Loanword Phonology versus Second Language Phonology

advertisement
Loanword Phonology versus Second Language Phonology:
Evidence from Brazilian Portuguese
Deborah Araujo
Stony Brook University
1. Introduction
Language contact and language transfer have been largely researched in the field of
linguistics because when a language encounters a phonological structure that is not part of its
phonology, speakers of that language will find ways to replace or fix the structure so it can be
pronounceable. Lexical borrowings as well as second language acquisition, pidgins and creoles
offered some of the data that fueled such research. In the last decades, second language
acquisition has incited a wave of new discussion about the interaction of language transfer with
other factors related to language development in the process of phonological change.
In the midst of understanding and describing the linguistic systems of language contact
situations, the literature on second language phonology and loanword phonology has not
explicitly distinguished loanword phonology from second language phonology and fails to offer
a clear comparison between the two linguistic systems and the phonological processes involved
in producing each system. Are they different or are they fundamentally the same? How do they
differ? Are the phonotactics used in adapting loanwords the same used by learners of second
language? These are valid questions that need to be addressed in order for linguists to have a
better understanding of how language works in the mind of the speaker.
2
The following study offers a contrastive analysis of the adaptations caused by
monolingual Brazilian speakers to English loanwords against those done by Brazilian speakers of
English as a second language. This study will first discuss in section 2 what the literature says
about second language phonology and loanword phonology. Then section 3 will describe the
experiment. Section 4 will compare the phonotactics employed by the participants when they
were not able to pronounce the word-final coda in the same way an American would pronounce
it. Finally, section 5 will offer a discussion of the findings, connecting them to the pertinent
question of how loanword phonology differs from second language phonology.
2. Literature on second language phonology and loanword phonology
2.1. Second language phonology
The field of second language acquisition has attracted growing attention in the last few
decades as researchers try to understand the sound patterns that characterize the speech of a
person learning a foreign language. Some speakers achieve complete fluency in the second
language, to the point of passing as a native speaker. Others, however, will always have a
nonnative pronunciation or an accent. The nonnative accent is the expression of the speaker’s
interlanguage, or the intermediate place between his native language and the target pronunciation
(Major 2001). This is due to cross-linguistic influences or language transfer, where phonological
structures of the speaker’s native language exert influence upon the pronunciation of the second
language (Odlin 1989). These nonnative utterances can often give clues into the language
3
background of the speaker. Members of a speech community have strong intuitions about what
sounds native and what does not.
In a discussion of second language phonology, Major (2001) says that errors are likely to
occur when the speaker’s first language and second language have different linguistic
phenomena. These errors are due to transfer of the native language phonology onto the second
language. The English language has the phoneme [θ] in the beginning of ‘think' but neither
Brazilian Portuguese speakers nor German speakers have this sound in their phonology. Thus,
when a speaker of either language tries to speak English without having mastered the
pronunciation of this sound, they will interpret the sound through their native phonological and
will substitute it for a familiar sound. Since German and Brazilian Portuguese have different
phonologies, each speaker will make a different change to the English word. Brazilian
Portuguese speakers will usually substitute [t] for [θ], pronouncing ‘tink’ instead of ‘think’,
while German speakers will usually substitue [s], as in ‘sink’. This exemplifies the influence of
transfer in second language acquisition. In Brazilian Portuguese, speakers focus on place of
articulation, and sin [t] is a dental consonant, it is preferred to [s], an alveolar, as a substitute for
[θ] since [θ] is interdental. In German, however, speakers will focus on the mode of articulation,
and [s] is a fricative just like [θ], thus its preferred as a replacement because [t] is a plosive.
Transfer can occur in phonology as well as in syntax, morphology, discourse and any
other component of language. Van Coetsem (1988) describes the concept of transfer in second
language phonology as “imposition”. The speaker will impose his articulatory habits and his
native phonology when learning a second language because he has not acquired the phonology of
the second language yet. As a result, he will adapt the second language sound by substituting it
for a more familiar one to compensate for incomplete acquisition of the foreign phonology.
4
Past research on second language phonology has focused on contrastive analysis of
languages, which compares the similarities and differences between languages in order to
understand and predict what kind of transfer errors speakers will make throughout the language
acquisition process as well as to build an understanding about the difficulty and order of
acquisition of sounds. The tenets of contrastive analysis relied on the claim that nonnative
accents, or second language phonology, originated solely from the interaction of the native
phonology with the phonology of the target language (Major 2001). Several critics of contrastive
analysis pointed out errors in second language acquisition that could not be explained by the
speaker’s native phonology or by the target phonology, thus weakening the predictive power of
Contrastive Analysis. Other researchers pointed out that some of the adaptations present in
second language phonology were similar to those made by children when learning a first
language, which produced the idea that second language acquisition might be similar to first
language acquisition. For example, adult learners of a second language, as well as children
learning a first language, usually acquire voicing contrast of consonants in syllable-initial
position more easily than in syllable-final position. These errors are called developmental errors
and are governed by language Universals, which determine the parameters or core principles that
govern all languages. One universal characteristic of all languages is the presence of syllables
formed by a consonant followed by a vowel. Consequently, speakers might add vowels or delete
extra consonants in complex syllables of a second language. Van Coetsem (1988) makes
reference to the existence of other factors in second language phonology when he discusses
strategies used by speakers, besides imposition, to adapt or change foreign sounds. He calls these
strategies “internally induced changes” and claims they are expressions of universals and not of
language transfer, which is the case of “imposition”. He states that such strategies are used
5
regularly in second language acquisition, such as devoicing of final consonants and
generalization of a rule.
Current research focuses on these new factors that might influence interlanguage as well
as transfer. Major (2001) proposes that second language phonology is affected by three
processes: transfer of the native language phonology, acquisition of target phonology (nativelike
pronunciation), and universals. The interaction of these components in all the different levels of
acquisition is still unclear, but Major (2001) points out that transfer is more likely to occur in
similar phenomena and at the initial stages of the acquisition process, since the speaker will
likely use his native phonology as a building block to acquiring the foreign phonology.
Situations of dissimilar phenomena, on the other hand, are dominated by universals, where
developmental processes and substitutions are likely to occur. Major (2001) proposes a
descriptive model of the interaction of these three components that he calls the Ontogeny
Phylogeny Model, where at the initial stages of acquisition, transfer exerts the largest influence,
and as acquisition of the target phonology increases, the effect of transfer decreases while the
role of universals first increases and then decreases. Current research continues to focus on
transfer and cross-linguistics influence, but now with attention to how it interacts with other
factors.
2.2. Loanword phonology
Transfer is also a topic of much interest in language contact research. One of the areas of
study of language contact is loan phonology. Loan words are lexical items borrowed from a
foreign language into a recipient language. The literature pertaining to loan words and language
6
contact describes loan phonology as the result of phonotactics and strategies used by speakers to
adapt the foreign phonological system to their native language. In language contact situations,
lexical items are by far the main target of borrowings. These loan words might have sounds that
are not present in the phonology of the recipient language, thus they will likely undergo
adaptations or substitutions to cause the lexical item to sound more native and less foreign
(Major, 2001). Odlin (1989) supports the view that borrowing relates largely to lexical items,
especially words in the fields of technology, government, education, and commerce. He says that
even when a language engages in massive lexical borrowing, cross-linguistic influence might
introduce changes to the syntax of the recipient language, but will probably not affect the
phonetics and phonology of the recipient language. Major (2001) argues that this happens
because native language transfer dominates loan phonology. The speaker will try to fit the
borrowed word into his native phonological system. For example, Japanese speakers might
borrow English words that have complex syllables but they will insert vowels as needed to break
up illegal consonant clusters and to conform to the preference Japanese phonology gives to CV
syllables.
Data of borrowed words shows that these lexical items do not always conform to the
phonology of the recipient language. Major (2001) explains that sometimes the speaker will
maintain some elements of the foreign phonology intact or partially adapted and eventually these
elements might be incorporated into the phonology of the recipient language. One example of
such case is the existence of the phoneme [ʒ] in word initial position in English, such as in the
name Gigi. This phoneme entered the English phonology through language contact and
borrowings of foreign words and is now allowed as a legal word-initial onset, although
occurrences of such phoneme are restricted to loan words. In addition the influence of native
7
language transfer and foreign language acquisition, universals may also be present in loan
phonology, but to a lesser extent (Major 2001). They determine to what extent universal plays a
role in loan word adaptation when the adaptation made to the word cannot be explained by the
phonology of the speaker’s native language or the phonology of the foreign language.
In a discussion of transfer in language contact, Van Coetsem (1988) presents a slightly
different perspective on loan phonology. He indicates that not all lexical borrowing implies
phonological borrowing. When a foreign word is borrowed but undergoes adaptation to fit the
recipient language phonology, then lexical loan has occurred. Phonological loan, on the other
hand, occurs when the speaker consciously imitates the foreign pronunciation of the borrowed
lexical item. He adds that speakers will try to preserve their phonology and articulatory habits
because these are the more stable domains of a language, thus engaging more often in lexical
borrowing that in phonological borrowing. From the perspective of van Coetsem (1988), loan
phonology is not about the changes that the native/recipient language will cause on the
pronunciation of the loan word. Rather loan phonology is formed from the influence that the
foreign sounds might have upon the recipient language.
2.3. Comparing the two phonological systems
The description of loan phonology offered by Major (2001), in which the direction of
transfer is from the recipient language towards the target language, is included in van Coetsem's
(1988) description of second language phonology. Most literature on transfer acknowledges the
influence that transfer has on both second language phonology and loan phonology, but without
a clear differentiation of how transfer affects each phonological system. Van Coetsem (1988)
8
argues that there are two different transfer types. One he defines as imposition or source
language agentivity and the other as borrowing or recipient language agentivity. Imposition
occurs in second language phonology, while borrowing occurs in loan phonology.
The main differences are in the direction of transfer, the elements that be transferred, and
the structural changes that result. In second language phonology the more stable domains of the
recipient language will be transferred to the target language, in other words, the phonology. The
speaker will impose his phonology and native articulatory habits upon the target language, thus
causing changes in the target language (e.g. errors in pronunciation) that are due to transfer as
well as the emergence of universals. Eventually the speaker will acquire the phonology of the
target language as learning takes place but traces of his native phonology will probably remain.
In loan phonology the less stable domains of the target language will be transferred into
the recipient language, in other words, lexical items. In the process of borrowing words from a
foreign language, the speaker will try to imitate the foreign pronunciation but adaptation will
eventually take place since the tendency to preserve his native phonology is stronger. In adapting
the loan phonology, constraints of perception or production of the foreign sound will filter out
the borrowed phonological structures leading to partial or total adaption.
While van Coetsem (1988) states that different factors are involved in the production of
second language phonology (language transfer and universals) versus that of loan phonology
(perception and production of sounds), Major (2001) argues that the same factors that cause the
emergence of second language phonology also cause loan phonology, but at different rates. He
states that transfer dominates loan phonology and target phonology or universals have a marginal
influence. In second language phonology, while language transfer will initially exert the largest
9
influence, its importance will diminish with the gradual acquisition of the target phonology and
the emergence of universals.
Even though the literature on language contact and second language acquisition
acknowledge to some extent that there is a difference between loan phonology and second
language phonology, it fails to describe how the lexicon that is borrowed differs from the
utterances produced during second language acquisition. This study will thus offer a description
of the two phonological systems by looking at how word-final consonants are adapted in English
loanwords used in Brazil and in the speech of Brazilian speakers of English. Brazilian
Portuguese does not allow word final codas with the exception of [s], [r], [n] and [l]. And even
then, [n] might not be pronounced if the segment preceding it is a vowel and it assimilates the
nasality of the consonant. The coda [l] will likely be velarized in word-final position resembling
the glide [w]. English, on the other hand, allows words to end in a variety of consonants. This
clear contrast between the two languages in relation to word final codas will likely cause
Brazilian speakers to make adaptations to English words.
3. Description of experiment
3.1. Subjects
In order to compare the phonotactics – strategies – of second language phonology to
those of loanword phonology, four speakers of Brazilian Portuguese were recorded during a
reading task. Two speakers live in Brazil and do not speak English, although they have had
contact with English through some formal education of English as a second language. They will
10
be referred to as the BR speakers from here onward. The other two speakers have been living on
Long Island, New York, for approximately five years and speak English proficiently. They will
be referred to as the US speakers. To avoid pronunciation differences due to a gap in generation,
all four speakers are closely related in age, ranging from 24 to 29 years old. The speakers are
from the same region in Brazil, Sao Paulo, and so have the same accent and pronunciation style
of Brazilian Portuguese.
A fifth American speaker within the age group of the Brazilian speakers and who lives on
Long Island, New York, also took the reading task. Her pronunciation was representative of the
target English pronunciation.
3.2. Method
A list of 184 English loanwords that are used daily in Brazil was presented to the
participants. The BR speakers had to read them as they are pronounced in Brazil, thus the
Portuguese pronunciation of these loanwords. The US speakers had to read them in English. The
fifth speaker, the American, also read them in English. The participants were recorded during the
reading task and later the recordings were transcribed to the International Phonetic Alphabet.
The pronunciation errors the US and BR speakers made in words that ended in a
consonant were noted and recorded for analysis. For the sake of focus, all other pronunciation
errors, such as vowel reduction, adaptation of word initial and word medial consonants, or error
with stress placement, were ignored during data analysis even though these errors can shed some
light into the differences and similarities between loanword phonology and second language
phonology.
11
3.3 Limitations
The method used for the collection of data had several limitations. The fact that the
participants had to read a list of words might have allowed influence of spelling on the
pronunciation. Such possibility will be discussed in section 4.5 and 4.6. Another limitation of
giving a word list is that it elicits a more formal pronunciation style, where native phonology
transfer is less likely to influence mispronunciation and, thus, second language learners are more
likely to achieve target pronunciation (Major 2001).
The level of English acquisition of the speakers should also be taken into consideration.
Different levels of English knowledge might raise different adaptations than the ones done by the
speakers that live in the United States. It would be interesting to compare the speech of several
speakers, from beginners to advanced students of English, in order to see if the phonology of
those speakers is different/ similar from loanword phonology at all levels or if the
differences/similarities are accentuated as the speaker acquires more English skills.
The number of people that participated in the study is too small to be used as a
representative of the population of Brazilian speakers. The influence of gender in pronunciation
styles, for example, can not be clearly analyzed because some of the curiosities seen in the
following section 4 could be due to gender but could also be due to the individual’s phonology.
Nevertheless, the following data analysis will provide a small sample from which to raise
generalizations that can be further explored in future researches.
12
4. Data analysis
Of the total 184 words given to the participants, 156 had word-final coda. The BR
speakers adapted the majority of final codas while the US speakers pronounced the majority of
the words faithfully. The data elicited a variety of strategies in adapting problematic words
ending in consonant and the speakers utilized more than one strategy throughout the reading
task. In order to more easily interpret the data, first I grouped the words that ended in the same
codas and then I grouped the codas that underwent the same adaptations. The following sections
describe the adaptations implemented in each grouping, while appendixes A through G list the
grouped data discussed bellow.
4.1. Words ending in [f], [v], [g], [k], [p], or [b]
Appendix A shows a table with all 40 words that ended in [f], [v], [g], [k], [p], or [b].
Brazilian speakers:
If we first look at the reading task performed by the BR speakers, we see that the main
strategy used to adapt the above cited codas was vowel epenthesis, since approximately 1/3 of
the words with the above cited final coda underwent epenthesis. Vowel epenthesis occurs when a
language inserts a vowel where there was none. Vowel epenthesis is a common strategy used in
Brazilian Portuguese to deal with illegal codas because when a vowel is inserted after the coda,
the consonant becomes the onset of a new syllable. For example, data (5) of appendix A shows
that the English pen drive [pɛndraɪv] was pronounced [pɛndraɪvi] by the BR speakers. The high
13
front vowel [i] is usually the epenthesized vowel in Brazilian Portuguese (Perini 2002). The
words in Appendix A that underwent epenthesis are highlighted in yellow or green.
The female and male BR speakers epenthesized 32 out of the 40 words that ended in the
above cited codas. The female speaker pronounced five words faithfully, without any
adaptations, while the male was not faithful in any of the examples.
A large number of the epenthetic vowels underwent vowel devoicing. The words
highlighted in green in Appendix A show epenthetic vowels that were devoiced. Some words
from appendix B also underwent vowel devoicing, but these will be discussed in section 4.2.
Devoiced vowels occur when air (like whispering) comes out of the mouth without the vibration
of the vocal cords and the articulators move to the position of a vowel, thus affecting the sound
of surrounding phonemes. Vowel devoicing is a common strategy in Brazilian Portuguese when
the final vowel is unstressed (Perini 2002).
To the unaware listener, voiceless vowels can be mistaken for aspiration or lengthening
of final consonant. I used a computer program called Praat that creates a spectrogram of the
recorded word. The spectrogram shows the sound waves of speech. In order to identify voiceless
vowels in the spectrogram, I looked for irregular striations (vocal cord pulses) following the final
consonant that resembled formants (a column that’s formed of dark horizontal bands). However,
my recordings were not good and exterior noise interfered in producing clear spectrograms.
Nevertheless, the amount of definite occurrences of voiceless vowels leads the supporting
evidence that vowel devoicing was employed by the participants as a strategy to fix final codas.
Curiously, vowel devoicing was used almost exclusively with epenthetic vowels. There
were only two cases of devoicing of an original vowel. In appendix A those are data (41) and
14
(42). Another interesting fact of vowel devoicing in this experiment is that it occurred mainly in
the speech of the male BR speaker and was predominant in words ending in [k], [p], and [ʧ].
In the few cases where epenthesis was not employed, other strategies were used to fix the
illegal coda. Some words underwent aspiration, even though it is not a phonotactics of
Portuguese phonology. This happens when a puff of air is released after the articulation of a
voiceless stop, such as in data (19), where the English word 'picnic' [pɪknɪk] was pronounced
[piknikʰ] by the female BR speaker. There were two cases of aspiration in the BR female’s
speech (data (19) and (26) and six in the BR male’s speech – data (15), (21), (25), (38), (40), and
(41).
The male BR speaker also applied consonant lengthening in three occasions, data (1),
(3), and (6). This strategy produces a longer final consonant, such as in data (3), where the
English surf [sɝf] becomes [sɝff] in the speech of the male BR speaker. This is not a common
strategy of Brazilian Portuguese phonology.
Note that both aspiration and consonant lengthening could be residues of a devoiced
vowel. However, Praat’s spectrogram of these words shows no signs of a voiceless vowel.
Consonant devoicing was employed once by the female BR speaker – data (6) – and
twice by the male BR speaker – data (6) and (40). In these cases a final voiced consonant looses
its voicing contrast, thus in data (6) the English making of [meɪkɪŋʌv] was pronounced [meɪkĩɔf]
by the female BR speaker. Consonant devoicing is rare in Brazilian Portuguese, but will occur in
very fast speech when the antepenultimate syllable is stressed and the lat vowel is devoiced. For
example, the word rápido [hapidu] ‘rapid’ will be pronounced [hapitu], with the bold-face here
representing the devoiced sounds (Perini 2002).
15
US speakers:
Now, let’s change focus from the BR speakers to the US speakers. As mentioned above,
epenthesis was the main strategy used by BR speakers. It was not present at all in any of the
utterances produced by US speakers. As seen in appendix A, instead of epenthesizing, the US
speakers opted for aspiration of final coda [k], lengthening of [f], or consonant devoicing of
[v] and [g] codas. The female US speaker did most of the adaptations present in this data. In her
speech there were eleven cases of aspiration, while the male US speaker had one aspiration in
data (24). The female US speaker devoiced four final consonants – data (4), (6), (9), and (10) –
and the male US speaker utilized it once in data (8). Finally, the female US speaker had two
lengthening in data (2) and (3) and the male US speaker had one in data (2).
4.2. Words ending in [t] and [d]
Appendix B provides the table with all the [t] and [d] final codas, the coronal plosives.
BR speakers:
Looking first at the BR speakers, it seems that affrication was almost exclusively the
only strategy BR speakers used to adapt word-final coronal plosives. The words that were
affricated are highlighted in yellow in appendix B.
Affrication consists of adding a fricative to the plosive consonant, such as in data (76) of
BR speakers the English flat [flæt] becoming [flɛʧ]. Of the 31 words present in the data that
ended in coronal plosives, the females BR speaker affricated 26 of them and the male BR
speaker 29 of them. Words that were not affricated were pronounced faithfully with the
exception of data (97), since the female speaker deleted the final consonant.
16
Affrication of [t] and [d] is a common allophonic process in most dialects of Brazil when
the alveolar stop is followed by a high front vowe [i] (Bettoni-Techio 2006). The consonant will
be pronounced normally in all other contexts. The process of affrication of alveolar stops
following a high front vowel is common in many languages of the world. Perini (2002) says that
Brazilian Portuguese can delete a devoiced vowel altogether in rapid speech. In this case, an
affricated [t] or [d] might surface word-finally. The verb pode [pɔʤi] ‘(you) can’ in fast speech
becomes [pɔʤ], and the verb conte [kõʧi] ‘(you) tell (imperative)’ becomes [kõʧ] in fast speech.
In the speech of the BR speakers interviewed for this study, a few of the affricated codas
also underwent vowel epenthesis. The male BR speaker applied vowel devoicing to all of his
epenthetic vowels, and the female BR speaker voiced all but one vowel. It is possible that there
are more voiceless vowels in this data than I was able to identify, due to the difficulties in
identifying voiceless vowel discussed in section 4.1. Since in the reading task of BR speakers we
find affrication with voiced vowels, affrication with voiceless vowels, and affrication alone, this
can be evidence that epenthesis was actually the main strategy used for [t] and [d] codas and that
the epenthetic vowel drove affrication. The vowel is then devoiced and eventually disappears,
which is corroborated by the existence of this process in the native phonology of Portuguese.
US speakers:
Bettoni-Techio et al (2002) found that in second language phonology affrication occurred
more often by itself than followed by an epenthetic vowel, which is reinforced by the results of
the BR speakers. The US speakers also support Bettoni-Techio’s findings since they didn’t
epenthesize at all. The US speakers only adapted a few of the words. The female did most of the
adaptations, with nine instances total. The male only adapted one word. Affrication was one of
17
the strategies used by the US speakers but not the only one. The female affricated five words –
data (70), (74), (91), (92), and (97). The male affricated one, data (70).
As for the other words mispronounced by the female, three final consonants underwent
aspiration – data (82), (84), and (87) – and two codas underwent consonant devoicing – data
(94) and (97). Note that aspiration only occurred with [t] codas and consonant devoicing with
[d].
4.3. Words ending in nasal consonants
Appendix C contains the data for this section.
BR speakers:
The BR speakers employed only two strategies to deal with nasals. One of them was
vowel epenthesis, which was also applied to other final codas as discussed in the previous two
sections. The other strategy, the prevalent one, was vowel nasalization. The words that were
nasalized are highlighted in yellow in the appendix. This is a common strategy in Brazilian
Portuguese (Cardoso 2005). Nasalization occurs when the vowel preceding a nasal is realized
through the nasal cavity instead of the oral one. In the case of Brazilian Portuguese, nasalization
of the vowel means that the final consonant is not produced (perini 2002).
Of the 26 words ending in a nasal consonant, the BR speakers nasalized two thirds of
them. Both the female and the male produced one word faithfully. The female had one case, data
(100), where she dropped the nasal but did not nasalize the preceding vowel.
18
The other one third of the words that were not nasalized, data (119) through (123),
underwent vowel epenthesis, as mentioned previously. In these cases, the nasal was preceded by
a diphthong [aɪ], [oʊ] or [eɪ].
US speakers:
The US speakers did not epenthesize, as has been the pattern so far. They did apply
nasalization to a few of the words. In this context the male US speaker adapted more words than
the female speaker did. He adapted four words – data (98), (99), (101) and (103) – while she
only had two cases of nasalization – data (98) and (103).
4.4. Words ending in [l]
Appendix D shows all the words ending in [l].
BR speakers:
Epenthesis was not the preferred strategy for dealing with laterals word finally. The data
provided by the BR speakers show that words ending in [l] tend to undergo further velarization,
when the consonant is produced with the tongue closer to the velum of the mouth. Velarized
words are highlighted in yellow in the appendix. The English final [l] is already velarized into
the dark lateral [ɫ]. When produced by the BR speakers, this English allophone became even
more velarized, approximating to the glide [w]. (Cardoso 2005). Data (128) shows the English
Google [gugǝɫ] being pronounced [gugoʊ] by the BR speakers. This process is common in the
phonology of Brazilians (Perini 2002). Words ending in [l] are always velarized, such as natal
[nataw] ‘Christmas.’
19
Every word ending in [l], with the exception of data (135), was velarized by the BR
speakers. In the casa of data (135), the word style [staɪɫ] was epenthesized into [staɪle] or
[istaɪli]. The [l] had been preceded by the diphthong [aɪ]. This was probably the reason for
epenthesis; just as the same diphthong drove epenthesis with nasal consonants (see section 4.3).
US speakers:
The US speakers also velarized some of the words. The female velarized a quarter of
final codas – data (124), (125), (130), and (135) – and the male adapted only two codas – data
(130) and (132). Neither of them epenthesized the lateral that followed a diphthong. On the
contrary, the female US speakers preferred to velarize (135) style rather than epenthesize.
4.5. Words ending in [z]
Appendix E provides the words that end in coda [z].
BR speakers:
Even though vowel epenthesis is also present in this data, as it has been in all the other
sections, it was not the primary strategy used. The primary strategy used was consonant
devoicing. The words that underwent devoicing are highlighted in yellow in the appendix.
Section 4.1 describes the process of consonant devoicing in more details. Appendix E shows that
four of the six words ending in [z] were devoiced to [s] by both BR speakers – see data (136)
through (139).
This is probably due to the influence of spelling during the reading tasks, for most of the
words listed in appendix E are spelled with an <s> in the end. It seems that this reasoning would
20
not explain jazz in data (136), but actually it does, because words spelled with final <z> in
Portuguese are pronounced with an [s] (Perini 2002).
Consonant devoicing is not a common phonotactics of Portuguese, except for the case of
following a devoiced final vowel (see section 4.1). It was rarely applied by the BR speakers in
the contexts described so far. There were only three cases all together, two in appendix A – data
(6) and (40) – and one in appendix B – data (92). It was more prominent among the US speakers,
who devoiced seven final consonants, five in appendix A – data (4), (6), (8), (9), and (10) – and
two in appendix B – data (94) and (97).
There were two words in appendix E that were not devoiced, data (140) and (141). In
those cases, the female BR speaker maintained a voiced [z] but epenthesized a high front vowel,
while the male BR speaker pronounced both words faithfully. This pronunciation error is likely
the influence of spelling as well. The pronunciation error made by the female BR speaker could
be due to spelling, since <s> between two vowels is pronounced as a [z] in Portuguese (Perini
2002). For example, the word dose ‘shot’ is pronounced [dɔzi].
US speakers:
The US speakers adapted about half of the words through consonant devoicing of the
final [z]. This can not be related to spelling because the contexts for devoicing are varied, even
including data (140) which was voiced and epenthesized by the female BR speakers. The other
words were pronounced faithfully. Vowel epenthesis was not an option for these speakers.
21
4.6. Words ending in [s] and [r]
Appendix F contains the words that ended in [s] or [r].
BR speakers:
These words were grouped together because unlike all codas so far, the majority of these
words did not undergo adaptation. [s] and [r] were usually pronounced faithfully. This was
expected since the only consonants that surface as word-final codas in Brazil as [s] and [r] (perini
2002), even though they may be deleted in casual speech in Brazil (Cardoso 1999).
There were five [s] final words, out of the twelve, that underwent adaption by the female
BR female speaker. She inserted a vowel after the [s] in four words – data (149) thought (152).
In another case she applied consonant voicing, pronouncing the [s] as [z], and then inserted a
vowel – data (153). Note that in section 4.5 the female BR speaker also inserted vowels after [z]
when the coda was not devoiced. The male BR speaker only epenthesized one word ending in
[s] – data (149) – but he also voiced the [s] in data (153) and inserted a vowel. These adaptations
could be the influence of spelling, as discussed in 3.5. Final ‘-ce’ in Brazilian Portuguese is
pronounced [si]. Compare for example the word doce [dosi] ‘sweet’ with the word presented in
section 4.5 dose [dɔzi] ‘shot’. With this comparison in mind, it is easy to see how some of the
data in appendix E and F might have been mispronounced because of spelling.
US speakers:
All words were pronounced faithfully with the exception of data (153), which was voiced
by the male speaker. It could be that he was influence by the Brazilian pronunciation of this
word, which also voices the final consonant, but he might have refrained from epenthesizing due
to acquisition of word-final coda pronunciation.
22
4.7. Other contexts – words ending in [ʧ], [ɵ] and [ʃ]
In appendix G, words ending in [ʧ], [ɵ] and [ʃ] were grouped together not because of
similarities in the strategies used to adapt these coda, like the other groupings. They were
grouped together because there were not enough examples of each show the speaker’s tendency
when dealing with these codas. See appendix G for the examples.
US speakers:
The US speakers pronounced all of them faithfully.
BR speakers:
The BR speakers had only one mispronunciation. The female speaker pronounced the
affricate [ʧ] as a fricative [ʃ] and inserted a vowel after.
4.8. Generalizations
The contrastive analysis between the strategies the US speakers used and those the BR
speakers used, as well as the frequency of occurrences, shows evidence of the difference between
loanword phonology and second language phonology in the speech of Brazilian Portuguese. The
two BR speakers adapted the majority of the words, while the US speakers were able to
pronounce most of them faithfully. This result is expected because the BR speakers have not
acquired the phonology of English, but the US speakers have clearly acquired some of it and
have consequently learned to pronounce words more faithfully.
The large number of adaptations the BR speakers made to the English word list confirms
the well known fact that Brazilian Portuguese prefers not to have word-final codas (Perini 2002,
23
Major 2005, Bettoni-Techio 2006). The only codas that were pronounced faithfully, without
adaptation, were [s] and [r] because they may occur in coda position in Brazilian phonology.
Even the voiced alveolar fricative [z] was pronounced as its voiceless counterpart, [s], with
exception of where spelling would interfere.
The two BR speakers had a clear preference for vowel epenthesis as a strategy to deal
with illegal final codas, especially if we take into consideration the possibility of epenthesis and
later vowel deletion in the affrication process of coronal plosives.
The US speakers avoided epenthesis completely. Where epenthesis would occur, they
used aspiration for voiceless coronal and velar stops or consonant devoicing for voiced stops and
fricatives. These strategies also appeared in the speech of BR speakers, but in a small frequency
in comparison to vowel epenthesis.
The contexts where US and BR speakers coincide in the choice of strategy used was
when the final consonant was a nasal or a lateral. Nasals were realized mostly through nasalized
final vowels and laterals were velarized into glides. Here the difference is that US speakers only
adapted a few of the words, while the BR speakers adapted the large majority of them.
One curiosity of the data that has not been mentioned yet was the possible influence of
gender in the choice of strategy as well as frequency of adaptation. Among the BR speakers,
gender seemed to influence the choice of strategy, since the male BR speaker was responsible for
most of the vowel devoicing that occurred. Among the US speakers, gender seemed to influence
the frequency of adaption, since the female speaker adapted words more often than the male did.
This is an interesting aspect to explore in future research.
24
5. Discussion
This study shows that one of the differences between loanword phonology and second
language phonology is in the nature of the errors made by the speakers. In the loanword
phonology of the BR speakers, most errors were due to transfer of the speaker’s native
phonology. Epenthesis, affrication, nasalization and velarization were all expected strategies in
dealing with illegal codas because they are present in the phonology of Portuguese. As for the
other words that underwent different adaptations, a small fraction of them were pronounced
faithfully, which could be due to previous contact with the English language and some
acquisition of the phonology. Another small fraction of the words undergo aspiration,
lengthening, and consonant devoicing, which could not be caused by transfer, since these
strategies are not present in Brazilian Portuguese. These strategies reflect the influence of
universals in the grammar of these speakers.
In the second language phonology of the US speakers, faithful pronunciation was
achieved for the majority of the words, which is due to acquisition of the English phonology.
Native strategies were minimized in the second language phonology of the US speakers,
sometimes to the point of non-existence, such as vowel epenthesis. The native strategies that did
emerge in second language phonology occurred in a much smaller frequency than it did in the
loanword phonology. Other strategies that were neither native nor from the foreign language
surfaced in the results of the US speakers, showing that they had incomplete acquisition of the
foreign phonology, such as aspiration, lengthening, and consonant devoicing. These strategies
are not common in the native phonology of Brazilian Portuguese. They are not present in the
25
phonology of English in relation to word final codas. They are due to language Universals
related to markedness constraints or developmental processes natural to all languages.
The findings of this study fit with Major’s (2001) definition of second language
phonology and loanword phonology. He claims they are ends of a spectrum where the native
phonology, the foreign phonology and universals interact to produce the phonotactics we see in
this experiment. The BR speakers are one end of the spectrum with native phonology surpassing
by far the influence of the foreign phonology and universals. Major says that once the person
starts to learn the second language, acquisition of the foreign phonology will increase as
influence of native phonology decreases but universals increase. At the end of the spectrum,
universals decrease again and then the foreign phonology will be responsible for most of the
pronunciation characteristics.
There is a practical advantage to understanding the specific differences between the two
linguistic systems. A description of the phonology of borrowed English words in Brazil can help
educators understand the mentality that Brazilian students bring to the classroom as they take the
first steps towards learning a second language. The difficulties and tendencies are better defined
and easier for instructors to call attention to or practice during pronunciation drills. By
understanding second language phonology, educators will know what difficulties Brazilian
students face once they’ve acquired some of the English phonology but still have a strong accent
when speaking English. In general, by exploring the explicit differences between the two
systems, linguistics might gain a little more understanding in how a second grammar develops in
the mind of an adult who is learning a second language.
26
6. Conclusion
The phonological adaptations done to borrowed lexical items and to a second language in
the acquisition process are the result of factors acting upon the phonology of the speaker. The
literature on language contact lacks contrastive comparison between the differences in the
adaptation of loanwords and that of second language learning. This study focused in comparing
the two phonological systems in order to understand if they are the same or if they are different.
The finds show that loanword phonology is different from second language acquisition in several
ways. The factors that drive adaptations are different. In loanword phonology transfer of the
native language drives adaptation. In second language phonology universals and developmental
processes are responsible for mispronunciation. Consequently, the phonotactics and frequency of
adaptation in the speech of monolingual Brazilians is different from that of the speech of
Brazilian English speakers.
Acknowledgements
A special thanks to Professor Lori Repetti from the Linguistics department for her
guidance and input, to Professor José Elías-Ullo from the department of Hispanic Languages and
Literature for his assistance in analyzing the data with the software Praat, and to Professor
Christina Bethin from the Linguistics department for her suggestion in the editing of the final
draft.
27
References
Bettoni-Techio, M. et al (2006). Palatalization in Brazilian Portuguese/English
interphonology. Revista Virtual de Estudos da Linguagem – ReVel. V. 4, n. 7
Cardoso, W. (1999). A quantitative analysis of word-final /r/-deletion in Brazilian Portuguese
[Electronic version]. Lnguistica Atlantica 21, 13-52.
Cardoso, W. (2005). The Variable Acquisition of English Word-final Stops by Brazilian
Portuguese Speakers. Proceedings of the 7th Generative Approaches to Second Language
Acquisition Conference (GASLA 2004), ed. Laurent Dekydtspotter et al., 38-49.
Somerville, MA: Cascadilla Proceedings Project.
Major, R. C. (2001). Foreign accent: The ontogeny and phylogeny of second language
phonology. Mahwah, NJ: Lawrence Erlbaum Associates
Nevins, A. and Braun, D. (2009). The Role of Underlying Representations in L2 Brazilian
English. In Calabrese, A., and Wetzels, W. L. (Eds.), Current issues in linguistics Theory
V. 307: Loan phonology (181-191). Philadelphia, PA: John Benjamins Publishing Co.
Odlin, T. (1989). Language transfer: Cross-linguistics influence in language learning.
Cambridge: Cambridge University Press.
Pap, L. (1949). Portuguese-American Speech: An outline of Speech Conditions Among
Portuguese Immigrants in New England and Elsewhere in the United States. New York:
King's Crown Press.
Perini, M. A. (2002). Modern Portuguese: A reference grammar. New Haven: Yale University
Van Coetsem, F. (1988). Loan phonology and the two transfer types in language contact.
Dordrecht, Holland: Foris Publications.
28
Appendix A
Words ending in [f], [v], [g], [k], [p], or [b]
29
Appendix B
Words ending in [t] and [d]
Appendix C
Words ending in nasal consonants
30
Appendix D
Words ending in [l]
Appendix E
Words ending in [z]
Appendix F
Words ending in [s] and [r]
31
Appendix G
Other contexts – words ending in [ʧ], [ɵ] and [ʃ]
Download