Beruflich Dokumente
Kultur Dokumente
Mark Antoniou
2. Theories of bilingualism 4
3.3 PAM-L2 16
5.2 Rationale 37
6.2 Experiment 1: Production of Greek and AusE stop voicing: /p, t, k/ vs. /b, d, g/ 40
6.2.1 Aim 41
2
6.2.2 Study design 41
6.2.3 Predictions 43
6.3 Experiments 2A-2C – Perception of Greek and AusE stop voicing distinctions 45
6.3.1 Aims 46
6.4.1 Aims 51
6.4.3 Predictions 51
7. Anticipated impact 53
8. Timeline 55
9. References 57
3
Linguistic context and bilingual speech
1. Overview
The present proposal will investigate the effects of linguistic context on the production
bilinguals. Contextual effects have been addressed in theories of bilingual word selection,
code switching, and other aspects of language use, but not by theories of phonetics and
phonology, which address the system of spoken consonants, vowels, and other sound
have investigated the influence that linguistic context can exert on bilingual speech
perception and production. A new theoretical framework will be presented that combines
the principles of existing theories of bilingual language selection with those of cross-
language of presentation.
2. Theories of bilingualism
and lexical inhibition to account for variability in the performance of bilinguals in their
two languages. Note that these models focus mainly on lexical (knowledge of and mental
access of words) and/or grammatical tasks, rather than specifically on the bilingual’s
phonological system.
4
Linguistic context and bilingual speech
Grosjean’s (1989; 1998; 2001) Language Mode framework postulates that at any time,
mode when interacting with monolingual speakers of one of their languages. In this case,
the other language is said to be deactivated (although never completely). Bilinguals will
be in bilingual mode when interacting with a bilingual speaker of the same two
languages. Under these circumstances, both languages are activated but one is used for
processing language and is therefore more active than the other. According to Grosjean,
the bilingual’s language mode affects both language comprehension and performance.
Any number of factors can position a bilingual speaker or listener at a particular point on
the language mode continuum: the person(s) being spoken or listened to, the situation, the
form and content of the message, the function of the language act, and specific research
factors (does the participant know the aims of the study taking place?). Even hinting at
the bilingual abilities of a bilingual participant may put them into bilingual mode
(Grosjean, 1998). Movement along the continuum can happen at any time depending on
the factors mentioned above, usually unconsciously, smoothly and effortlessly. This may
changes in context and/or interlocutor, e.g., child-directed speaking register versus adult-
directed register.
5
Linguistic context and bilingual speech
performance across a range of linguistic tasks, the Inhibitory Control model (Green,
1998) attempts to explain how bilinguals are able to select the correct word (i.e., lexical
item) in the correct language. The Inhibitory Control model assumes that the two (or
more) languages of an individual are subsets of the language system as a whole. Lexical
(consonant-vowel structure), and phonetic (spoken form) properties. All lemmas are
tagged with a language label (language A or language B). A language task schema alters
the activation levels of representations within a given language system. Lemmas with
incorrect language tags are inhibited by the language task schemas. The activation of
specific lemmas requires input either from an external source (heard or read words) or
from the conceptual system, and is therefore reactive. In this way, word candidates in
Unfortunately, the Language Mode and Inhibitory Control models are inadequate for
guiding research at the phonological and phonetic levels. They were designed to address
lexical and syntactic aspects of language use, rather than the sound structure aspects.
These models are not capable, in their current form, of putting forth any testable
although interesting, merely knowing that bilinguals can move between language modes
6
Linguistic context and bilingual speech
is not yet sufficiently developed to make predictions about how the language of
proposed research will be the first systematic investigation of linguistic context on early
bilingual speech perception and production at the phonetic and phonological levels. In the
present research, an important distinction is made between people who are actively
acquiring a second language (L2) – L2 learners – and those who have acquired their
languages and achieved a level of language stability – bilinguals. It is the latter group
which will serve as the focus of the proposed research. Existing theories of cross-
language speech production and perception have focused on L2 learners (Flege, 1995)
and nonnative listeners (Best, 1995), but have not addressed the performance of fluent,
stable bilinguals.
bilinguals
the perception and production of speech. Language experience shapes perception so that
information relevant to the native language is picked up from the speech stream while
other information is ignored, which in turn affects speech production. The majority of
cross-language research has focused on monolinguals. However, about two thirds of the
present a special opportunity to test the predictions of the theoretical models of nonnative
speech perception and production as they possess more than one language.
7
Linguistic context and bilingual speech
keep their languages separate (Flege & Eefting, 1987b; Pallier, Colome & Sebastian-
Gallés, 2001; Mack, 1989). Bilinguals rarely possess an equal command of both
languages. This is because each language serves different social functions. Each language
is used with different people, in different places, for different purposes (Grosjean, 1982;
1989). However, researchers have often ignored the linguistic context in which the
research has been conducted. Evidence exists that bilinguals are able to shift their
production and perception along a continuum to adjust to the linguistic context (Grosjean,
1989; 1998; 2001; Green, 1998). This implies that bilinguals are able to select language-
specific information from the speech stream and that this selectivity in attention to
phonetic details may vary depending on the linguistic context. Obviously, this ability to
shift between languages is unique to people who speak more than one language. Less
fluent people who are still acquiring their second language (L2 learners) may not be able
to shift their language mode yet, and this may contribute to their difficulties in correctly
perceiving and producing many L2 phones. If fluent bilinguals are able to shift their
necessary to provide a theoretical explanation of how this occurs and to examine what
The Speech Learning Model (SLM: Flege, 1995) attempts to explain age-related
perception). By first predicting whether first language (L1) and L2 phones will be
8
Linguistic context and bilingual speech
elements of the L1, guiding proper production of those L1 elements. It is assumed that
these speech-learning mechanisms remain intact throughout life. This does not
necessarily mean that native-like L2 production will ever be achieved, especially by late
L2 learners, because there are ongoing changes throughout life in the way that the L1 and
SLM predicts that production and perception difficulties arise from L1/L2 phonetic
sound. This leads to one of the central assumptions of SLM, that many (although not all)
(consonant or vowel) perception. Put simply, the accuracy with which L2 segments are
According to SLM, L2 learners possess one phonological system. This one system
phone, production of the corresponding L1 phone will move away from the L1 phonetic
segments, but they typically approximate L2 phonetic norms for certain L2 phonetic
9
Linguistic context and bilingual speech
segments more closely over time as they gain experience in the L2. This may, however,
SLM predicts that L2 phonetic segments and contrasts that are very different from any L1
phonetic category will come to be perceived and produced relatively accurately, whereas
phones for new category formation to occur. It is predicted that early L2 learners will be
Initially, learners will perceptually relate positional allophones (similar phones perceived
the L2 to the closest positionally defined allophone in the L1. If an L2 phone sounds
classified” in Flege’s original terms) into the existing L1 category. Sounds that are
perceptually linked to one another across L1 and L2 are called diaphones, and come to
resemble each other in production. Despite audible differences between the L2 phone and
the closest L1 phone, a new category may fail to be established. This is called category
assimilation and results in a merged L1/L2 phonetic category, which may be unlike that
phonetic category will be used to produce corresponding speech sounds in the L1 and L2.
SLM would predict that discrimination of these L1 and L2 phones would be difficult, and
production would be accented. The proposed research will investigate whether Greek-
Australian English bilinguals are able to make such distinctions of stop consonants in
10
Linguistic context and bilingual speech
their two languages, and whether they are systematically affected by the syllable position
With experience, L2 learners may gradually discriminate between certain L2 sounds and
the closest L1 sound. When this occurs, a phonetic category representation may be
previously for L1 sounds. The likelihood of a category being formed for an L2 phone
increases with the degree of perceived dissimilarity from the closest L1 phone. If the L2
phone is very different from any L1 phonetic category, then it will not be assimilated to
any L1 phonetic category, and will establish a new, unique phonetic category. A newly
established L2 phonetic category and the nearest L1 phonetic category will shift away
from one another in the phonetic space, called phonetic category dissimilation. SLM
posits that category dissimilation occurs because L2 learners strive to maintain phonetic
contrast between all of the L1 and L2 phonetic categories in the common phonetic space.
from an L1 phone and will therefore differ from a native speaker’s phonetic category for
In summary, according to SLM, some L2 phonemes will acquire their own phonetic
category while others will not, and this will have implications for both L1 and L2
bilingual perception or production may shift according to the linguistic context, nor even
11
Linguistic context and bilingual speech
Like SLM, the Perceptual Assimilation Model (PAM: Best, 1993; 1994; 1995) of
nonnative speech perception proposes that perceivers possess one phonological system.
PAM also argues that nonnative speech segments are categorised according to their
perceived similarities to, or differences from, the phonetic segments in the native
phonological system. The PAM framework is built upon direct realist principles. Direct
realism is a philosophy that posits that we know the external world directly, without
informational media, such as air and light, to our perceptual systems. These informational
media are structured by distal events in the external world. For speech, the acoustic and
optical signal of the talker is structured by the gestures of their vocal tract, which are then
perceived through lawful changes in air pressure and the reflection of light. Therefore,
when we perceive speech, we directly perceive the vocal tract gestures – the distal event.
The most notable difference between PAM and SLM is PAM’s assumption that there is
speech perception, while SLM proposes only a phonetic level. Traditionally, it has been
assumed that speech has both a physical structure and an abstract, or cognitive, structure.
However, PAM is based upon direct realist metatheoretical assumptions. How can PAM
account for these two seemingly different (physical and abstract) structures of speech?
In articulatory phonology (Browman & Goldstein, 1986, 1989, 1992, 1993, 1995, 2000;
Goldstein & Browman, 1986), these seemingly different domains are the macroscopic
degree of resolution differs at these two levels. The microscopic (phonetic) level captures
12
Linguistic context and bilingual speech
much more of the detailed articulatory properties – much more detail than is needed to
distinguish words. This is the level of gradient physical details. The macroscopic
level is only concerned with the details needed to make phonemic distinctions (used to
differentiate words). This is a global level in which the structure is more abstract. The
Traditionally, phonemes have been defined as sets of static features – targets that do not
change over time. By defining phonemes as static feature bundles, it is difficult to explain
historical language changes, phonological patterning (the way sounds function within a
constriction in the vocal tract. By defining gestures dynamically, it is possible to link the
macroscopic and microscopic elements of speech, and explain language change and
phasing of gestures relative to each other. A linguistic utterance is a constellation that can
configuration. Two speech segments will contrast if the gestural constellations differ, e.g.
if a gesture is absent in one segment, if there are differences in the degree of constriction
among gestures, or if gestures are ordered or phased differently (Browman & Goldstein,
1992). Gestures are dynamic physical events that occur in space and time, and can
overlap. Overlap may occur because several gestures can be active in the vocal tract at
13
Linguistic context and bilingual speech
any point in time. This overlap may result in microscopic (phonetic-acoustic) variations
According to PAM, the native phonological system, as well as phonetic details – even
Perceivers can assimilate nonnative segments into native phonological categories, used to
distinguish words (e.g. native /b/ vs. native /p/). However, perceivers may also be
differences (e.g. accented vs. unaccented consonants and vowels). Nonnative segments
may be:
2) Perceived as a speech sound that does not fit into any single existing native
category; or
14
Linguistic context and bilingual speech
1. Two-category assimilation (TC) in which the nonnative segments are assimilated into
excellent;
assimilated to the same native phonological category, but they differ in that one
segment is a good exemplar of that native category while the other is deviant, and
discrimination should be moderate to very good depending on how big the difference
of category-goodness is;
to the same native phonological category but are equally discrepant from the native
assimilated to a native phonological category, whilst the other falls within the
phonetic space but not into any native phonological category, and discrimination
For a full description of possible perceptual assimilations see Best (1993; 1995).
PAM predicts that listeners will discriminate nonnative phones in reference to the
phonological categories of their native language. PAM was not originally designed to
15
Linguistic context and bilingual speech
predict L2 learning. However, the original PAM framework has been extended to account
3.3 PAM-L2
PAM-L2 (Best & Tyler, in press) attempts to predict how extended L2 experience will
change a phonological system that has developed based on native language input. PAM-
L2 predicts that L2 learners will learn to perceive L2 segments with varying degrees of
category, but one is perceived as being more deviant than the other. The deviant phone
With experience, the learner should develop a new phonological category for the deviant.
category, but as equally good or poor instances of that category. The learner will have
4) Perceive both L2 phones as speech segments, but not assimilate either of them to any
L1 phonological categories. The phones may be heard as speech segments; but not as
16
Linguistic context and bilingual speech
clear examples of L1 categories. As a result, the learner may form one or two new L2
phonological categories.
The predictions put forth by PAM-L2 will be tested by the present research, applied to
fluent Greek-Australian English early sequential bilinguals. However, the early bilinguals
who will be tested in the present research had far less L1 exposure when they began
acquiring their L2, as compared to the adult L2 learners that PAM-L2 was designed for.
The results of the present research are expected to extend the principles of PAM-L2 – and
The research proposed in this thesis aims to improve our understanding of the structure of
the bilingual phonological system(s) and their finer-grained phonetic details. Research
involving bilingual participants has yielded mixed results regarding the nature of their
phonological system(s).
It is difficult to characterise the findings in the bilingual literature because some authors
only describe differences at the phonetic level while others address differences at both the
phonetic and phonological levels. This distinction is not always made clear and can lead
sometimes present seemingly ambiguous data. For example, bilinguals may appear to
have merged categories in one experiment, but may appear to maintain separate
categories in the following experiment. Part of this ambiguity can be attributed to the
17
Linguistic context and bilingual speech
Voice onset time (VOT) has long been established as a useful measure for characterising
= voiced; spread vocal folds = voiceless) between stop consonants which involve
complete closure of the articulators at some position in the vocal tract, temporarily
stopping the airflow (Lisker & Abramson, 1964). VOT is defined as the timing between
the release of the closure and the onset of vibration of the vocal folds (Lisker &
Abramson, 1964). It distinguishes voiced and/or prevoiced from voiceless stops within
most languages (e.g. /b/ vs. /p/), and provides a common basis for describing certain
is especially interesting in cases where the bilingual’s two languages make phonological
distinctions at different phonetic points along the VOT continuum, as in Greek versus
English.
languages, this is taken as evidence of merged L1-L2 phonetic categories that are used by
with significantly longer VOT values in English words than in Spanish words, however,
their English word VOT values were significantly shorter than those of American English
(AmE) monolinguals (Flege & Eefting, 1987b). Note that Spanish /p, t, k/ are
implemented with short lag unaspirated VOT in utterance-initial position, while English
/p, t, k/ have long lag aspirated VOTs. In a separate task, the Spanish bilinguals’ category
monolinguals, consistent with the VOT differences between the languages (Flege &
Eefting, 1987b).
18
Linguistic context and bilingual speech
but not the other is interpreted as evidence of a merged phonological category. Numerous
Spanish-dominant early sequential bilinguals who have learned Catalan assimilate the
Catalan-specific /e/ and /ε/ vowels to the Spanish /e/ phonological category (Pallier,
bilinguals responded faster for pairs of words that only differed in a Catalan-specific
contrast, which indicated that they thought they had encountered the word for a second
time because they failed to discriminate the Catalan contrast, having treated the two
bilinguals did not respond faster for these minimal pairs, meaning that they discriminated
the two words because they had separate phonological categories of Catalan /e/ and /ε/
(Pallier, Colome & Sebastian-Gallés, 2001). It appears that the L1 shapes the perceptual
nonnative phonemic contrasts, even if there is early and prolonged exposure to the L2.
These findings provide strong evidence that early sequential bilinguals have an integrated
In spite of evidence that bilinguals have a phonological system with merged categories
(Flege & Eefting, 1987b; Pallier, Colome & Sebastian-Gallés, 2001), there is other
evidence suggesting that bilinguals maintain separate categories for their two languages.
19
Linguistic context and bilingual speech
discrimination of /d/-/t/ and /i/-/ɪ/ continua, the location of category boundary for /d/-/t/,
and the slope and fit of /i/-/ɪ/ regression lines (Mack, 1989). Note that French /d/ is
produced with voicing lead, /t/ is short lag unaspirated, and /ɪ/ does not exist in French. In
addition, the two groups were similar in their productions with respect to VOT, vowel
duration, and almost all spectral features associated with vowels (Mack, 1989). Note that
/i/ is one of the longest vowels in English, but one of the shortest in French. These
Thus, studies are inconclusive with regard to the phonological systems of bilinguals. On
the one hand, some evidence exists that suggests that bilinguals possess one phonological
system containing merged phonetic and phonological categories from both languages
(Flege & Eefting, 1987b; Pallier, Colome and Sebastian-Gallés, 2001). However, it has
been suggested that bilinguals may keep their L1 and L2 categories separate (Mack,
1989). As a consequence, this issue remains unresolved and warrants further scientific
investigation.
Nevertheless, it is possible to draw some general conclusions from the existing literature.
There is strong evidence that bilinguals are uniquely configured as a result of their
perceptual attunement so that bilinguals are able to pick up contextual information from
the environment which affords them selective use of their two languages. If this is
20
Linguistic context and bilingual speech
correct, then a shift in linguistic context should cause a corresponding shift in the
phonetic and phonological information that a bilingual uses in speech perception and
production.
Compatible with that notion, change in language use and language setting results in
the United States and Brazil several times over the course of a year. Brazilian Portuguese
voiceless stops are short lag unaspirated, whereas in AmE, they tend to be long lag
aspirated. During her 4.5 and 4-month stays in the United States, the participant mainly
spoke AmE, and immediately upon her return to Brazil, native Brazilian Portuguese
American-sounding than recordings before her trip to the United States. Acoustic
measurements confirmed that the speaker’s voiceless stop VOT values in her native
language, Portuguese, had increased toward the longer values of English after her stay in
the United States (Sancier & Fowler, 1997). After a 2.5 month stay in Brazil, during
which she mainly spoke Portuguese, the participant returned to the United Sates.
Acoustic measures indicated that the speaker’s /p/ and /t/ in both languages had shifted
just as much, this time toward the shorter Portuguese VOT values (Sancier & Fowler,
1997). The speaker had experienced “gestural drift” in both of her languages towards the
VOTs of the language environment. These findings strongly suggest that a bilingual
are merged or at least linked. Given that productions in both languages were affected
21
Linguistic context and bilingual speech
when only one language was used for a period of a few months, the L1 and L2
addition, learning appears to be an ongoing process that may result in temporary and
context-dependent changes in speech production well into adulthood. In this sense, the
Linguistic context may provide a potentially important account for some of the
ambiguous findings in research involving bilinguals (Flege & Eefting, 1987b; Pallier,
Colome, & Sebastian-Gallés, 2001; Mack, 1989). Researchers should heed language
mode at all times, as it may influence the results of a bilingual study, and if a researcher
does not wish to make such a manipulation, then it must be controlled for (Grosjean,
1998). Theories of speech perception (PAM; PAM-L2) and production (SLM) have not
considered the role that linguistic context can play in the performance of bilingual
language mode in their methodological designs. Some of the most influential speech
perception studies involving bilinguals have been those from researchers at the
conducted by the Barcelona group, the language used by the experimenter has not been
reported at all (Pallier, Bosch, & Sebastian-Gallés, 1997; Sebastian-Gallés & Soto-
& Bosch, 2005). This suggests that the researchers have not considered the role played by
the linguistic setting in which research takes place. This lack of interest is puzzling when
22
Linguistic context and bilingual speech
one considers that the majority of the few studies that have manipulated the language of
One reason why linguistic context has been overlooked by researchers may be because
the initial two studies which attempted to put bilinguals in a monolingual mode yielded
results that suggested that the language of presentation did not exert a significant
Carbone, 1973). The researchers tested the bilingual group using an English experimenter
in an English university and a French experimenter in a French high school. All stimuli,
materials and instructions were in the language of the experimenter. The bilinguals
English words, the bilinguals’ VOTs shifted, but did not completely match the English
monolinguals. These results indicate that the two languages are not separate but inter-
dependent in bilinguals, and provide some (albeit limited) evidence that language mode
can influence bilingual production. However, the bilinguals’ perception scores were in
between the two groups of monolinguals. It seemed that bilinguals were able to adapt
their production mechanisms, but not their perceptual mechanisms, to the second
language. In perception, the stimulus, rather than the language mode, seemed to
productions made by bilinguals in both languages were similar to the results obtained
23
Linguistic context and bilingual speech
divided a synthetic VOT continuum into voiced and voiceless domains at a compromise
that question the validity of Williams’ data. The bilingual group contained some
Language dominance may exert a stronger effect than language of presentation (Hazan &
Boulakia, 1993), and therefore, may have confounded any potential linguistic context
(like Caramazza et al., 1973) was not sufficient to keep the bilingual listeners in a
monolingual mode. Providing instructions in one language will only determine what the
base language will be but will not deactivate the other language and keep participants in a
monolingual mode (Grosjean, 1998). Some of the stimuli would have been heard several
minutes after the language-specific instructions had been given, thereby reducing the
intermediate results between those of the two monolingual groups in both language
presentation conditions. Had these researchers used natural tokens and provided constant
production was conducted by Magloire and Green (1999) who examined the effect of
speaking rate on VOT of Spanish and English voiced and voiceless bilabial stops.
24
Linguistic context and bilingual speech
were contacted, instructed, and spoken to only in English. The experimenter made no
mention of their bilingual status during the testing procedure. During the debriefing, the
experimenter pointed out that an item on the questionnaire indicated that the participant
spoke Spanish, and then asked the participant to participate in a future Spanish study.
Participants were kept naive of the fact that the two experiments were part of the same
study. Spanish-English bilinguals produced VOT values in each language that were
nearly identical to their monolingual counterparts, even when their articulation was
stressed under very fast and slow speaking rates (Magloire & Green, 1999).
The first perceptual study which firmly established its participants in a monolingual
language mode was conducted by Elman, Diehl and Buchwald (1977). Naturally
produced stimuli were preceded by a carrier phrase in each language (Spanish and
English). In addition, same-language filler words were placed between test items. The
five test syllables had VOT values of -69, +15, +19, +26, and +66 milliseconds. In a pilot
study, both English monolinguals and Spanish monolinguals identified the - 69 stimulus
as /b/ and the +66 stimulus as /p/. The three intermediate stimuli were ambiguous in the
sense that English monolinguals heard them as /b/, whereas Spanish monolinguals heard
depending on the language of presentation (Elman, Diehl, & Buchwald, 1977). The
natural stimuli, the carrier sentence and the filler words provided constant language-
specific information which activated one language much more than the other and kept the
25
Linguistic context and bilingual speech
processed the stimuli in a more monolingual-like fashion, both in Spanish and English,
Of all the studies that have manipulated the language of presentation, only one was
designed with the sole purpose of examining the effect of the language of presentation on
bilingual perception and production (Flege & Eefting, 1987a). Dutch-English bilinguals
continuum. Note that Dutch /t/ is short lag unaspirated and /d/ is prevoiced. The Dutch-
English bilinguals had begun learning English at the age of 12. Instead of carrier phrases,
were required to answer. This required that participants concentrate during the task.
Surprisingly, recorded instructions were used and the experimenter minimised his contact
with participants. The reason for this was to reduce the possibility of psychosocial and
contradictory, as any effect of language mode must surely involve some interaction with
human speakers. Despite the facts that participants were made aware of the researchers'
bilinguals' category boundaries occurred at longer values in the English presentation than
in the Dutch (Flege & Eefting, 1987a). This finding directly contradicts those of previous
language of presentation studies that used synthetic stimuli yet failed to find perceptual
effects of language mode (Caramazza et al., 1973; Williams, 1977). This may have been
because the procedures used in this study were more effective (e.g. questions between
trial blocks), and the synthetic stimuli were more natural-sounding (Flege & Eefting,
1987a). The cross-language difference in VOT seen in speech production was larger than
26
Linguistic context and bilingual speech
the perceptual category boundary shift between /d/-/t/. Despite their ability to shift VOT
in Dutch and English, Dutch-English bilinguals produced English /t/ with shorter VOTs
with the best English accents produced Dutch /t/ with shorter VOTs than bilinguals with
stronger nonnative accents. It appeared that learning English had influenced their
Given the findings of Elman et al. (1977) and Flege & Eefting (1987a), it appeared that a
identification of edited natural /bεn/-/pεn/ tokens (Hazan & Boulakia, 1993). Separate test
sessions were conducted in French and English: all conversations, instructions and
questionnaires were only in one language. Each stimulus was preceded by the word
monolingual-like VOTs in French words, but only intermediate VOT values in English
words. However, these were significantly different from their French productions. In
addition, small but significant perceptual shifts in phoneme boundaries were found for
showed greater evidence of boundary shifts in perception. These results indicate that
27
Linguistic context and bilingual speech
If language mode can shift perception of L1 and L2 segments, then it follows that it
should also influence the perception of nonnative contrasts. The perceptual abilities of
identification tasks using bilabial stop voicing contrasts from Xhosa (Calderon & Best,
1996). Bilinguals were divided into Spanish or English language presentation groups,
meaning that all contact and instructions were given in only one of the bilingual’s
languages throughout the course of the study. The Spanish presentation condition, but not
the English condition had a negative effect on performance with the Xhosa prevoiced
implosive /ɓ/- prenasalised plosive /mb/ contrast. Note that neither of these consonants
occur in Spanish or English. The VOT values of Xhosa /ɓ/ and /mb/ (-60ms and -109ms,
respectively) are located towards the prevoiced Spanish /b/ (-130ms). The VOT range of
Spanish /b/ may have impeded the bilinguals’ ability to discriminate this Xhosa
distinction. Bilinguals in the English presentation condition discriminated the Xhosa /ɓ/-
/mb/ contrast because the prenasalised consonant /mb/ can be heard in English as an m+b
cluster, which differs phonologically from b, the English consonant to which the
implosive /ɓ/ was assimilated. The language of presentation put bilinguals into Spanish or
English monolingual mode, which affected the way that their phonological system
constrained their perception of the nonnative /ɓ/-/mb/ contrast. These findings provide
strong evidence indicating that bilinguals are sensitive to linguistic context, even when
While no study that has manipulated the language of presentation has involved Greek-
28
Linguistic context and bilingual speech
perception of nonnative Thai bilabial stop voicing contrasts (Beach, Burnham &
Kitamura, 2001). Because the Greek-AusE bilinguals produced stops in both languages,
this study is useful for hypothesising how Greek-Australian English bilinguals will
produced almost identical VOTs for AusE [ph], 76ms and 75ms respectively. Greek-
AusE bilinguals produced shorter VOT values for the unaspirated voiceless Greek /p/
(38ms) than their production of the long lag aspirated AusE [ph] (76ms). AusE [b] was
monolinguals (-13ms). The Greek bilinguals produced Greek [b] (-102ms) with more
voicing lead than their production of AusE [b] (-56ms). This suggested that Greek-AusE
bilinguals had separate phonetic categories for Greek [b] and AusE [b]. In a second
production task, participants imitated a native Thai speaker’s productions of Thai bilabial
stops. Greek-AusE bilinguals produced VOTs that were more extreme (more negative for
voiced, more positive for voiceless) than the Thai model’s and the AusE monolinguals’.
/ba/-/pha/, /pa/-/pha/, and /ba/-/pa/. The VOT values of the Thai stops are /ba/ (-103ms),
/pa/ (7.8ms), and /pha/ (71.9ms). Greek-Australian bilinguals were more successful at
discriminating the Thai /ba/-/pa/ contrast than AusE monolinguals. Interestingly, a link
was found between production and perception: Greek-AusE bilinguals who exaggerated
their productions of Greek versus English VOT discriminated Thai /ba/-/pa/ like Thai
native speakers (Beach, Burnham and Kitamura, 2001). However, the researchers did not
29
Linguistic context and bilingual speech
bilingual group, two of the participants were simultaneous bilinguals, and no Greek
monolingual data were collected. There was a high degree of individual variation in the
Greek-AusE bilinguals’ productions which may have been reduced if the bilinguals had
the same language dominance, and had acquired their languages in the same order and at
similar ages. The present research will address all of these shortcomings.
In all, the evidence presented here indicates that the effect of language mode on both
has been argued, no author has presented a theoretical account of how and why language
mode effects occur in bilingual speech production and perception. Providing such an
The proposed study will systematically examine the effects of language of presentation
on early bilingual perception and production of stop voicing contrasts at the phonetic
level of gradient physical details. With the exception of Flege and Eefting (1987a), past
studies that have manipulated the language of presentation have had other primary aims
(Caramazza et al. 1973; Williams, 1977; Elman, Diehl & Buchwald, 1977; Hazan &
Boulakia, 1993; Calderon & Best, 1996; Magloire & Green, 1999), and have included a
language of presentation manipulation as a secondary goal. Note also that the Dutch-
English bilinguals in the Flege and Eefting (1987a) study had acquired their L2 late, at
the age of 12. Our research will focus on the production and perception of stop
consonants by Greek-AusE bilinguals who have acquired their L2 English much earlier,
by 4-5 years of age. Recall that Beach, Burnham and Kitamura (2001) tested Greek-AusE
30
Linguistic context and bilingual speech
bilinguals. In the present research, the performance of the English (L2) dominant, Greek-
AusE early sequential bilinguals will be compared to that of groups of AusE and Greek
monolinguals.
The proposed study will involve acoustic analyses of Greek and AusE stop consonant
productions, as well as the identification and discrimination of AusE, Greek, and Xhosa
stop consonants in a variety of contexts. Most bilingual research has examined Spanish-
Catalan, French-English, or Spanish-English language pairs. Greek and English are more
distant languages with less historical contact and fewer structural similarities than these
other language pairs (Algeo & Pyles, 1993). While English contains Greek borrowings,
they are largely restricted to scientific terms, and are typically pronounced according to
English phonology and phonetics, not Greek. For these reasons, Greek and English are
useful languages for testing the predictions of SLM, PAM and PAM-L2 extended to
bilinguals. We have decided to investigate stop consonants that differ in voicing because,
English differ in the phonetic settings of VOT used for stop voicing distinctions.
However, importantly, Greek and English also differ in the phonological status of voiced
versus voiceless stop consonants in ways that are not evident in the other languages that
Australian English (AusE) is a non-rhotic dialect of English, spoken by most (but not all)
native-born Australians. It differs from other varieties of English primarily by its vowel
31
Linguistic context and bilingual speech
and 6 long vowels /iː, eː, ɐː, oː, ʉː, ɜː/ – and 6 diphthongs /æɪ, ɑe, ɔɪ, əʉ, æɔ, ɪə/ (Cox,
2006). AusE contains 24 consonant phonemes. The consonant system of AusE does not
differ greatly from most other dialects of English. The proposed research is concerned
with stop consonants, so I will only describe the AusE stop consonants.
AusE contains three voiceless (plosive) stops: the bilabial /p/, the alveolar /t/, and the
velar /k/. Voiceless stops may be long lag aspirated VOT (where there is a delay between
the release of the consonantal constriction and the beginning of voicing) or short lag
unaspirated VOT (where voicing begins at the time of the release or shortly afterwards)
depending on the context in which they occur. Voiceless stops are aspirated in word-
initial position and in stressed syllable-initial position (e.g. “push” [phʊʃ], “support”
[sɐˈphoːt], “tin” [thɪn], “factorial” [ˌfækˈthoːrɪʲəɬ], “cow” [khæɔ], and “accord” [əˈkhoːd]:
Cox & Palethorpe, in press). At the beginning of an unstressed syllable, voiceless stops
are usually shorter-VOT and weakly aspirated (e.g. “beeper” [ˈbiːpə], “looking” [ˈlʊkɪŋ]).
Voiceless stops are always unaspirated when preceded by [s] within the same syllable
(e.g. “spin” [spɪn], “stop” [stɔp̚], and “skin” [skɪn]; Cox & Palethorpe, in press).
AusE contains three voiced (plosive) stops: the bilabial /b/, the alveolar /d/, and the velar
/g/. Voiced stops have short lag unaspirated VOT in word -initial position, where /b/, /d/,
and /g/ are realised as [p], [t] and [k], e.g. “back” [pæk̚], “dig” [tɪg]), “go” [kəʉ]. But in
intervocalic (between-vowel) stress-initial position they are realized as fully voiced (e.g.
“abort” [ɐˈboːt̚], “adore” [ɐˈdoː], “again” [ɐˈgen]; Cox & Palethorpe, in press).
32
Linguistic context and bilingual speech
Note that the alveolar stops /t/ and /d/ may, in some contexts, be realised as taps
(Tollfree, 2001; Cox & Palethorpe, in press). Alveolar stops may be realised as an
alveolar tap [ɾ] at the beginning of an unstressed syllable that is not at the beginning of a
Thus, in general, AusE VOT values become more extreme (voiced – negative, voiceless –
Greek is an independent branch of the Indo-European language family. Over its 3,300
year written history, the Greek language has undergone a number of changes, however,
its dialects have never developed into mutually incomprehensible languages. Standard
Modern Greek, the official language of the Hellenic Republic of Greece, is spoken by
Yugoslav Republic of Macedonia, Italy and Turkey. There are also many Greek emigrant
communities around the world including the United States, Canada, Germany, and
Australia.
Standard Modern Greek (henceforth MG) contains the same 24 letters in its writing
26 phonemes, with most accounts ranging from 20 (Newton, 1961; Warburton, 1970;
1962; Householder, Kazazis, and Koutsoudas, 1964; Setatos, 1974; Mackridge, 1985). It
33
Linguistic context and bilingual speech
is uncontested that there are only five vowels in MG: /a, e, i, o, u/. Vowel length is not a
distinctive feature, although vowels are slightly longer when stressed. Where there is
The present proposal will investigate the production and perception of Greek stop
contains the voiceless stops bilabial /p/, dental /t/, and dorso-velar /k/. The VOTs of
voiceless stops are always short-lag unaspirated in MG, regardless of stress or syllable
position. The phonological status of each of the voiceless stops is widely accepted by
Greek linguists.
Over the past fifty years, however, one of the most passionately debated aspects of Greek
phonology has been the status of voiced stops in MG. There is no denying that the voiced
stops [b, d, g] occur phonetically in MG. However, the phonological status of these
There are two sources of voiced stops in MG. Historically, most voiced stops came from
Classical Greek sequences of nasal + voiceless stop. This provides an intuitive and
uncomplicated method of predicting the occurrence of voiced stops by rule. This rule is
doubly tempting when one examines Greek orthography, where voiced stops are
d = ντ (/n/+/t/), and g = γκ. In the Athenian dialect of MG, word-initial position voiced
stops are pronounced as [b, d, g]. In word-medial position, they are prenasalised [mb, nd,
34
Linguistic context and bilingual speech
n
g], except when following a liquid (i.e., /r/ or /l/). Note also that prenasalised voiceless
However, voiced stops also entered the MG language through foreign loan words,
particularly during the period of Turkish occupation (1453-1829) and also from the
Venetian dialect of Italian. Both of these languages contain voiceless and voiced stops as
contrastive segments. This is where the tension arises, between elegant simplicity and
empirical completeness (Viechnicki, 1995). The reality is that loan words have been
accepted into the Greek language and must be accounted for. Loan words exist in
To further complicate matters, when variations in dialect, idiolect, rate of speech, and
social register are taken into account, the variation in pronunciation is tremendous. For
example, the word άντρας ('man') would be pronounced [andras] in the Peloponnese,
[adras] in Crete, and [andras] in Athens with [d] slightly prenasalised (Viechnicki, 1995).
There have been reports that more prenasalisation occurs in slow speech, and also that it
varies according to context, possibly with an emphatic function. Variation also exists
with some words, in some dialects, that permit the use of prenasalisation of a voiced stop
in word-initial position.
So far, I have only discussed individual words. Nasals and voiceless stops may also
combine across word boundaries, resulting in prenasalised voiced stops. For example, if
35
Linguistic context and bilingual speech
the accusative article τον [ton] is followed by the name Πέτρο [Petro], this may be
pronounced [tombetro].
A number of proposals have attempted to account for voiced stops in MG, however, none
can address all these factors. A common argument for the recognition of [b, d, g] as
separate phonemes is the existence of minimal pairs such as μπετό [beˈto] - πετώ [peˈto],
κάπως [ˈkapos] – κάμπος [ˈkampos], and Τίνος [ˈtinos] - Ντίνος [ˈdinos]. To complicate
linguists do not use the term at all, and those who do use it describe different things.
When it comes to the phonological status of MG voiced stops there is anything but
consensus.
Regardless of the phonemic status of the MG sounds [p, t, k] and [b, d, g], they
incontestably exist as phonetic categories in Greek. Given that these are phonemes in
AusE, but that they are produced with different VOTs than those of MG, questions
regarding the production and perception of voiced and voiceless stops of both languages
Xhosa sounds have been chosen for perceptual tests of nonnative stop voicing contrasts
our different groups of listeners. Also, participants are unlikely to have had previous
36
Linguistic context and bilingual speech
Xhosa is a South African language spoken by people of Bantu origin. The three bilabial
Xhosa sounds relevant to this research will now be described. The implosive /ɓ/ is
produced with voicing lead (-60ms), which is presumably similar to the Greek /b/.
Voiced implosives require the larynx to be lowered during the oral closure, while the
vocal folds are vibrating. The voiced plosive /b/ is produced with short lag VOT (+9ms),
which is similar to English /b/, and presumably Greek /p/. The prenasalised /mb/ (-109ms)
contains a full murmured pre-voicing period, thus similar to initial voiced /b/ or medial
prenasalised /mb/ in Greek. The nasalisation of /mb/ is a result of the velum being lowered
5.2 Rationale
One central goal of the proposed research is to combine principles of SLM, PAM, and
PAM-L2 with language mode and inhibitory control, to account for how linguistic
continuum and/or change their base language (Grosjean, 1989, 1998, 2001). The
language-specific input (heard or read words) of the linguistic context will inhibit words
that have different language tags, enabling bilingual word selection (Green, 1998). Given
that cross-linguistic research has demonstrated that bilinguals are sensitive to the
linguistic context when producing speech (Flege & Eefting, 1987a; Hazan & Boulakia
1993; Magloire & Green, 1999), perceiving native phones (Elman, Diehl, & Buchwald,
1977; Flege & Eefting, 1987a; Hazan & Boulakia, 1993), and perceiving nonnative
37
Linguistic context and bilingual speech
phones (Calderon & Best, 1996), any theory attempting to account for bilingual speech
Existing cross-linguistic models have attempted to predict how naive listeners will
perceive nonnative contrasts (PAM), how people who are acquiring an L2 will
perceived correctly, while others will not (Best & Tyler, in press). If L2 phones are
1995). However, fluent bilinguals are not actively acquiring either of their languages.
They have reached a level of language stability and are comfortable using both of their
languages every day, often interchangeably. This, of course, does not imply that they do
not have a dominant language. It is argued that bilinguals always have a dominant
language (be it the L1 or the L2). Nor does it imply that bilinguals ever stop learning. On
the contrary, any effect of linguistic context is evidence that bilinguals are able to learn
and adapt. However, this learning differs from L2 acquisition, and for this reason fluent
bilinguals are not L2 learners. In order to predict the performance of fluent early
bilinguals in perceptual and production tasks, a new framework is required – one which
The aim of the proposed PAM-EB framework is to explain and predict early bilinguals’
Consistent with the assumptions of PAM (Best, 1995) and PAM-L2 (Best & Tyler, in
38
Linguistic context and bilingual speech
press), PAM-EB assumes that bilinguals possess an integrated phonological system. This
integrated system will contain L1, L2, and merged phonetic and phonological categories,
which have been formed as a direct result of the bilingual’s linguistic experience.
The core tenet of PAM-EB is that fluent bilinguals are sensitive to the linguistic context,
and are able to shift their perception and production when in different language modes.
These shifts are reliable, and directly depend upon the linguistic context. These shifts are
possible because phonetic categories are tagged with a language label. It is hypothesised
that speech will be assimilated to categories with the same language tag as the language
mainstream cognitive sciences. The effects of this reattunement are that bilinguals will be
It is hypothesised that fluent bilinguals are able to produce L2 phones approaching the
and L2 (e.g. AusE /t/ alveolar vs. Greek /t/ dental), fluent bilinguals will shift the place of
When bilinguals fail to identify differences in the gestural constellations of L1-L2 phones
(e.g. AusE /b/ vs. Greek /b/ which differ in their phasing of laryngeal gesture to bilabial
39
Linguistic context and bilingual speech
Language specific contrasts (L1-L1 or L2-L2) will be best discriminated in that language
wrong language mode (e.g. [b] vs. [p] will be discriminated in Greek mode, but not in
impeded if the bilingual is in a language mode in which the nonnative phones do not
contrast.
Humans of June 1999, UWS requires that all PhD students obtain their own ethics
approval from the University of Western Sydney’s Human Research Ethics Committee.
consent form etc.) is in preparation. No research will commence until the Ethics
Application has been approved. Separate ethics applications will be prepared for testing
that will occur in Greece and in Australia. This is indicated in the research timeline.
Submission of the Ethics Application for the research in Australia is expected to happen
in time for the next HREC application deadline of 2nd February 2007.
6.2 Experiment 1: Production of Greek and AusE stop voicing: /p, t, k/ vs. /b, d, g/
40
Linguistic context and bilingual speech
6.2.1 Aim
stop consonant production has ever been conducted on Greek speakers. It is required to
41
Linguistic context and bilingual speech
The reason for using these three word contexts is that stress has been shown to result in
longer VOT values in English (Cooper, 1991). Note that in the word-medial contexts,
some stops are nasalised to test the effects of Greek phonotactic rules on Greek listeners’
production and perception. In addition, the rationale for including stops from two places
of articulation is that VOT values increase as the place of closure moves further back into
the oral cavity, in English (Dorman & Raphael, 1980). We need to assess whether this
effect also occurs in Greek. Furthermore, Greek and AusE differ in the place of
articulation for /t/ and /d/ (AusE – alveolar, Greek – dental) and we plan to examine
whether this occurs systematically in the productions of our monolinguals and bilinguals.
The Greek and AusE monolinguals will be tested in their respective languages. Greek-
AusE bilinguals will be tested in the language of presentation (either Greek or AusE).
This means that all contact, instructions, stimuli, sheets, feedback and conversations will
only be in one language. For example, the carrier phrases for the respective languages
Participants will have time to familiarise themselves with the words on a sheet before
they are recorded. Participants will be instructed to produce the words at a normal
Three productions of each stop consonant in each context will be recorded. This will
contexts x 32 speakers). Target words will be excised from the recordings using PRAAT.
Acoustic measurements will be made for acoustic variables known to cue the voicing
distinction: target VOTs, F0 contour, F1 onset, and F1 cutback (Lisker, 1978).We will
42
Linguistic context and bilingual speech
determine whether place of articulation differs between Greek and English /d/-/t/ but not
/b/-/p/. We will also attempt to measure nasalisation in voiced stops in all positions.
6.2.3 Predictions
While the exact VOT values produced by Greek monolinguals are not yet known, it has
been claimed that Greeks produce voiced stops with voicing lead and voiceless stops with
short lag (Beach, Burnham, & Kitamura, 2001). Recall that AusE monolinguals produce
stress-initial voiced stops with short lag and voiceless stops with long lag. Based on these
cross-language differences, it is possible to put forth the following predictions from PAM
for monolinguals:
2. AusE monolinguals will produce more extreme VOTs (more positive – voiceless,
more negative – voiced) in medial stops when the syllable beginning with the stop
3. Given that we do not yet know the effects that syllabic stress has on Greek VOT,
we cannot make a specific prediction regarding how stress will affect the Greek
monolinguals’ productions.
Consistent with the proposed PAM-EB framework, the language of presentation will
43
Linguistic context and bilingual speech
presentation condition;
5. For the voiceless stops [ph] and [th], Greek-AusE bilinguals in the English
monolinguals;
6. For prevoiced stops [b] and [d], Greek-AusE bilinguals in the Greek presentation
8. It is not yet known what effect syllabic stress has on Greek VOTs, but it can be
9. The coronal stops /d/-/t/ will be more fronted (dental) in the Greek presentation
and more backed (alveolar) in the AusE presentation. This will be indicated in the
Note that SLM would predict that Greek-AusE bilinguals’ productions would not be
equal to that of Greek monolinguals (Greek predictions 5 and 7) due to their extended
experience with their L2 – AusE. Moreover, SLM would predict that Greek-AusE
bilinguals’ productions of English long lag aspirated stops would be equal to AusE
monolinguals’ (AusE predictions 4 and 6) because long lap stops are very different from
44
Linguistic context and bilingual speech
any L1 (Greek) phonetic category, i.e., “new” L2 categories should be formed in these
cases.
The AusE monolingual group (n = 8) will be recruited from the student population at
UWS. They will be monolingual native speakers of AusE, between 20-40 years of age,
The Greek monolingual group (n = 8) will be recruited from the student population at the
National and Kapodistrian University of Athens. They will be native speakers of Greek
with limited proficiency and exposure to English, aged between 20-40 years, with no
history of speech, language or hearing disorders. Ethics approval will be obtained before
The Greek-AusE bilinguals (n = 16) will be recruited from within the UWS student
population, the Greek societies of other universities within Sydney, and the Greek
pamphlets/posters. To be eligible, they must have been exposed to, and acquired, Greek
since birth and have acquired English by the age of five. In addition, bilinguals must be
6.3 Experiments 2A-2C – Perception of Greek and AusE stop voicing distinctions
45
Linguistic context and bilingual speech
6.3.1 Aims
Experiment 2 will test the discrimination predictions of PAM-EB (adapted from PAM-
L2) concerning perceptual shifts brought about by the language of presentation effects in
Greek stops). New groups from the same participant populations described above will be
recruited and tested on an identification task with ratings (testing for assimilation to
native language phonemes or to language of presentation for bilinguals) and also an AXB
discrimination task, for all stimulus contrasts described below. Both tasks are necessary,
as the combined results will provide the information needed to determine how the sounds
were perceptually assimilated and discriminated. Experiment 2 will be divided into three
word-medial unstressed.
If the results of Experiment 1 show that Greek-AusE bilinguals (in the respective
language presentations) shift their VOTs towards those of monolinguals, the speech of
2. If the bilinguals fail to shift their VOTs, four Greek and four AusE monolingual
speakers will be recorded instead (this possibility has been anticipated in the timeline).
Three tokens will be recorded of each speaker producing each stop in each context. Note
that if the speech of monolingual speakers is recorded, tokens from different speakers
46
Linguistic context and bilingual speech
will be used to make up the triads of tokens for the AXB discrimination task. This will
ensure that listeners will not be using speaker-specific cues to make their responses.
The target sounds /p/-/b/ and /t/-/d/ will be presented in both languages in three word
contexts. Testing will only involve within-language pairings. Note that while the short lag
unaspirated phonetic entities are similar in AusE and Greek, the phonological categories
of the respective monolingual groups will differ (e.g. [p] is an allophone of AusE word-
Greek or AusE). For example, their answer sheet will indicate “tick what you hear” or
Predictions
1. AusE monolinguals will discriminate word-initial long lag aspirated stops (e.g.
[ph]) from short lag unaspirated and voicing lead stops. AusE listeners will be
unable to distinguish the voicing lead stops from the short lag unaspirated stops –
47
Linguistic context and bilingual speech
2. Greek monolinguals will discriminate stops produced with voicing lead from
short lag unaspirated and long lag aspirated. Greeks may distinguish short lag
assimilated as good (short lag) versus poor (long lag) exemplars to the same
Based on the hypotheses of PAM-EB, it is anticipated that bilinguals will shift their
3. Bilinguals in the English presentation will discriminate long lag aspirated stops
from short lag unaspirated, but will not distinguish the voicing lead stops from the
4. Bilinguals in the Greek presentation will discriminate voicing lead stops from
short lag unaspirated. They will distinguish the short lag unaspirated stops from
Note that Experiment 2B-2C predictions concerning Greek stops may change depending
Predictions
stops (e.g. AusE [ph]) from the short lag unaspirated (e.g. AusE [p]);
48
Linguistic context and bilingual speech
2. AusE monolinguals will discriminate voiceless and voiced stops from nasalised
stops, as they phonologically contrast in AusE (e.g. ‘daddy’ vs. ‘dandy’ and
3. Greek monolinguals will not be able to discriminate AusE nasal + voiceless stop
vs. nasal + voiced stop because prenasalised voiceless stops are phonotactically
illegal in Greek;
4. Greek monolinguals will discriminate voiceless but not voiced oral stops from
6. Bilinguals in the Greek presentation will show poorer discrimination for non-
nasalised voiced stops vs. prenasalised stops, as the Greek linguistic context will
Predictions
49
Linguistic context and bilingual speech
1. AusE monolinguals will discriminate stressed intervocalic long lag aspirated stops
from the short lag unaspirated and voicing lead stops. AusE listeners will not be
able to distinguish the voicing lead stops from the short lag unaspirated;
2. AusE monolinguals will discriminate voiced stops from nasalised voiced stops;
3. Greek monolinguals will discriminate better than in the ˈVCV context, because
the stressed context will increase aspiration, maximizing the dissimilarity between
4. Greek monolinguals will discriminate voiceless but not voiced stops from
Participants will be recruited using the same methods outlined in Experiment 1. Each
group will contain 20 participants who will complete the perceptual tasks.
50
Linguistic context and bilingual speech
6.4.1 Aims
The aim of Experiment 3 is to test the predictions of nonnative speech perception made
of nonnative contrasts.
The two Xhosa contrasts that have been chosen are the implosive prevoiced /ɓ/ (-60ms)
vs. short lag voiced plosive /b/ (+9ms), and the implosive prevoiced /ɓ/ (-60ms) vs.
prenasalised /mb/ (-109ms). These contrasts will be recorded from three male, native
Xhosa speakers in both word-initial and word-medial positions. Xhosa consonants will be
consistent with Xhosa phonology and provides the greatest phonetic dissimilarity.
Greek and AusE monolinguals will be tested in their respective languages. Greek-AusE
6.4.3 Predictions
The following are based on the predictions made by PAM concerning the perception of
1. The implosive prevoiced /ɓ/ vs. short lag voiced plosive /b/ contrast:
51
Linguistic context and bilingual speech
Xhosa stops, and make a single category assimilation to the AusE /b/
phonological category;
b. In all positional contexts, the /ɓ/ - /b/ contrast will be assimilated by Greek
be excellent.
As expected by PAM-EB:
/ɓ/-/b/;
52
Linguistic context and bilingual speech
d. In word-medial position, it is expected that the latter two groups will make
7. Anticipated impact
The proposed series of experiments will be the first systematic exploration of the
and production of stop voicing contrasts. Models of nonnative speech production (SLM)
and perception (PAM; PAM-L2) will be extended to explain bilingual production and
perception. PAM-EB will be the first model to address the bilingual ability to shift
perception and production of speech dependent on linguistic context, and will provide a
significant theoretical contribution to an area that has been largely overlooked by most
researchers.
The proposed studies will also provide insights into existing ambiguous findings in the
merge phonetic (Flege & Eefting, 1987b) and phonological categories (Pallier, Colome &
Sebastian-Gallés, 2001), or maintain separate categories (Mack, 1989), but have ignored
the linguistic context in which the research occurred, should be interpreted with caution.
53
Linguistic context and bilingual speech
The data from this series of experiments also has the potential to address the longstanding
linguistic question concerning the phonological status of voiced stops in Modern Greek.
and unstressed) nasalised vs. non-nasalised stop consonants may provide insight as to
monolinguals are unable to perceive word medial AusE /p/-/b/ when following a nasal,
this would indicate that word-medial Greek [b] is an allophone of Greek /p/. If Greek-
AusE bilinguals in the AusE presentation can make this distinction, but bilinguals in the
Greek presentation cannot, this would provide additional support that the word-medial
54
8. Timeline
55
Return to Australia
Upon return to Australia, apply for ethics approval - UWS
56
9. References
Algeo, J., & Pyles, T. (2004). The Origins and Development of the English Language.
Boston: Heinle.
Beach, E., Burnham, D., & Kitamura, C. (2001). Bilingualism and the relationship
Best, C. T., & Tyler, M. D. (in press). Nonnative and second-language speech perception:
57
Browman, C. P., & Goldstein, L. (2000). Competing constraints on intergestural
Browman, C. P., & Goldstein, L. (1995). Gestural syllable position effects in American
Browman, C. P., & Goldstein, L. (1993). Dynamics and articulatory phonology. In Status
Reports on Speech Research (Vol. SR-113, pp. 51-62). New Haven: Haskins
Laboratories.
Caramazza, A., Yeni-Komshian, G. H., Zurif, E. B., & Carbone, E. (1973). The
58
French-English bilinguals. The Journal of the Acoustical Society of America,
54(2), 421-428.
Cooper, A., M. (1991). Laryngeal and oral gestures in English /p, t, k/. Actes Du XII
Cox, F. M. (2006). The acoustic characteristics of /hVd/ vowels in the speech of some
Cox, F. M., & Palethorpe, S. (in press). Australian English. Journal of the International
Phonetic Association.
Dorman, M. F., & Raphael, L. J. (1980). Distribution of acoustic cues for stop consonant
Elman, J., Diehl, R., & Buchwald, S. (1977). Perceptual switching in bilinguals. The
Flege, J. E. (1995). Second language speech learning: Theory, findings, and problems. In
6, 185-202.
59
Flege, J. E., & Eefting, W. (1987b). Production and perception of English stops by native
Grosjean, F. (1989). Neurolinguists beware: The bilingual is not two monolinguals in one
Hazan, V. L., & Boulakia, G. (1993). Perception and production of a voicing contrast by
60
Lisker, L. (1978). Rapid vs. Rabid. A catalogue of acoustic features that may cue the
Lisker, L., & Abramson, A. S. (1964). A cross-language study of voicing in initial stops:
Mack, M. (1989). Consonant and vowel perception and production: Early English-French
200.
effects on the production of voice onset time in English and Spanish. Phonetica,
56, 158-185.
Pallier, C., Bosch, L., & Sebastian-Gallés, N. (1997). A limit on behavioral plasticity in
Pallier, C., Colome, A., & Sebastian-Gallés, N. (2001). The Influence of Native-language
Sancier, M. L., & Fowler, C. A. (1997). Gestural drift in a bilingual speaker of Brazilian
61
Sebastian-Gallés, N., & Soto-Faracao, S. (1999). Online processing of native and non-
Sebastian-Gallés, N., Echeverria, S., & Bosch, L. (2005). The influence of initial
/t/. In D. Blair & P. Collins (Eds.), English in Australia (pp. 45-67). Amsterdam:
University Publications.
62
Appendix A. Language background information sheet
Participant ID:_____________
Date:____________
1. Age: ___________
_____________________________________________________________________
4. Did/do you or any of your immediate family members have any special problems
with language development (e.g. delayed language onset, serious difficulties in
learning new words or remembering the names of objects)? □ YES □ NO
If YES, who has/had the problem and what is/was the nature of the problem? At what
age did it occur? Did it/does it require a language therapist? ____________________
_____________________________________________________________________
_____________________________________________________________________
5. Did/do you or any of your family members have any special problem with speaking
(e.g. stuttering lisping, etc.)? □ YES □ NO
If YES, who? What type of problem? At what age did it occur? Did/does it require a
speech therapist? ______________________________________________________
63
____________________________________________________________________
____________________________________________________________________
6. Did/do you or any immediate family members have any special problems with
learning to read (e.g. confusing certain sounds or letters, dyslexia)?
□ YES □ NO
If YES, who? What type of problem? At what age did it occur? Did/ does I require
special education or reading tutoring?______________________________________
_____________________________________________________________________
___________________________________________________________________
Father: ______________________________________________________________
Mother: ______________________________________________________________
10. What countries have you lived in, for how long and during what age range?
64
11. What languages do you know, how long you have used or studied them, at what age
did you start learning each one, and how well you speak, read and write them?
12. Please estimate to the nearest 10% how much you speak any English in these places
or situations. Try to base your estimate on your use of English over the past 5 years.
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
While at home
Visiting family
At work
At church
Visiting friends
While on vacation
While shopping
At parties/social gatherings
65
Note: this page is to be completed by bilingual participants after the experiment has
been conducted
13. How did you learn your languages (e.g. from parents, friends, school)? Specify.
Greek: _______________________________________________________________
English: ____________________________________________________________
14. Please estimate to the nearest 10% how much you speak any Greek in these places or
situations. Try to base your estimate on your use of Greek over the past 5 years.
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
While at home
Visiting family
At work
At church
Visiting friends
While on vacation
While shopping
At parties/social gatherings
66
Appendix B. Πληροφόριση γλωσσικής καταγωγής
Ταυτότητα συμμετέχων:_____________
Ημερομηνία:____________
1. Ηλικία: ___________
____________________________________________________________________
____________________________________________________________________
4. Μήπως εσείς ή κάποιος στην οικογένεια σας έχει δυσκολία στην εκμάθηση της
γλώσσας; □ ΝΑΙ □ ΟΧΙ
Εαν ΝΑΙ, ποιος έχει τη δυσκολία; Σε ποια ηλικία συνέβη αυτό; Χρειάστηκε ιδικός
θεραπευτής γλώσσας; __________________________________________________
_____________________________________________________________________
_____________________________________________________________________
67
____________________________________________________________________
____________________________________________________________________
_____________________________________________________________________
_____________________________________________________________________
Πατέρας: ____________________________________________________________
Μητέρα: _____________________________________________________________
10. Σε ποιες χώρες έχετε ζήσει; Για πόσο χρονικό διάστημα; Και σε ποια ηλικία;
68
11. Ποιες γλώσσες μιλάτε, πόσο καιρό μιλατε αυτές τις γλώσσες, σε τι ηλικία αρχίσατε
να τις μιλάτε, και πόσο καλά μιλάτε, γράφετε κι διαβάζετε;
12. Δώστε μας περίπου το ποσοστό που μιλάτε Ελληνικά για τα τελευταία πέντε χρόνια
σε αυτές τις περιπτώσεις:
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Στο σπίτι
Επίσκεψη σε συγγενείς
Στη δουλειά
Στην εκκλησία
Με φίλους
Σε διακοπές
Στα μαγαζιά/ψώνια
Σε γιορτές/γενέθλια
69
13. Πως μάθατε αυτές τις γλώσσες (π.χ. από γωνείς, σε σχολείο);
Ελληνικά: ____________________________________________________________
Αγγλικά: _____________________________________________________________
14. Δώστε μας περίπου το ποσοστό που μιλάτε Αγγλικά για τα τελευταία πέντε χρόνια σε
αυτές τις περιπτώσεις:
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Στο σπίτι
Επίσκεψη σε συγγενείς
Στη δουλειά
Στην εκκλησία
Με φίλους
Σε διακοπές
Στα μαγαζιά/ψώνια
Σε γιορτές/γενέθλια
70