Sound change and syllable structure in Germanic
phonology. Language, 59(3), 514-528.
Nespor, Marina (1999) Fonologia [Phonology]. Bologna: Il Mulino (1993). Adapted in Greek by Marina
Nespor and Aggeliki Ralli (translated in Greek by Aggeliki Ralli, Athanasios Natsis, Anna
Papastavrou). Athens: Patakis.
Nikolou, Kalomoira (2013). Fonologikoi kanones tis Alvanikis [Albanian phonological rules]. In
Revithiadou, Anthi & Vasileios Spiropoulos (eds.). Antiparavoliki meleti grammatikon domon
Alvanikis-Ellinikis [Contrastive study of Albanian and Greek grammatical structures]. Aristotle
University Thessaloniki. 79-88.
87
O’Grady, William, Hye-Young Kwak, On-Soon Lee, and Miseon Lee (2011). An emergentist perspective
on heritage language acquisition. Studies in second language acquisition 33(2), 223-245.
Oh, Janet S., and Andrew J. Fuligni (2010). The role of heritage language development in the ethnic
identity and family relationships of adolescents from immigrant backgrounds. Social
Development 19 (1), 202-220.
Oh, Janet S., Sun-Ah Jun, Leah M. Knightly, and Terry Kit-Fong Au (2003). Holding on to childhood
language memory. Cognition, 86(3), B53-B64.
Oh, Janet S., Terry Kit-Fong Au, and Sun-Ah Jun (2010). Early childhood language memory in the speech
perception of international adoptees. Journal of child language, 37(5), 1123-1132.
Otheguy, Ricardo and Ana Celia Zentella (2012). Spanish in New York. New York, NY: Oxford University
Press.
Pagoni, Stamatia (1993). Phonological Variation in Modern Greek: a Government Phonology Approach.
Doctoral dissertation, UCL, London.
Pan, Ho-Hsien (2007). Focus and Taiwanese unchecked tones. In Chungmin Lee, Daniel Büring, and
Matthew Gordon (eds.), Topic and Focus: Cross-Linguistic Perspectives on Meaning and
Intonation. Dordrecht: Springer. 195–213.
Papafilis, Konstantinos (2003). Leksiko Ellino-Alvaniko/Fjalor Shqip-Greqisht [Albanian-Greek
Dictionary]. M. Sideris editions.
Peña, Marcela, Janet F. Werker and Ghislaine Dehaene-Lambertz (2012). Earlier speech exposure does
not accelerate speech acquisition. Journal of Neuroscience, 32(33), 11159-11163.
Polinsky, Maria (2006). Incomplete acquisition: American Russian. Journal of Slavic linguistics, 14(2),
191-262.
Polinsky, Maria (2008). Relative clauses in heritage Russian: Fossilization or divergent grammar.
Formal approaches to Slavic linguistics 16, 333-358.
Polinsky, Maria (2011). Reanalysis in adult heritage language: New evidence in support of
attrition. Studies in second language acquisition, 33(2), 305-328.
Polinsky, Maria (2018). Heritage Languages and Their Speakers. Cambridge University Press
Prince, Alan and Paul Smolensky (1993) Optimality Theory: Constraint interaction in generative
grammar. Optimality Theory in phonology, 3.
Rao, Rajiv (2013). Allophonic variation in the voiced stop phonemes of heritage speakers of
Spanish. 7th Heritage language Summer Institute, University of Illinois at Chicago, June, 18.
Rao, Rajiv (2014). On the status of the phoneme /b/ in heritage speakers of Spanish. Sintagma: revista
de linguística, 37-54.
Rao, Rajiv (2015). Manifestations of /bdg/ in heritage speakers of Spanish. Heritage Language
Journal, 12(1), 48-74.
88
Revithiadou, Anthi & Marina Tzakosta (2004a). Markedness hierarchies vs. positional faithfulness and
the role of multiple grammars in the acquisition of Greek. In Sergio Baauw and Jacqueline van
Kampen (eds.) Proceedings of GALA 2003. Generative Approaches to Language Acquisition 2,
Utrecht University, 377-388.
Revithiadou, Anthi & Marina Tzakosta (2004b). Alternative grammars in acquisition: markedness- vs.
faithfulness-oriented learning. In Alejna Brugos, Linnea Micciulla, and Christine E. Smith (eds.)
Proceedings of the 28th Annual Boston University Conference on Language Development:
Supplement. Somerville: Cascadilla Press.
Roca, Iggy (1994). Generative Phonetics. Routledge.
Roca, Iggy, and Wyn Johnson (1999). A course in phonology. Wiley-Blackwell.
Ronquest, Rebecca (2013). An acoustic examination of unstressed vowel reduction in heritage
Spanish. In Chad Howe, Sarah E. Blackwell, and Margaret Lubbers Quesada (eds.) Selected
proceedings of the 15th Hispanic linguistics symposium. 151-171.
Saadah, Eman (2011). The production of Arabic vowels by English L2 learners and heritage speakers of
Arabic. Doctoral dissertation, University of Illinois at Urbana-Champaign.
Saussure, Ferdinand de (1959). Course in general linguistics. Charles Bally and Albert Sechehaye (Eds.).
Wade Baskin (Trans). NY: The Philosophical Society.
Schachter,
Jacquelyn
(1990).
On
the
issue
of
completeness
in
second
language
acquisition. Interlanguage studies bulletin (Utrecht), 6(2), 93-124.
Schlyter, Suzanne (1993). The weaker language in bilingual Swedish-French children. In Ake Viberg,
and Kenneth Hyltenstam (eds.) Progression & regression in language: sociocultural,
neuropsychological, & linguistic perspectives. Cambridge: Cambridge University Press, 289-308.
Selkirk, Elisabeth O. (1982). On the major class features and syllable theory. In Mark Aronoff and
Richard T. Oerle (eds.) Language Sound Structure. Cambridge, MA.: MIT Press. 107-136.
Setatos, Marios (1974). Phonology of Modern Greek. Athens: Papazisis.
Shelton, Michael, David Counselman, and Nicolás Gutiérrez Palma (2017). Metalinguistic intuitions
and dominant language transfer in heritage Spanish syllabification. Heritage language
Journal, 14(3), 288-306.
Shin, Naomi Lapidus (2014). Grammatical complexification in Spanish in New York: 3sg pronoun
expression and verbal ambiguity. Language Variation and Change, 26(3), 303-330.
Silva-Corvalán, Carmen (1994). Language Contact and Change: Spanish in Los Angeles. Oxford:
Clarendon.
Skourtou, Eleni (2002). Bilingual pupils in Greek School [Diglossoi mathites sto sxolio]. Epistimes Ayogis
4, 11–20.
Sorace, Antonella (1993). Incomplete vs. divergent representations of unaccusativity in non-native
grammars of Italian. Second Language Research, 9(1), 22-47.
89
Steriade, Donca (1982). Greek Prosodies and the Nature of Syllabification. Doctoral Dissertation, MIT,
Cambridge, MA.
Tees, Richard C., and Janet F. Werker (1984). Perceptual flexibility: maintenance or recovery of the
ability to discriminate non-native speech sounds. Canadian Journal of Psychology/Revue
canadienne de psychologie, 38(4), 579-590.
Thepboriruk, Kanjana (2015). Thai in Diaspora: Language and Identity in Los Angeles, California. Ph.D.
Dissertation, University of Hawaii Mānoa.
Thomason, Sarah Grey, and Terrence Kaufman (2001). Language contact. Edinburgh: Edinburgh
University Press.
Trask, Robert Lawrence (1996). A dictionary of phonetics and phonology. London and New York:
Routledge.
Triantafyllidis, Manolis (1990). Neoelliniki gramatiki [Grammar of modern Greek]. Athens: Organismos
Ekdoseos Didaktikon Vivlion (ΟΕΔΒ).
Trommer, Jochen & Angela Grimm (2004). Albanian word stress. Paper presented at the Manchester
Phonology Meeting, May 20-22. [URL: https://home.uni-leipzig.de/jtrommer/papers/alb.pdf]
Tse, Holman (2016b). Contact-induced splits in Toronto Heritage Cantonese mid-vowels. Linguistica
Atlantica 35, 133–55.
Tse, Holman (2017a). Variation and change in Toronto heritage Cantonese. Asia-Pacific Language
Variation 2, 124–56.
Tse, Holman (2016a). Phonetic vs. phonological considerations in inter-generational vowel change in
Toronto Heritage Cantonese. Paper presented in the Linguistic Society of America (LSA) 2016
Annual Meeting, 07 January 2016 - 10 January 2016, Washington, DC.
Tse, Holman (2017b). Heritage language maintenance and phonological maintenance in Toronto
Cantonese monophthongs? - But they still have an “accent”! Paper presented at the Linguistic
Society of America 2017 Annual Meeting, January 2017, Austin, TX.
Tsokalidou, Roula (2005). Raising ‘bilingual awareness’ in Greek primary schools. International Journal
of Bilingual Education and Bilingualism, 8: 48–61.
Tsujimura, Natsuko (2013). An introduction to Japanese linguistics. John Wiley and Sons.
Tzakosta, Marina (2004). Multiple Parallel Grammars in the Acquisition of Stress in Greek L1. Ph.D.
thesis. University of Leiden. LOT Dissertation Series 92.
Tzakosta, Marina (2006). Developmental Paths in L1 and L2 Phonological Acquisition: Consonant
Clusters in the Speech of Native Speakers and Turkish and Dutch Learners of Greek. In Adriana
Belletti, Elisa Bennatti, Cristiano Chesi, Elisa Di Domenico & I. Ferrari (eds.), Language
Acquisition and Development: Proceedings of GALA 2005. Generative Approaches to Language
Acquisition. Cambridge Scholars Press. 630-642.
90
Tzakosta, Marina (2007). Silaviki domi kai fonotaktikoi periorismoi stin ekmathisi tis ellinikis os
defteris-ksenis glossas-mia antiparavoliki meleti [Syllabic structure and phonotactics in Greek
as a second/foreign language learning-a contrastive study]. In Revithiadou, Anthi and Marina
Tzakosta (2007). Phonology in teaching Greek as a foreign language [I fonologia sti didaskalia
tis ellinikis os ksenis glossas]. Athens: Patakis. 71-117.
Vago, Robert M. (1991). Paradigmatic regularity in first language attrition. In Herbert W. Seliger and
Robert M. Vago (eds.) First language attrition. Cambridge: Cambridge University Press. 241-252.
van Rijswijk, Remy, Antje Muntendam, and Ton Dijkstra (2017). Focus marking in Dutch by heritage
speakers of Turkish and Dutch L1 speakers. Journal of Phonetics, 61, 48-70.
Vennemann, Theo (1988) The rule dependence of syllable structure. In Robert P. Stockwell (ed.) On
language: Rhetorica, phonologica, syntactica: A festschrift for Robert P. Stockwell from his
friends and colleagues. 257-83.
Walker, Willard, and James Sarbaugh (1993). The early history of the Cherokee syllabary. Ethnohistory,
70-94.
Werker, Janet F. (1989). Becoming a native listener. American Scientist 77, 54–59.
Werker, Janet F., and Richard C. Tees (1984). Phonemic and phonetic factors in adult cross-language
speech perception. The Journal of the Acoustical Society of America, 75(6), 1866-1878.
Whitney, William Dwight (1874). Oriental and Linguistic Studies. Scribner: Armstrong.
Xhaferaj,
Artan (2018). Syllabification of Bi-Consonantal Clusters
Between Vowels
in
Albanian. International journal of English linguistics 8 (5). 230-237.
Yang, Bei (2015). Perception and production of Mandarin tones by native speakers and L2 learners.
Berlin, Heidelberg: Springer.
Yeni-Komshian, Grace H., James E. Flege, and Serena Liu (2000). Pronunciation proficiency in the first
and second languages of Korean–English bilinguals. Bilingualism: Language and cognition, 3(2),
131-149.
Yip, Virginia, & Steven Matthews (2007). The bilingual child: Early development and language contact.
Cambridge, UK: Cambridge University Press.
Zec, Draga (2007). The syllable. In Paul de Lacy (ed.) The handbook of phonological theory. Cambridge
University Press. 161-194.
91
APPENDIX I ‐ STANDARD ALBANIAN AND STANDARD MODERN GREEK CONSONANTS (PHONEMES AND ALLOPHONES)
In the cases of pairs, the first consonant is voiceless and the second consonant is voiced. Table adapted from Lengeris (2013: 38)
92
APPENDIX II - EXPERIMENTAL STIMULI
Cluster type
Cluster
[nasal + stop + liquid]
[mbl]
[mpl]
[nasal + stop + fricative]
[ndj]
[mbj]
[stop + fricative +fricative]
[z + fricative + fricative]
[kθj]
[zvj]
stressed penultimate
1.
komblo
2.
tumble
3.
pamblo
4.
pimble
5.
tomblo
6.
cemble
7.
komplo
8.
tumple
9.
pamplo
10.
pimple
11.
tomplo
12.
cemple
stressed final
1.
kambla
2.
tumblep
3.
pemblik
4.
pambla
5.
tembli
6.
cimblek
7.
kampla
8.
tumplep
9.
pemplik
10.
pampla
11.
templi
12.
cimplek
13.
kondjo
13.
kandja
14.
15.
16.
17.
18.
tundje
tundjep
cendje
14.
15.
16.
17.
18.
19.
kombjo
19.
kambja
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
tumbje
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
tumbjep
pandjo
pindje
tondjo
pambjo
pimbje
tombjo
cembje
kokθjo
tukθje
pakθjo
pikθje
tokθjo
cekθje
kozvjo
tuzvje
pazvjo
pendjik
pandja
tendji
cindjek
pembjik
pambja
tembji
cimbjek
kakθja
tukθjep
pekθjik
pakθja
tekθji
cikθjek
kazvja
tuzvjep
pezvjik
93
[stop + s + stop]
[bst]
[bsk]
[kst]
[ksk]
[ksp]
34.
35.
36.
37.
38.
39.
pabsto
40.
pibste
41.
tobsto
42.
cebste
43.
kobsko
44.
tubsce
45.
pabsko
46.
pibsce
47.
tobsko
48.
cebsce
49.
koksto
50.
tukste
51.
paksto
52.
pikste
53.
toksto
54.
cekste
55.
koksko
56.
tuksce
57.
paksko
58.
piksce
59.
toksko
60.
ceksce
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
kokspo
61.
kakspa
62.
tukspe
tukspep
63.
pakspo
64.
pikspe
65.
tokspo
66.
cekspe
62.
63.
64.
65.
66.
pizvje
tozvjo
cezvje
kobsto
tubste
pazvja
tezvji
cizvjek
kabsta
tubstep
pebstik
pabsta
tebsti
cibstek
kabska
tubscep
pebscik
pabska
tebsci
cibscek
kaksta
tukstep
pekstik
paksta
teksti
cikstek
kakska
tukscep
pekscik
pakska
teksci
cikscek
pekspik
pakspa
tekspi
cikspek
94
APPENDIX III ‐ TABLE OF DATA
Cluster type
[nasal+stop+liquid]
Cluster
[mbl]
stressed penultimate
stressed final
[mpl]
stressed penultimate
stressed final
[nasal+stop+fricative]
[ndj]
stressed penultimate
stressed final
[mbj]
stressed penultimate
stressed final
[stop+fricative+fricative]
[kθj]
stressed penultimate
stressed final
[fricative+fricative+fricative]
[zvj]
stressed penultimate
Group 1
Input
Florida
Participant 1
Vasiliki
Thodoris
Control
Evelina
['ko.mblo]
['tu.mble]
['pa.mblo]
[ka'mbla]
[tu.'mblep]
[pe.'mblik]
['ko.blo]
['tu.mble]
['pam.blo]
[kam.'bla]
[tum.'blep]
[pem.'blik]
['ko.blo]
['tum.ble]
['pam.blo]
[kam.'bla]
[tu.'blep]
[pem.'blik]
['ko.mblo]
['tu.mble]
['pam.blo]
[kam.'bla]
[tum.'blep]
[pem.'blik]
['ko.mblo]
['tum.ble]
['pam.plo]
[ka.'mbla]
['tum.blep]
[pem.'blik]
['komb.lo]
['tu.mble]
['pam.blo]
[ka.mbla]
[tumb.lep]
[pemb.lik]
['ko.mplo]
['tu.mple]
['pa.mplo]
[ka.'mpla]
[tu.'mplep]
[pe'mplik]
['kom.plo]
['tu.ble]
['pa.blo]
[kam.'pla]
[tum.'pleb]
[pem.'plik]
['kom.plo]
['tu.ble]
['pa.blo]
[kam.'pla]
[tum.'plep]
[pem.'plik]
['kom.plo] ['kom.plo] ['kom.plo]
['ko.mplo]
['tum.ble]
['tum.ple] ['tu.mble]
['tu.mple]
['pam.plo] ['pam.plo] ['pa.mblo] ['pa.mplo]
[kam.'pla]
[kam.'pla] [ka.mpla]
[ka.'mpla]
[tum.'pleb] ['tum.plep] [tum.'plep] [tu.'mplep]
[pem.'plik] [pem.'plik] [pem.blik] [pe.'mplik]
['ko.ndjo]
['tu.ndje]
['pa.ndjo]
[ka.'ndja]
[tu'ndjep]
[pe.'ndjik]
['kon.djo]
['tun.dje]
['pan.djo]
[kan.'dja]
[tun.'djep]
[pen.'djik]
['kon.djo]
['tun.dje]
['pan.djo]
[kan.'dja]
[tun.'djep]
[pen.'djik]
['kon.djo]
['tun.dje]
['pan.djo]
[kan.'dja]
[tun.'djep]
[pen.'djik]
['ko.ndjə]
['kon.djo]
['tu.ndje]
['tund.je]
['pand.jə] ['pand.jo]
[kan.'dja] [kand.ja]
[tu.'ndjep] [tund.'jep]
['pen.djik] [pend.jik]
['ko.ndjo]
['tun.dje]
['pa.ndjo]
[kan.dja]
[tun.djep]
[pen.djik]
['ko.mbjo]
['tu.mbje]
['pa.mbjo]
[ka.'mbja]
[tu.'mbjep]
[pe.'mbjik]
['ko.bjo]
['tu.bje]
['pam.bjo]
[kam.'bja]
[tu.'bjep]
[pem.'bjik]
['kom.bjo]
['tu.bje]
['pa.bjo]
[ka.'bja]
[tum.'bjep]
[pem.'bjik]
['kom.bjo]
['tum.bje]
['pam.bjo]
[kam.'bja]
[tu.'bjep]
[pemb.'jik]
['kom.bjo]
['tu.mbje]
['pam.bjo]
[ka.'mbja]
[du.'mbjep]
[pem.'bjik]
['ko.kθjo]
['tu.kθje]
['pa.kθjo]
[ka.'kθja]
[tu.'kθjep]
[pe.'kθjik]
['kok.θço]
['tuk.θçe]
['pak.θço]
[kak.'θça]
[tuk.'θçep]
[pek.'θçik]
['kok.θço]
['tu.kθçe]
['pak.θço]
[kak.'θça]
[tuk.'θçep]
[pek.'θjik]
['kok.θço]
['tuk.θçe]
['pa.kθço]
[kak.'θça]
[tuk.'θçep]
[pe.'kθjik]
['ko.zvjo]
['koz.vjo]
['koz.vjo]
['koz.vjo]
Group 2
Input
Aggeliki
Alex
['pi.mble]
['to.mblo]
['ce.mble]
[pa.'mbla]
[te.'mbli]
[ci.'mblek]
[pim.ble]
[tom.blo]
[cem.ble]
[pam.bla]
[tem.bli]
[cim.blek]
['pi.mple]
['to.mplo]
['ce.mple]
[pa.'mpla]
[te.'mpli]
[ci.'mplek]
['pi.mple]
['to.mplo]
['ce.mple]
[pa.'mpla]
[te.'mpli]
[ci.'mplek]
[pim.ple]
[tom.plo]
[cem.ple]
[pam.pla]
[tem.pli]
[cim.plek]
['pi.ndje]
['to.ndjo]
['ce.ndje]
[pa.'ndja]
[te.'ndji]
[ci.'ndjek]
['pi.ndje]
['to.ndjo]
['ce.ndje]
[pa.'ndja]
[te.'ndji]
[ci.'ndjek]
[pin.dje]
[ton.djo]
[ce.ndje]
[pan.dja]
[te.ndji]
[cin.djek]
['ko.mbjo] ['pi.mbje]
['tu.mbje]
['to.mbjo]
['pa.mbjo] ['ce.mbje]
[ka.'mbja] [pa.'mbja]
[tu.'mbjep] [te.'mbji]
[pe.'mbjik] [ci.'mbjek]
['pi.mbje]
['to.mbjo]
['ce.mbje]
[pa.'mbja]
[te.'mbji]
[ci.'mbjek]
[pim.bje]
[tom.bjo]
[cem.bje]
[pa.mbja]
[tem.bji]
[cim.bjek]
['kok.θçə] ['kok.θço]
['tu.kθje] ['tuk.θje]
['pak.θjo] ['pak.θço]
['kak.θça] [kak.'θça]
[tuk.'θçep] [tu.kθçep]
[pe.'kθjik] [pek.'θjik]
['ko.kθjo]
['tu.kθje]
['pa.kθjo]
[ka.'kθja]
[tu.'kθjep]
[pek.θjik]
['pi.kθje]
['to.kθjo]
['ce.kθje]
[pa.'kθja]
[te.'kθji]
[ci.'kθjek]
['pi.kθje]
['to.kθjo]
['ce.kθje]
[pa.'kθja]
[te.'kθji]
[ci.'kθjek]
[pik.θje]
[tok.θço]
[cek.θçe]
[pak.θça]
[tek.θjə]
[cik.θjek]
['ko.zvjo]
['koz.vjo]
['pi.zvje]
['pi.zvje]
[pis.vje]
['komb.jo]
['tumb.je]
['pa.mbjo]
[ka.'mbja]
[tum.'bjep]
[pemb.jik]
['ko.zvjo]
Alma
['ko.mblo]
['pi.mble]
['tu.mble] ['to.mblo]
['pa.mblo] ['ce.mble]
[ka.'mbla] [pa.'mbla]
[tu.'mblep] [te.'mbli]
[pe.'mblik] [ci.'mblek]
95
stressed final
[stop+fricative+stop]
[bst]
stressed penultimate
stressed final
[bsk]
stressed penultimate
stressed final
[kst]
stressed penultimate
stressed final
[ksk]
stressed penultimate
stressed final
[ksp]
stressed penultimate
stressed final
['tu.zvje]
['pa.zvjo]
[ka.'zvja]
[tu.'zvjep]
[pe.'zvjik]
['tuz.vje]
['paz.vjo]
[ka.'zvja]
[tuz.'vjep]
[pez.'vjik]
['tuz.vje]
['pa.zvjo]
[kaz.'vja]
[tuz.'vjep]
[pez.'vjik]
['tuz.vje]
['paz.vjo]
[ka.'zvja]
[tuz.'vjep]
[pez.'vjik]
['tu.zvje]
['paz.vjo]
[kaz.'vja]
['tuz.vjep]
[bez.'vjik]
['tuz.vje]
['paz.vjo]
[kaz.'vja]
[tuz.vjep]
[pez.'vjik]
['tuz.vje]
['paz.vjo]
[ka.'zvja]
[tuz.vjep]
[pez.vjik]
['to.zvjo]
['ce.zvje]
[pa.'zvja]
[te.'zvji]
[ci.'zvjek]
['to.zvjo]
['ce.zvje]
[pa.'zvja]
[te.'zvji]
[ci.'zvjek]
[tos.vjo]
[ces.vje]
[pas.vja]
[tes.vji]
[cis.vjek]
['ko.bsto]
['tu.bste]
['pa.bsto]
[ka.'bsta]
[tu.'bstep]
[pe.'bstik]
['kob.sto]
['tu.bste]
['pab.sto]
[kab.'sta]
[tub.'step]
[peb.'stik]
['kob.sto]
['tub.ste]
['pab.sto]
[kab.'sta]
[tub.'step]
[peb.'stik]
['kob.sto]
['tub.ste]
['pab.sto]
[kab.'sta]
[tub.'step]
[peb.'stik]
['kob.sto]
['tum.ste]
['pab.ʃto]
[kap.'sta]
[tub.'step]
[peb.'stik]
['kobs.to]
['tubs.te]
['pabs.to]
[kab.'sta]
[tub.'step]
[pebs.tik]
['kob.sto]
['tub.ste]
['pa.bsto]
[kab.sta]
[tub.step]
[pe.'bstik]
['pi.bste]
['to.bsto]
['ce.bste]
[pa.'bsta]
[te.'bsti]
[ci.'bstek]
['pi.bste]
['to.bsto]
['ce.bste]
[pa.'bsta]
[te.'bsti]
[ci.'bstek]
[pi.bste]
[tob.sto]
[ceb.ste]
[pab.sta]
[tebs.ti]
[cib.stek]
['ko.bsko]
['tu.bsce]
['pa.bsko]
[ka.'bska]
[tu.'bscep]
[pe.'bscik]
['kob.sko]
['tu.bsce]
['pab.sko]
[kab.'ska]
[tub.'scep]
[peb.'ʃcik]
['kob.sko]
['tu.bsce]
['pab.sko]
[kab.'ska]
[tub.'scep]
[peb.'scik]
['kob.sko]
['tub.sce]
['pab.sko]
[kab.'ska]
[tub.'scep]
[peb.'scik]
['kob.ʃko]
['du.bsce]
['pab.ʃko]
[kab.'ska]
['tub.scep]
[peb.'scik]
['kobs.ko]
['tubs.ce]
['pab.sko]
[kabs.ka]
[tubs.cep]
[peb.'scik]
['kob.sko]
['tub.sce]
['pa.bsko]
[kab.ska]
[tub.scep]
[peb.scik]
['pi.bsce]
['to.bsko]
['ce.bsce]
[pa.'bska]
[te.'bsci]
[ci.'bscek]
['pibs.ce]
['to.bsko]
['cebs.ce]
[pa.'bska]
[te.'bsci]
[ci.'bscek]
[pib.sce]
[tob.sko]
[ce.bsce]
[pab.ska]
[teb.sci]
[cib.scek]
['ko.ksto]
['tu.kste]
['pa.ksto]
[ka.'ksta]
[tu.'kstep]
[pe.'kstik]
['ko.ksto]
['tu.kste]
['pak.sto]
[kak.'sta]
[tuk.'step]
[pek.'stik]
['ko.ksto]
['tu.kste]
['pak.sto]
[kak.'sta]
[tuk.'step]
[peks.'tik]
['kok.sto]
['tuk.ste]
['paks.to]
[kak.'sta]
[tuk.'step]
[pek.'stik]
['kok.ʃto]
['tu.kste]
['pak.ʃto]
[kak.'.ʃta]
[tuk.'step]
[pek.'stik]
['koks.to]
['tuks.te]
['pak.sto]
[kak.'sta]
[tuk.'step]
[peks.'tik]
['kok.sto]
['tu.kste]
['pak.sto]
[kak.sta]
[tuk.step]
[pe.'kstik]
['pi.kste]
['to.ksto]
['ce.kste]
[pa.'ksta]
[te.'ksti]
[ci.'kstek]
['pi.kste]
['to.ksto]
['ce.kste]
[paks.'ta]
[te.'ksti]
[ci.'kstek]
[piks.te]
[toks.to]
[ceks.te]
[paks.ta]
[teks.ti]
[cik.stek]
['ko.ksko]
['tu.ksce]
['pa.ksko]
[ka.'kska]
[tu.'kscep]
[pe.'kscik]
['kok.sko]
['tuk.sce]
['pak.sko]
[ka.'kska]
[tuk.'scep]
[pek.'ʃcik]
['kok.sko]
['tuk.sce]
['pak.sko]
[kak.'ska]
[tuk.'scep]
[pek.'scik]
['kok.sko]
['tuk.sce]
['pak.sko]
[kak.'ska]
[tuk.'scep]
[peks.'cik]
['kok.ʃko]
['tuk.sce]
['pak.ʃko]
[kak.'ska]
[tuk.'scep]
[bek.scik]
['koks.ko]
['tuk.sce]
['paks.ko]
[kaks.ka]
[tuks.cep]
[peks.cik]
['kok.sko]
['tuk.sce]
['pak.sko]
[kak.ska]
[tuk.scep]
[pek.scik]
['pi.ksce]
['to.ksko]
['ce.ksce]
[pa.'kska]
[te.'ksci]
[ci.'kscek]
['pi.ksce]
['toks.sko]
['ce.ksce]
[paks.'ka]
[teks.'ci]
[ciks.'cek]
[piks.ce]
[toks.ko]
[ceks.ce]
[paks.ka]
[teks.ci]
[cik.scek]
['ko.kspo]
['tu.kspe]
['pa.kspo]
[ka.'kspa]
[tu.'kspep]
[pe.'kspik]
['kok.spo]
['tuk.spe]
['pak.spo]
[kak.'spa]
[tuk.'spep]
[pek.'ʃpik]
['kok.spo]
['tuk.spe]
['pak.spo]
[kak.'spa]
[tu.'kspep]
[peks.'pik]
['kok.spo]
['tuk.spe]
['pak.spo]
[kak.'spa]
[tuk.'spep]
[pek.'spik]
['kok.ʃbo]
['tuk.spe]
['pak.ʃpo]
[kak.'spa]
[tuks.pep]
[pek.'spik]
['koks.po]
['tuks.pe]
['paks.po]
[kaks.pa]
[tuks.pep]
[peks.'pik]
['kok.spo]
['tuk.spe]
['pak.spo]
[kak.spa]
[tuk.spep]
[pe.'kspik]
['pi.kspe]
['to.kspo]
['ce.kspe]
[pa.'kspa]
[te.'kspi]
[ci.'kspek]
['pi.kspe]
['to.kspo]
['ce.kspe]
[paks.'pa]
[teks.'pi]
[ciks.'pek]
[piks.pe]
[toks.po]
[cek.spe]
[paks.pa]
[teks.pi]
[cik.spek]
96
Katerina Iliopoulou