Foreign accent in adult simultaneous bilinguals
Tanja Kupisch, Dagmar Barton, Ewgenia Klaschik, Tatjana Lein, Ilse Stangen &
Joost van de Weijer
Forthcoming in Heritage Language Journal
Abstract: The study reported in this paper examines foreign accent (FA) in adult simultaneous
bilinguals (2L1ers). Specifically, we investigate how accent is affected if a first language is
acquired as a minority (heritage) language as compared to a majority (dominant) language. We
compare the perceived FA in both languages of 38 adult 2L1ers (German-French and German-
Italian) to that of monolingual native speakers (L1ers) and late second language learners (L2ers).
Naturalistic speech samples are judged by 84 native speakers of the respective languages. Results
indicate that the majority language is always spoken without a FA, while results for the heritage
language fall between those of L1 and L2 speakers. For the heritage language, we further show
that a native accent correlates with length of residence in the heritage country during childhood
but not during adulthood. Furthermore, raters have comparatively more difficulties when judging
the accent of a heritage speaker. The results of this study add to our current understanding of
what factors shape the phonology of a heritage language system in adulthood.
2
1. Introduction
It is commonly assumed that a foreign accent (FA) in a second language (L2) cannot be
overcome easily after puberty (Lenneberg, 1967, p. 176). FA research shows that the earlier in
life a language is acquired the less accented it will be (e.g. Thompson, 1991; Flege, Munro, &
Kay, 1995; Flege, Yeni-Komshian, & Liu, 1999; Abrahamsson & Hyltenstam, 2009, as well as
the overviews in Moyer, 1999: 82-84; Piske, MacKay, & Flege, 2001; Jesney, 2004).
Phonological acquisition may be constrained as early as age 6, earlier than syntax or morphology
(Long, 1990, p. 280). There is even evidence that the accent of those learning English at age 3 is
significantly different from the accent of English speakers with exposure from birth (e.g. Flege,
Munro, & MacKay, 1995; Flege, Yeni-Komshian, & Liu, 1999). Various explanations for
problems with phonetic and phonological properties in L2 acquisition have been offered, the
classical one being the loss of neural plasticity with aging (Lenneberg, 1967; Scovel, 1969; for a
review see Birdsong, 1999). Another explanation is that late learners generally receive less L2
input and/or more foreign-accented L2 input than early learners (e.g., Oyama, 1976; Snow &
Hoefnagel-Höhle, 1977). Amount of continued L1 use is another important factor in the degree of
foreign accent in an L2 (Flege, Frieda, & Nozawa, 1997; Yeni-Komshian, Flege, & Liu 2000;
Piske et al., 2001; Flege, MacKay, & Piske, 2002). Since each of these explanations is directly or
indirectly related to age, one may expect early bilinguals to have a great advantage over L2
learners (henceforth, L2ers).
According to Au, Knightly, Jun and Oh (2002) even passive exposure to a language in
childhood can lead to lasting advantages in native-like production of sounds. Their L1 English
subjects “overheard” informal Spanish ~9 hours per week between birth and age 6, and ~4 hours
per week between 6 and 12 years before starting to learn Spanish at age 14. Around the age of 20,
these overhearers were found to have more native-like VOTs in their L2 than L2ers who had
3
learnt Spanish after age 14 without any prior exposure to the language. Moreover, they produced
Spanish /b, d, g / authentically, i.e. as lenited consonants more often. This study suggests that
early bilinguals have a great advantage over late learners when acquiring the sound system of a
second language, even if they actively used this language little or not at all during childhood.
Although FA has been an extensively researched topic for the past two decades, there have
been few systematic studies of adult bilinguals with simultaneous exposure to two languages
from birth (henceforth, 2L1ers), investigating both languages. Similarly, the growing field of
research on heritage bilinguals has focused mostly on morpho-syntax, while —at least until
recently— phonology has been an understudied area (see Montrul, 2010; Benmamoun, Montrul,
& Polinsky, 2013 for overviews). In this paper, we investigate the perceived FA of 2L1ers who
acquired German and French or German and Italian from birth, comparing them to monolingual
speakers (henceforth, L1ers) and late L2ers in each of their languages. Moreover, we compare the
perceived accent in the 2L1ers’ two languages to see whether native-like pronunciation in one of
them implies a FA in the other.
But before we proceed, a few notes on terminology are in place: FA studies often use the
terms L2 learner and L2 speaker for the chronologically second language, even if immigration
and Age of Onset (AoO) in the language of the L2 country happened very early in life, e.g. at age
two. By contrast, within the context of research on bilingual children, distinctions are often drawn
between bilinguals and early L2 learners. The age dividing these groups depends on assumptions
about the critical age (e.g. around age 4 according to Meisel 2011), and ultimate proficiency need
not correlate with these distinctions (Grosjean, 1982). Importantly, a person with an AoO at age
two may count as 2L1 learner in one study, but as an L2 learner in another study. When
summarizing relevant research in the following, we will use the respective authors’ terminology.
Our own study is limited to cases of bilingualism from birth (i.e., simultaneous bilingualism) and
4
late L2 learners (AoO and/or Age of Arrival (AoA) at age 11 or later), thus avoiding any
borderline cases.
The bilingual participants in the present study all grew up with two languages at home,
their parents following the one person-one language principle. Despite being very fluent in both
languages, these bilinguals are generally more proficient (stronger or dominant) in the national or
majority language of their childhood environment. Their less proficient (weaker) language
matches the concept of a heritage language according to some definitions. According to some
authors, a language counts as heritage language if it is acquired naturalistically in the home
context and corresponds to a minority language in the national society (e.g., Montrul, 2008, 2012;
Rothman, 2009; see Kupisch 2013 for a discussion of who counts as “heritage speaker”). Others,
by contrast, have used the term heritage speaker exclusively with reference to bilinguals raised in
a home where only one language is spoken (e.g., Polinsky & Kagan, 2007), which does not apply
to our case. So, it is important to keep in mind that, having been raised with two languages at
home, our simultaneous bilinguals have “dual heritage”. Moreover, some of our participants may
be considered bilinguals who acquired their heritage language under “privileged” conditions
because they had had the opportunity to attend schools where the minority language was the
language of instruction, thereby differing from heritage speakers that have predominantly been
discussed in the literature (e.g., Benmamoun et al., 2013). Nevertheless, we believe that our study
adds to the current understanding of heritage languages, in testing under which conditions early
bilinguals can attain monolingual-like1 end state systems in phonology.
1 We are aware of the ongoing discussion of whether true monolingualism still exists, and whether it is justified to
compare monolinguals to bilinguals (e.g., Birdsong & Gertken, 2013). Our study is orthogonal to this discussion, as
we mainly compare bilingual speakers to each other. We use the term “monolingual” to refer to speakers who have
not been exposed to a second language during early childhood, and we assume that monolingual native speakers
5
In what follows, we provide an overview of previous research, concentrating on studies that
could potentially compromise the idea of a sensitive period for accent. Section 3 formulates our
research questions. Section 4 introduces participants and the experimental design; section 5
presents our study. Results are discussed in section 6. We conclude in section 7.
2. Age of onset and foreign accent
Long (1990) proposed that age 6 is critical for phonology: “Learners starting later than age 6
often become communicatively fluent, but also often finish with measureable accents in
phonology […]” (Long, 1990, p. 266). Contrary to this claim, many studies provided evidence
that speakers with an AoO later than 6 years can achieve high, even native-like, levels of
pronunciation and intonation (e.g., Schneiderman & Desmarais, 1988; Ioup, Boustagui, El Tigi,
& Moselle, 1994; Bongaerts, Planken, & Schils, 1995; Bongaerts, van Summeren, Planken, &
Schils, 1997; Birdsong, 2003; Moyer, 1999, 2004; Flege et al., 1995). Thus, even if it is
uncontroversial that earlier learners tend to attain more authentic pronunciation than late learners,
evidence from late learners achieving native-like accents challenges the idea of a critical cut-off
point for the acquisition of a native-like accent. Furthermore, as we outline in the following, (i)
some speakers have a FA in a language (L1 or L2) to which they were exposed before the age of
6, and (ii) under certain conditions, speakers who grew up in a monolingual context may manifest
a FA in their L1 later in life.
have relatively consistent judgments about who sounds “native” and who sounds “foreign”. We use the term “native
(L1) speaker” to refer to someone who has been exposed to a language from birth. Being a native speaker does not
imply sounding “native-like”.
6
2.1 The earlier the better?
There is considerable evidence that early exposure does not guarantee monolingual-like native
performance during adulthood. In Thompson’s (1991) study of 36 Russian immigrants to the U.S.
(mean age 45.3 years), 3 were below age 6 when arriving in the US, and only 2 of these 3 were
deemed native-like in English. Flege, Munro and MacKay (1995) looked at the L2 English
pronunciation of 240 Italian speakers who immigrated to Canada between 2 and 23 years.
According to average ratings of 10 listeners, only 78% of the Italian speakers younger than 4 on
arrival met the criterion for authentic pronunciation (ibid. 1995, p. 3129). 40 learners who
immigrated to Canada between the ages of 2;6 and 9;6 were further differentiated in terms of
self-reported use of their L1 Italian after arriving in Canada (“high” vs. “low” use of L1 Italian
subjects) (Flege et al., 1997). Both groups were found to have “detectable foreign accents” in
English. Accents were milder for earlier arrivals (average age of arrival (AoA): 3;2) than for later
arrivals (Flege et al., 1997, p. 182), and high use L1 Italian subjects had significantly stronger
foreign accents in English than did low use native Italian subjects. These results suggest that
relative frequency of L1 and L2 use has an impact on accent as well.
Only few studies on FA included individuals with an AoA earlier than 3 years. In Flege,
MacKay and Piske’s (2002) study (L1 Italian, L2 English), 36 speakers were between 2 and 13
years on arrival in Canada, and only 11 of them fell within the native speaker range of
accentedness ratings in their L2. Abrahamsson and Hyltenstam (2009) examined FA in
naturalistic speech samples (20-30 sec) of 195 speakers with L1 Spanish L2 Swedish (AoO 1-47)
who considered themselves native-like in Swedish. Most speakers with an AoO before 12 were
perceived as native-like in Swedish. However, none of the 53 speakers with AoO below 6 was
perceived as native by all 10 raters, and there was a visible (though statistically non-significant)
7
difference in the mean scores of native Swedish controls and L2ers with an AoO between 1 and 5
years.
Flege, Yeni-Komshian and Liu (1999) investigated L2 English in 240 Korean speakers who
immigrated to the U.S. between the ages of 1 and 23 years. All but a few early arriving Koreans
received lower ratings than the native English controls. Those who arrived in Canada between
age 1 and 5 were similar but nonetheless deviant from the native controls. The same speakers
were compared in terms of their pronunciation in Korean and English (Yeni-Komshian, Flege, &
Liu, 2000), and there was an almost inverse relationship between the pronunciation in the two
languages. The Koreans arriving in the U.S. between age 1 and 5 had very weak FAs in English
(cf. Flege et al., 1999), but a distinct FA in Korean; Koreans arriving between age 12 and 23 had
clear FAs in English and no foreign accent in Korean. In other words, there was a linear
correlation between age of arrival and FA in both languages.
There are other studies showing that the phonetic and phonological systems of early adult
bilinguals are different from those of monolingual speakers. Mack (1989) investigated adult
English-French early bilinguals in the U.S. in their perception and production of voice onset time
(VOT) for /t/ and /d/ as well as vowel duration and formant frequency for /i/ and /ɪ/ in English.
The phonetic system of these speakers approximated that of monolinguals, although it did not
perfectly resemble it. Unlike the participants in the present study, the “early bilinguals” in
Mack’s (1989) study had an AoO of 4.5 years for their second language (English or French being
the L1), and only one bilingual was exposed to both languages before age 3. Sundara et al. (2006)
investigated adult simultaneous English-French bilinguals in Canada, i.e., an environment where
neither English nor French would count as a minority language. High scores on an accent rating
task (6 out of 1-7 points) in both languages were used as inclusionary criteria for subject
selection. Those who were included (5 out of 10) did not differ in their VOT productions from
8
monolinguals concerning /t/. However, the bilinguals did differ in the VOT production of the /d/
in English (producing it with lead voicing more often than monolinguals), and they used fewer
acoustic cues than monolinguals when differentiating /t/ from /d/. In another study on VOT in
Canada, English-French adult simultaneous bilinguals were shown to differ from monolinguals in
their production of the voiceless stops /p, t, k/ (Fowler, Sramko, Ostry, Rowland, & Hallé, 2008).
In summary, there is evidence that at least some bilinguals exhibit compromise values in their
speech production.
Although there is a lot of research showing that early exposure does not guarantee a native-
like accent later in life, the research specifically dedicated to heritage speakers suggests that these
speakers have advantages over L2 speakers in phonology (see Montrul 2012 for an overview).
For example, Chang et al. (2011) investigated the production of fricatives, vowels and VOT in 28
heritage speakers of Mandarin Chinese in the US, looking at their production in both English and
Chinese. The heritage speakers were better than L2ers in producing the phonetic norms of their
two languages, thus maintaining cross-linguistic contrasts between similar categories in L1 and
L2 which L2ers tend to merge.
In summary, previous research suggests that early exposure helps but does not guarantee a
native-like accent later in life. Furthermore, there is some evidence suggesting (i) that relatively
high use of a minority language during childhood can prevent a native accent in the community
language, (ii) that there is an inverse relationship between nativelikeness of one’s accent in a
minority/heritage language and in the language of the community, and that (iii) heritage speakers
nevertheless have advantages over late L2ers in approaching a monolingual speaker’s perception
of a “native accent”.
The participants in the present study had previously been investigated with respect to
various morpho-syntactic and semantic properties in their two native languages. All patterned
9
with monolinguals in the dominant language of their childhood environment, while in their
minority language they ranged in performance between natives and L2ers (e.g., Bianchi, 2013;
Kupisch, 2012, 2014). In this study, we will investigate whether the same holds true for their
global accent.
2.2 Can one lose a native accent later in life?
With regard to the question of whether early AoO guarantees a native-like accent, it is also
relevant to ask what may happen to a native accent under a drastic change in input conditions
after the L1 has been fully acquired. Such a situation may occur after immigration to a foreign
country, but also in the case of second-generation immigrants who have been exposed to a
minority language at home but cease to use it when starting school in the majority language.
Godson (2004) showed for Western Armenian heritage speakers, bilingual with English,
that their English affected their Western Armenian vowel system. Interrupted (n=10) and
uninterrupted speakers (n=10) of Western Armenian in an English-dominant environment
produced their vowels differently. Not only did the interrupted acquirers' vowels sound more like
English vowels, but the uninterrupted acquirers' vowels were also different from those of a
Western Armenian monolingual baseline speaker. According to Godson (2004), this finding
“supports the view that phonetic modification of a fully acquired language is a process that
continues over a lifetime.” (Godson, 2004, p. 18).
Oh, Jun, Knightly and Au (2003) studied native Korean students in first-year Korean
classes who ceased to use and hear Korean at age 7 or earlier when beginning to use English
intensively (childhood speakers), comparing them to students who had only heard Korean during
childhood without using the language actively (childhood hearers). With respect to VOT in
production and perception, childhood speakers were indistinguishable from monolingual native
10
speakers of Korean and significantly better than L2 learners and childhood hearers of Korean.
But in terms of perceived accent, the childhood speakers were significantly different from natives
despite being more native-like than the L2ers and the childhood hearers. This study, too, shows
that interrupted exposure in childhood in combination with intensive input and use of a second
language can affect accent, even if exposure to this second language only happens after age 6.
Major (1992) studied VOT in 5 L1 English speakers who had moved to Brazil as adults and
had been staying there for between 12 and 35 years. The speakers’ L2, Brazilian Portuguese,
which has little aspiration and a relatively short VOT, had an impact on the speakers’ VOT in
English, which has more aspiration and a relatively long VOT. Moreover, the proficiency level in
L2 Portuguese was positively correlated with the degree of L1 loss with respect to VOT. In
informal style, the degree of loss was more evident than in formal situations, such as teaching.
The results confirm Godson’s (2004) view that modification of a fully acquired language is a
process that continues over the lifespan.
De Leeuw, Schmid and Mennen (2010) investigated perceived FA in two groups of L1
German speakers who had moved to the Netherlands or Anglophone Canada as adults, and had
been living there for 37 years on average. When compared to L1 German speakers residing in
Germany, these “potential attriters” of German had a significantly more noticeable FA. However,
not all individual subjects were judged as having a FA: 20 speakers were judged as native, 23 as
unclear, and 14 as non-native. Hence, L1 attrition is not an inevitable consequence of
immigration. FA could not be reliably predicted on the basis of AoA or on the basis of length of
residence in the L2 country.
Hopp and Schmid (2013) compared perceived FA in potential L1 German attriters and L2
learners of German (L1 English or Dutch). The L1 German speakers had emigrated to
Anglophone Canada or the Netherlands after age 17 and had been living there for more than 15
11
years. There were significant differences between L2ers and potential attriters (as well as
between L2ers and native controls), but no significant difference between potential attriters and
the L1 control group: 72.5% of the potential L1 attriters scored within the control group range,
while only 37.5% of the L2ers did so. Hence, an L1 can be accented after extensive and long-
term exposure to another language, but in terms of nativelikeness it continues to bear more
similarities with an L1 than with an L2.
Summarizing, there is some evidence that phonetic and phonological changes can happen
throughout the lifespan. Nevertheless, there is a tendency for L1 attriters to resemble L1 speakers
more closely than L2 speakers do.
3. Research questions
The research summarized above suggests that a native language is vulnerable to phonetic and
phonological changes later in life, and that factors other than AoO have an influence on the
degree of accentedness. In this study, we will question the role of AoO further by comparing
simultaneous bilinguals (2L1ers) who acquired a particular language as a majority language to
2L1ers who acquired the same language as a minority language, as well as to monolingual L1 and
L2 speakers.
To our knowledge, no previous study has systematically examined global FA in adult
bilinguals who have acquired two languages simultaneously from birth, and no previous study
has compared the two languages of simultaneous bilinguals with respect to FA. Our first three
research questions can be summarized as follows:
(i) Are 2L1ers (foreign) accent-free in their two languages?
12
(ii) If not, does it matter whether a given language of the 2L1 represented the
majority/dominant language of the society (e.g. French in France) or a
minority/heritage language (e.g. French in Germany)
a. during childhood and the school years?
b. later in life, i.e. after finishing school?
(iii) Which sociolinguistic or affective factors are correlated with the presence of a
foreign accent in 2L1ers?
Based on previous research on FA in early bilinguals (e.g., Yeni-Komshian, Flege, & Liu, 2000),
we predict that few bilinguals will be taken for native speakers in both languages, although they
will outperform L2ers even in their minority language (Montrul, 2012). On par with the research
on morpho-syntax in simultaneous bilinguals (e.g., Bianchi, 2013; Kupisch, 2012, 2014; Kupisch,
Akpinar, & Stöhr, 2013), we expect native-like accents in the bilinguals’ majority language,
while accents in their minority language may be subject to individual variation. Given this
individual variation, we suspect that factors beyond the minority/majority language distinction
determine the accent during adulthood. Finally, since research on L1 attrition has shown that it is
possible but rare for monolinguals to develop a foreign accent in their L1 later in life, despite
intensive exposure to an L2, we expect the effect of the environment language to be less decisive
during adulthood than during childhood.
Another question we wish to address concerns confidence ratings in experiments with 2L1
speakers. Although confidence ratings were used in previous studies, they were not used to
explore whether raters show differences in their ability to judge the accent of 2L1ers as compared
to L1 and L2 speakers. This motivates our last question:
(iv) Are raters less confident and do they need more time when judging the accent of a
simultaneous bilingual (as compared to the accent of L1 or L2 speakers)?
13
We assume that it is more difficult to determine the accent of early bilinguals, because they have
acquired their languages in a naturalistic setting and are generally very fluent speakers with good
mastery of the colloquial language (e.g. Montrul 2008), which could make their speech sound
more “authentic” compared to that of L2ers.
4. Foreign accent study
4.1 Participants (speakers)
The data to be judged for foreign accent were gathered from three groups of adult participants: 20
monolinguals (L1ers), 20 second-language speakers (L2ers) and 38 simultaneous bilinguals
(2L1ers). The 2L1ers were investigated in both of their languages. We made the preliminary
assumption that the country of residence during childhood and adolescence is crucial for their
relative proficiency in each language and grouped them accordingly.2 We also assumed that
2L1ers may speak their minority language with a FA, but that their accent in this language is
subject to individual variation, as was found for their performance in morphosyntax (see above).
We therefore used the data from L2ers and monolingual L1ers not only as control data, but also
to normalize the ratings of the bilinguals, presenting them with clear examples of accented speech
(L2ers) and native-like speech (monolinguals L1ers).
The monolingual native speakers spoke only one language (French, German or Italian)
during their childhood at home and had this language as the only language of instruction at
school (the criterion for “monolingualism” proposed by Abrahamsson & Hyltenstam (2009)).
2 This assumption was confirmed by a cloze test as well as by previous studies on their morpho-syntactic abilities
(Bianchi, 2013; Kupisch, 2012, 2014; Kupisch et al., 2013). Most 2L1ers performed better in the majority language
of their childhood environment. Only a few 2L1ers performed equally well in both languages in some tests.
14
Some had a minor but recognizable regional accent, e.g. Franconian, Tuscan, Romanesco, but
this was also true for some of the 2L1ers. The L2ers were also monolingual in the above sense
and had no contact with the L2 until the age of 11 years. They were very advanced speakers of
the L2 according to self-assessment and cloze test results.
The 2L1ers grew up in bi-national families in France, Germany, or Italy as German-French
or German-Italian bilinguals, and they heard and used both languages from birth. According to
self-reports, each parent used his/her native language with them, while the family language
varied, being either the minority or the majority language.3 All 38 2L1ers were active and fluent
speakers of both languages. During adulthood, some 2L1ers had left their childhood environment
to live in the country where their heritage language was spoken (henceforth: heritage country, or
HC). This was the case for all 10 German-French bilinguals who grew up in France, and for one
German-Italian bilingual who grew up in Italy. Comparing these speakers to all others, we can
investigate whether a change in the environment language during adulthood can have an impact
on accent. Table 1 provides an overview of the participants.
[ INSERT TABLE 1 ]
3 Some studies have shown that the home language is crucial for the development and maintenance of the weaker
language (e.g., Gathercole & Thomas, 2009). Since we have no systematic information on language use at home, we
cannot include this factor in our analysis. All we can say is that the parents were L2 speakers of their partner’s native
language and they had to choose one language when communicating to each other. It is therefore likely that the
2L1ers were also exposed to accented speech in at least one of their languages. Again, we have no record of their
parents’ speech. However, it seems that accented speech in first generation of immigrants does not affect the second
generation’s speech in a majority language, but only in a minority language. This suggests that exposure to accented
speech at home can only be one of several possible explanations for non-nativelike pronunciation during adulthood.
15
4.2 Predictors
Since we wanted to find out which sociolinguistic and affective factors correlate with a FA in the
minority/heritage language, we collected more detailed information on the variables outlined
below. We attributed numerical values to each of them in order to later determine whether they
correlate with a FA. The information is based on self-reports.
1) Language preference (LP): The 2L1ers generally preferred to speak the majority
language of their childhood environment, but there were exceptions, mostly among the 2L1ers
who grew up in France and moved to Germany during adulthood. The numerical values we used
were “1” (preference for the majority language of their childhood environment), “2” (no
preference/likes to use both languages), “3” (prefers to speak the minority language of their
childhood environment).
2) Frequency of contact and use during childhood (abbreviated as ChFU “Childhood
Frequency of Use”) was determined on the basis of the type of schooling and language practices
at home. The numerical values used were: “1” if participants reported that the heritage language
was used inconsistently at home and if participants had little or no contact with the heritage
language outside their homes, “2” if parents used their respective languages consistently or if the
participants attended a school in which the heritage language was used as a means of instruction,
“3” if both conditions in “2” applied.
3) Frequency of use during data collection (abbreviated as CuFU “Current Frequency of
Use”): Depending on social relations, work, and place of residence, 2L1ers differed in terms of
how much they used their languages at the time of testing. We assigned “1” if they reported using
their heritage language less than 5 hours per week and if they had no friends or family with
whom they used the language; “2” if they had little but regular (i.e. daily) contact with the
heritage language through work, university, friends or family; “3” if they moved to the heritage
16
country but did not socialize with other speakers of the language spoken there; “4” if they lived in
the heritage country and were fully immersed, using the heritage privately and at work.
4) Type of schooling in the heritage language (in the following TSH): We assigned “1” if
participants had no formal schooling in the heritage language, “2” if they had formal schooling
through weekly afternoon or weekend schools during childhood, or if they studied the heritage
language as a foreign language at school or university, “3” if they attended a regular daily school
during childhood where the heritage language was a/the language of instruction.
5) Absolute age indicates the participants’ age at the time of data collection.
6 and 7 and 8) Length of residence in a country where the heritage language is an official
language (LoRHC “Length of Residence Heritage Country”). Here, we distinguished between
the period before and after age 19;0. For the period after 19;0, we calculated absolute time in
years and relative time in % (dependent on age). For the period between 0-18;11, we
distinguished 2L1ers who travelled to the HC only for vacation (which constituted the majority of
the speakers) from 2L1ers who had lived there for a longer period. For the former, we assigned
the value “1”, meaning that they had spent between 0 and 10% of the period before 19 years in
the HC.4 For the latter, we assigned “2” if the period constituted between 11 and 20% of the
period before 19 years, “3” if it constituted between 21-30% of that period, etc. (up to 50%).
Table 2 repeats the 8 factors, the numerical values and provides group means.
[ INSERT TABLE 2 ]
4 The reason for not calculating specific percentages was that participants could not remember exactly for every
single year in their childhood how many weeks they had spent in the heritage country.
17
The table shows that the 2L1ers from France stand out, as they sometimes prefer to speak their
heritage language German. Being resident in Germany, they also differ from the other groups in
using their heritage language more often at the time of testing and in having spent more time in
the heritage country during adulthood. The German majority speakers with French as their
heritage language differ from the other groups with respect to their childhood: they had a lot of
contact with their heritage language and spent a lot of time in Francophone countries. Both
groups of German-Italian bilinguals spent relatively little time in the heritage country, but the
majority of the German heritage speakers from Italy attended a bilingual school.
4.3 Preparation
We designed four separate experiments to test FA in French, German and Italian. There were two
separate experiments for German, one with speakers of German who had a potential French
accent, and another one with speakers of German who had a potential Italian accent. The two
experiments testing German-French speakers each contained speech samples of 30 speakers; the
two experiments testing German-Italian speakers each contained samples of 28 speakers.
Each of the four experiments included samples from 5 monolingual speakers, 5 L2
speakers, and between 8 and 10 samples of the 2L1ers’ for whom the language to be judged was
the minority or majority language, respectively (see Table 3 for an overview).5
[ INSERT TABLE 3 ]
5 The 2L1 and L2 data is part of the HABLA-corpus (Hamburg Adult Bilingual LAnguage) (Kupisch, Barton,
Bianchi, & Stangen, 2012), which can be accessed via the following link: http://www.corpora.uni-
hamburg.de/sfb538/en_e11_habla.html.
18
Speech samples were cut out from naturalistic interviews of 20–30 minutes in duration. Two
passages from each speaker were selected, one lasting 15 seconds and one lasting 30 seconds
(±10%).6 These passages did not contain any information on cultural background, no morpho-
syntactic errors, few or no hesitations and no long pauses. In some cases the samples were
adjusted (e.g., by shortening the pauses or cutting out an utterance with a grammatical error),
while making sure that they still sounded natural.7 In individual cases, the volume had to be
adjusted.
4.4 Raters
Each experiment was rated by 20–23 monolingual (L1) speakers of the respective language.
Again, the criterion for being monolingual was that only one language was used at home during
childhood and this was the only language of instruction at school. 23 French raters were recruited
in Paris (France), 20 Italian raters were recruited in Siena (Italy), and 41 German raters were
recruited in Hamburg (Germany).
Since there is some controversy regarding the question of whether it is preferable to have
phonetically trained raters or “ordinary” individuals (Thompson, 1991, p. 183), we recruited two
6 This duration was motivated on the basis of a pilot study: Raters were given a sample lasting 60 seconds. They
were instructed to listen to the samples carefully and to say “stop” as soon as they noticed a foreign accent. We
found that most raters detected an accent within the first 15 seconds, but they needed more time when rating 2L1
speakers. Raters rarely detected an accent within the last 15 seconds of the 60-second sample, which is why we opted
for a total of 45 seconds of speech sample in the experiment.
7 We are aware that the manipulation of pauses in the speech samples might have influenced the results. The reason
for reducing pauses was that the samples were comparatively short, and we wanted the raters judge a comparable
amount of spoken data from each 2L1er.
19
groups of raters for each experiment: speakers who had had intensive contact with the language
potentially causing the foreign accent (experienced) and raters who had never had any experience
with the language potentially causing the accent (inexperienced). Most of the experienced raters
were university students and students of linguistics. For example, experienced raters in the Italian
experiment were native-speakers of Italian who had studied German at university. Inexperienced
raters, by contrast, had no knowledge of the contact language.
4.5 Procedure
Stimuli were presented over headphones with a PowerPoint presentation. There was a training
period with two examples, one from a L1 speaker, and one from a L2er with a clearly identifiable
foreign accent. These two speakers were not part of the experiment. Prior to testing, we
emphasized that a regional accent should be counted as native.
The method we used was inspired by De Leeuw et al. (2010) with some modifications. The
raters were asked to judge samples for foreign accent in four steps. They first heard a speaker for
15 seconds and had to decide whether this speaker sounded foreign or native. In a second step,
they had to indicate how certain they were between certain, semi-certain and uncertain. If a rater
considered a speaker to sound foreign, we asked him to explain which phonetic or phonological
features he based his decision on.8 Raters were encouraged to mimic or repeat certain words or
passages if they had problems using linguistic terminology. Raters then heard another sample
8 Due to limitations on space, we will not report the detailed results of this part of the study. Generally speaking, the
raters perceived properties in which the two contact languages of a speaker differed: e.g., nasals, intonation and
pronunciation of /r/ in French; pronunciation of /ʎ/ and /r/, geminates and intonation in Italian; pronunciation of /r/,
[ç] and /h/, intonation, vowel duration and the absence of a glottal stop in German. Sometimes raters also commented
on the choice of words and contents, although they had not been told to do so.
20
from the same speaker, this time for 30 seconds. After listening to the second sample, they had
the chance to revise their judgment. This option was added to find out whether it takes longer to
determine the accent of a 2L1 speaker.
Figure 1 shows the PowerPoint-slides of the experimental design (to be read from left to
right, top to bottom; colors have been changed).
[ INSERT FIGURE 1 ]
The experiment took between 30 and 40 minutes and there were two semi-randomized test
versions for each of the four experiments.
4.6 Analysis
For the comparison of the four groups, we performed mixed-effects regression analyses. In these
analyses, group was the main predictor of interest (fixed factor). The levels of the group predictor
were coded such that the regression coefficients reflected repeated contrasts. That is, the L1ers
were compared with the 2L1ers in their majority language; the 2L1ers in their majority language
with the 2L1ers in their minority language; and the 2L1ers in their minority language with the
L2ers. In this way, the output of the analysis reflects the order of the expected differences, and
there is no need for an experiment-wise correction of the significance level (p = 0.05), since the
contrasts are planned. The analysis of foreign accent was a logistic regression analysis, since
speakers were classified as either native or foreign. The analysis of the confidence ratings was a
regular mixed-effects regression analysis. In both analyses, items and subjects were random
factors. For the analysis of the background predictors, we looked at nonparametric correlation
21
coefficients (Spearman) between each of the predictors and the relative frequency with which a
speaker was classified as having an accent.
5. Results
5.1 Accent ratings
For an overview, Figure 2 shows the results of the first part of the experiment, where raters had to
judge after 15 seconds whether a talker sounded foreign or not, comparing the four groups
(L1ers, L2ers, 2L1ers in their minority language, 2L1ers in their majority language) in all four
experiments. Speakers of German with a potential Italian accent (Italian with German) have been
separated from speakers of German with a potential French accent (French with German), as in
the experiments. Figure 2 shows a general tendency for 2L1ers resemble L1ers in the majority
language, while resembling L2ers in the minority language, despite differences across languages
(especially for the L1ers and L2ers, who constitute smaller groups).9
[ INSERT FIGURE 2 ]
Figure 3 shows the combined results of the four experiments. The difference between the L1ers
and the 2L1ers in their majority language was not significant (B=-0.3445, SE=0.5660, z=-0.609,
p = .5428). The difference between the 2L1ers in their majority language and the 2L1ers in their
9 One monolingual L1 speaker of German with a regional (Franconian) accent sounded foreign to some of the raters.
The same speaker was sometimes considered foreign because her pronunciation of a foreign (English) name sounded
“too authentic” for a German speaker.
22
minority language was significant (B=4.6484, SE=0.2598, z=18.605, p<.0001), and so was the
difference between the 2L1ers in their minority language and the L2ers (B = 1.3791, SE =
0.5452, z = 2.530, p = 0.0114).
[ INSERT FIGURE 3 ]
Clearly, 2L1ers were rated as “having a foreign accent” less often while speaking their majority
language compared to speaking their minority language.
We explored accent ratings in majority and minority languages further by comparing the
ratings in the 2L1 speakers’ majority and minority language. Table 4 shows the proportion of
times speakers in a group were judged as native while speaking their majority language, and
while speaking their minority language. Not surprisingly, the proportions are higher when the
speakers speak their majority language, with the German-French bilinguals receiving the highest
scores, and the German-Italian bilinguals receiving the lowest scores. There is considerably more
variation when the speakers speak their minority language. Once more, the German-French
bilinguals received the highest scores, and the Italian-German bilinguals received the lowest
scores.
[ INSERT TABLE 4 ]
Figure 4 compares the proportion of times each individual 2L1er was considered foreign in
his/her two languages. Consistent with Table 4, for most speakers the right bar (representing the
minority language) is higher than the left bar (representing the majority language). Only three
23
speakers constitute an exception to this overall pattern, having either no foreign accent in the
minority language or a greater foreign accent in the majority language. No speaker has a
pronounced accent (above 50%) in both languages.
[ INSERT FIGURE 4 ]
We further explored how consistently individual speakers were judged as foreign or native.
We took 80% consistency as a benchmark for grouping speakers: They were grouped as native
when taken for a native speaker by 80% of the raters, as non-native when taken for a native
speaker by less than 20% of the raters, and as unclear when taken for a native speaker by
between 20% and 79% of the raters. This categorization yields the results illustrated in Table 5.
[ INSERT TABLE 5 ]
Clearly, varying (i.e. unclear) judgments are most typical for 2L1ers in their minority language.
L1 monolingual speakers are predominantly classified as native, L2ers predominantly as foreign,
and 2L1ers in their majority language as native. A few unclear cases appear in all three groups.
By contrast, 2L1ers tend to sound foreign in the minority language (55%), but sometimes pass as
natives (11%) or receive varying judgments (34%). The number of varying judgments is higher in
this group than in all others. Judgments for the minority language differ between the German-
Italian and the German-French 2L1ers, the former being classified as foreign-sounding more
often.
5.2 Correlation with other factors
24
In this part of the analysis we investigated which of the variables introduced in Table 2 correlated
with the proportion of times an individual was identified as having a foreign accent. Table 6
shows the correlation matrix (nonparametric correlation coefficients). The variables that correlate
most strongly with a native accent are language preference (LP), language use at the time of
testing (CuFU) and length of residence in the country where the heritage language was spoken
before age 19;0 (LoRHC 0-18;11).
[ INSERT TABLE 6 ]
5.3 Confidence ratings
Figure 5 shows how confidently the raters judged individual speaker groups as foreign or native.
Note that for this analysis confidence can be high regardless of whether someone was judged as
native or foreign. In other words, speakers judged “native” and speakers judged “foreign” are
treated alike. The expectation was for the 2L1ers to be judged (as foreign or native) with less
confidence than the other groups.
[ INSERT FIGURE 5 ]
On a scale between 1 (certain) and 3 (uncertain), 2L1 speakers in their minority language were
judged with an average degree of uncertainty of 1.53, which was higher than that obtained for the
L1 monolingual speakers (1.32), the 2L1ers in their majority language (1.32) and the L2ers
(1.38).
25
The results from this second analysis of uncertainty parallel those from the first analysis of
accent ratings. There was no significant difference between the L1ers and the 2L1ers in the
majority language (B = -0.02831, SE = 0.07576, t = -.037, p = .7087), but there were significant
differences between the 2L1ers in their majority language and the 2L1ers in their minority
language (B = 0.26636, SE = 0.03384, t = 7.87, p < .0001) and between the 2L1ers in their
minority language and the L2ers (B = -0.18005, SE = 0.07576, t = -2.38, p = 0.0176). In other
words, the comparisons suggest that the foreign-accent classifications for the 2L1ers in their
minority language were made with a slightly, but significantly, higher degree of uncertainty than
the classifications of the other speakers.
Finally, we investigated whether raters revised their judgments significantly more often
when judging the accent of the 2L1ers as compared to the other groups. Figure 6 shows the
results of this comparison.
[ INSERT FIGURE 6 ]
The results showed no significant difference between the L1 and the 2L1 speakers in the
majority language (B = -0.05752, SE = 0.37199, z = -0.155, p = 0.8771). On the other hand, the
difference between the 2L1ers in their majority language and in their minority language was
significant (B = 0.54234, SE = 0.22677, z = 2.392, p = 0.0168). The difference between the
2L1ers in their minority language and the L2ers was marginally significant (B = -0.70982, SE =
0.37488, z = 01.893, p = 0.0583). These comparisons confirm that the raters were more uncertain
in their foreign accent classifications of 2L1ers speaking their minority language.
6. Discussion
26
Our study set out to explore if adult simultaneous bilinguals (2L1ers) have a native accent in their
two languages. Results showed that language input from birth does not guarantee a native accent.
On the other hand, some 2L1ers have a native accent in both of their languages, which raises the
question of which factors besides AoO determine one’s accent.
6.1 Minority vs. majority language setting
We further asked whether differences in accentedness could be observed depending on whether
the language of a 2L1er represented the dominant/majority language of the society (e.g. French in
France) or a heritage/minority language (e.g. French in Germany) at two time points: a) in
childhood, when the two languages were acquired, and b) in adulthood, after the two languages
had been acquired.
On the group level, we found that 2L1ers were within the range of native speakers in the
majority language of their childhood environment, while being more similar to L2ers (i.e., in
terms of having a perceivable foreign accent) in their minority language (Figures 2,3, Table 4).
We can therefore also conclude that whether or not a language constitutes a speaker’s minority
language during childhood and adolescence plays a role for his accent in this language later in
life.
On the individual level, speakers had a native accent in the majority language, while their
accent ranged from being native-like to being foreign in their minority language (Figure 4),
suggesting that factors other than AoO must be involved. The same comparison showed that
having a native accent in the majority language need not imply a foreign accent in the minority
language, as some 2L1ers had a native-like accent in both languages.
Among the speakers we tested, some had left their childhood environment to live in a
country where the heritage language was spoken as a majority language. This was the case for all
27
German-French 2L1ers who grew up in France. Since these speakers tended to have a FA in their
heritage language German (36% on average), while continuing to have a native accent in French,
we can conclude that changes in the language environment during adulthood have a
comparatively small impact on accent. This is also confirmed by the German-Italian speaker who
grew up in Italy but moved to Germany at age 19, where he continued to live for 20 years
(D13_STE in Table 4). Furthermore, native accent in the heritage language was found to
correlate with LoR in the heritage country during childhood but not during adulthood (Table 6).
Finally, findings for the French-German 2L1ers from France allow us to conclude that 2L1ers can
avoid attrition when they leave their childhood environment to live in the country where their
heritage language is spoken, even though their input space during childhood was divided between
two languages.
6.2 Sociolinguistic and affective factors
Although we found a tendency for the heritage language to exhibit a FA, this was not the case
across the board. We therefore explored which factors other than AoO could play a role, finding
that a native accent correlated with language preference (LP), frequency of contact and language
use at the time of testing (CuFU) and length of residence in the heritage country during childhood
and adolescence (LoRHC 0-18;11). The relevance of LoRHC 0-18;11 could imply that phonetic
and phonological properties —in addition to being acquired early— also need a period of
stabilisation with input from several different native speakers early in life.
To our surprise, attending a bilingual school (TSH) was unrelated to accent. The German
heritage speakers in Italy provide a good example because out of 8 speakers 7 were
unambiguously rated as foreign in German, although 5 attended a German school. We were also
surprised to find that frequency of use and contact during childhood (ChFU) did not correlate
28
with accent, but we suspect that integrating the school situation into this variable may have
influenced the results. Looking at language use in the home context exclusively might have
yielded a different result (see also footnote 3). LoR in the heritage country during adulthood,
neither in relative nor in absolute numbers, as well as age, were unrelated to accent.
At first sight, finding a correlation between a native-like accent and CuFU, while finding no
correlation between native accent and LoR after 19;0, may look like a contradiction, given that
the group with the highest CuFU and the longest LoR are one and the same group, namely the
heritage speakers of German from France (see Table 2). It is possible, however, that this
contradiction is due to speakers of French from Germany. Although as adults they have not been
in France for a longer period of time, many of them went to French schools and are part of a
network of French-speaking people in Hamburg.
6.3 Rating the accent of a 2L1 speaker
Another question we investigated was whether it is more difficult to determine the accent of
2L1ers as compared to L1ers and L2ers. To address this question, we determined the raters’
degree of confidence when judging a speaker’s accent, and how often they revised their judgment
when they had the chance to listen to the speaker for a longer time. We found that the foreign-
accent classifications for the 2L1ers in their minority language were made with a slightly, but
significantly, higher degree of uncertainty than the classifications of the other speakers. By
contrast, in the majority language 2L1ers were judged with the same degree of confidence as
monolingual speakers. Furthermore, raters revised their judgments significantly more often when
judging the accent of 2L1ers in their minority language than when judging them in their majority
language, while showing no significant differences in judging 2L1ers in their majority language
29
and L1ers. These findings suggest that it is more difficult to determine the accent of a 2L1er, but
only in his minority language.
6.4 Language-specific properties
We compared speakers with different native languages and there were differences across
language groups (Figure 2). The question may therefore be raised whether a native accent is more
difficult to achieve in some language pairs than in others because of differences in the complexity
and/or markedness of their phonetic and phonological systems (e.g., in terms of phoneme
inventory, syllable types and rhythmic properties). In other words, could cross-linguistic
phonological similarity be an explanation for why the 2L1 groups differ in terms of
accentedness? Let us consider the results. First, the German-French 2L1ers are perceived as
relatively less accented in their minority language (see Figure 2). A possible explanation could be
that their minority language (whether German or French) is more similar to their majority
language in the complexity of their vowel inventories (as compared to Italian). Second, the two
groups of 2L1ers with German as a minority language performed differently: the German-French
group outperformed the German-Italian group. The explanation could be that knowing French —
a language with a complex phoneme inventory— is more beneficial to speaking German accent-
free than knowing Italian —a language with a less complex phoneme inventory (especially with
regard to vowels).
Although this is a possibility, there are several reasons why we are wary of drawing such a
conclusion. Although the sound systems of both German and French may be considered complex
compared to the Italian sound system, they do not overlap. For instance, French has nasal vowels,
German does not. In German, vowel length is distinctive, in French it is not. The German sound
inventory includes /ʔ/ and /h/, which are hard to perceive and pronounce by speakers of French.
30
Moreover, the pronunciation of <r> is phonologically conditioned and the rules are not the same
for German and French. Finally, the languages have very different intonation patterns. In fact, all
these properties were perceived by the raters when judging the 2L1ers’ accents, indicating that
the differences between the two systems pose problems. Moreover, an explanation based on the
complexity of the sounds systems does not explain why minority speakers of French
outperformed minority speakers of Italian, although both have German as a contact language.
Finally, the phonological system of Italian, being less complex, should be relatively easy to
acquire, but French heritage speakers outperformed Italian heritage speakers. Thus, if
phonologically similarity/complexity played a role, it could have been only be one of several
factors that determined acquisition outcomes.
A related possible assumption is that a comparatively high amount of dialectal variation
makes it harder to acquire a native-like accent in some languages. In fact, both German and
Italian display more dialectal variation than French. It seems fair to say that especially Italians are
a lot more conscious about linguistic variation in their country than speakers of French. Not only
is Italy home to many additional Romance languages (Sardinian, Ladin, Occitan, Franco-
Provençal etc.), but it is also a country in which many different dialects are still used in daily life.
Moreover, the La Spezia-Rimini line —a boundary that demarcates several important isoglosses
dividing eastern and western Romance languages— cuts through Italy, which partially explains
the impressive diversity between Northern and Southern dialects, as well as between regional
varieties of Italian (e.g. Sicilian vs. Piedmontese dialects). Since the heritage speakers of French
had the most native-like accent, it is possible that less variation in the input language leads to a
more native-sounding accent later in life. However, the difference between the two groups of
German heritage speakers would still not be accounted for. Thus, like phonological complexity,
31
dialectal variation in the input could only be one of several possible explanations for the
acquisition outcomes.
7. Conclusions
We have compared the perceived global accent of 38 2L1 speakers in both of their languages to
that of L1 monolingual and L2 speakers. Based on judgments of 84 monolingual raters, we found
that the 2L1ers were within the range of native speakers in the majority language of their
childhood environment, while being mostly within the range of L2 speakers in their heritage
language. Results suggest that besides age sufficient and continued input is a necessary
prerequisite for attaining a native-like accent. Moreover, in order to achieve a native-like
pronunciation, it may be beneficial if the input comes from a variety of different (native) speaker
sources. This is the case when a majority language is acquired, but not when a minority language
is acquired. Thus, receiving too little (and/or more foreign-accented) input constitutes a general
challenge — not only for late learners but also for early learners.
We explored the correlation of foreign accent with extra-linguistic factors, such as
schooling and amount of language contact and use during childhood and adulthood. Results
indicate that schooling has a lower impact than residence in a country where the heritage
language is spoken, and living in the heritage language country is more beneficial to a native-like
accent if it happens during childhood than if it happens during adulthood. Nevertheless, current
use and exposure can positively affect the accent even during adulthood, as witnessed by the
German heritage speakers from France, who lived in Germany at the time of the study, and
whose accent sounded less foreign than the accent of the German heritage speakers from Italy
32
(Figure 2). A native-like accent also correlates with language preference, suggesting that the role
of affective factors cannot be denied either.
Our results are interesting from the perspective of attrition research, since we have shown
that a native-like accent in the majority language of the childhood environment can be
maintained, even after long-term residence (up to 20 years) in the heritage language country.
Finally, we have shown that raters have more difficulties when judging 2L1ers in the minority
language than when judging 2L1ers in the majority language, and L2ers. Our study leaves open
the question of whether minority language speakers are indistinguishable from native speakers in
their perception of phonemic contrast, as reported by Flores and Rauber (2011) for heritage
speakers of German—an aspect left to future research.
References
Abrahamsson N., & Hyltenstam, K. (2009). Age of Onset and Nativelikeness in L2: Listener
Perception Versus Linguistic Scrutiny. Language Learning, 59, 249–306.
Au, T. K., Knightly, L. M., Jun, S.-A., & Oh, J. S. (2002). Overhearing a language during
childhood. Psychological Science, 13, 238–243.
Benmamoun, E., Montrul, S., & Polinsky, M. (2013). Heritage languages and their speakers:
Opportunities and challenges for linguistics. Theoretical Linguistics, 39(3/4), 29–181.
Bianchi, G. (2013). Gender in Italian-German Bilinguals: a comparison with German L2 learners
of Italian. Bilingualism: Language and Cognition, 16(3), 538–557.
Birdsong, D. (1999). Introduction: Whys and why nots of the critical period hypothesis for
second language acquisition. In D. Birdsong (Ed.), Second language acquisition and the
critical period hypothesis (pp. 1–22). Mahwah, NJ: Lawrence Erlbaum.
33
Birdsong, D. (2003). Authenticité de prononciation en français L2 chez des apprenants tardifs
anglophones: analyses segmentales et globales. Acquisition et interaction en langue
étrangère, 18, 2–14.
Birdsong, D., & Gertken, L. M. (2013). In faint praise of folly. A critical review of native/non-
native speaker comparisons, with examples from examples of native and bilingual
processing of French complex syntax. Language, Interaction and Acquisition, 4(2), 107–
133.
Bongaerts, T., Planken B., & Schils, E. (1995). Can late learners attain a native accent in a
foreign language? A test of the critical period hypothesis. In D. Singleton & Z. Lengyel
(Eds.), The age factor in second language acquisition (pp. 30–50). Clevedon, U.K.:
Multilingual Matters.
Bongaerts, T., van Summeren, C., Planken, B., & Schils, E. (1997). Age and ultimate attainment
in the pronunciation of a foreign language. Studies in Second Language Acquisition, 19,
447–465.
Chang, C. B., Yao, Y., Haynes, E. F., & Rhodes, R. (2011). Production of phonetic and
phonological contrast by heritage speakers of Mandarin. Journal of the Acoustical Society
of America, 129, 3964–3980.
De Leeuw, E., Schmid, M., & Mennen, I. (2010). Perception of foreign accent in native speech.
Bilingualism: Language and Cognition, 13(1), 33–40.
Flege, J. E., Munro, M. J., & MacKay, I. R. A. (1995). Factors affecting strength of perceived
foreign accent in a second language. Journal of the Acoustical Society of America, 97,
3125–3134.
Flege, J., Yeni-Komshian, G., & Liu, S. (1999). Age constraints on second
language learning. Journal of Memory and Language, 41, 78–104.
34
Flege, J.E., Frieda, E.M., & Nozawa, T. (1997). Amount of native-language (L1) use affects the
pronunciation of an L2. Journal of Phonetics, 25, 169–186.
Flege, J.E., MacKay, I.R.A., & Piske, T. (2002). Assessing Bilingual Dominance. Applied
Psycholinguistics, 23(4), 567–98.
Flores, C. M., & Rauber, A. S. (2011). Perception of German vowels by bilingual Portuguese-
German returnees: A case of phonological attrition? In E. Rinke & T. Kupisch (Eds.), The
Development of Grammar: Language Acquisition and Diachronic Change - Volume in
honor of Jürgen M. Meisel (pp. 287–305). Amsterdam: John Benjamins.
Fowler, A. C., Sramko, V., Ostry, D. J., Rowland, S. A., & Hallé, P. (2008). Cross language
phonetic influences on the speech of French-English bilinguals. Journal of Phonetics, 36,
649–663.
Gathercole, V. C. M. & Thomas, E.M. (2009). Bilingual first language development: Dominant
language takeover, threatened minority language take-up. Bilingualism: Language and
Cognition, 12, 213–238.
Godson, L. (2004). Vowel Production in the Speech of Western Armenian Heritage Speakers.
Heritage Language Journal 2. http://www.international.ucla.edu/languages/ heritage
languages/journal/article.asp?parentid=14648
Grosjean, F. (1982). Life with two Languages. An Introduction to Bilingualism. Cambridge,
Mass: Harvard University Press.
Hopp, H., & Schmid, M. (2013). Perceived foreign accent in L1 attrition and L2 acquisition: the
impact of age of acquisition and bilingualism. Applied Psycholinguistics, 34(2), 361–394.
Ioup, G., Boustagui, E., El Tigi, M., & Moselle, M. (1994). Reexamining the critical period
hypothesis: A case study in a naturalistic environment. Studies in Second Language
Acquisition, 16, 73–98.
35
Jesney, K. (2004). The use of global foreign accent rating in studies of L2 acquisition. Annotated
bibliography, Language Research Centre University of Calgary.
Kupisch, T. (2012). Generic subjects in the Italian of early German-Italian bilinguals and German
learners of Italian as a second language. Bilingualism: Language and Cognition, 15(4), 736–
756.
Kupisch, T. (2013). A new term for a better distinction? A view from the higher end of the
proficiency scale. Theoretical Linguistics, 39(3-4), 203-214.
Kupisch, T. (2014). Adjective placement in simultaneous bilinguals (German-Italian) and the
concept of cross-linguistic overcorrection. Bilingualism: Language and Cognition, 17(1), 222–
233
Kupisch, T., D. Akpinar, & A. Stöhr. 2013. Gender assignment and gender agreement in adult
bilingual and second language speakers of French. Linguistic Approaches to Bilingualism,
3(2), 150–179.
Kupisch, T., Barton, D., Bianchi, G., & Stangen, I. (2012). The HABLA-corpus of adult
bilinguals (German- French and German-Italian). In T. Schmidt & K. Wörner (Eds.),
Multilingual Corpora and Multilingual Corpus Analyses, Hamburg Studies in
Multilingualism 14. (pp. 153–169), Amsterdam: Benjamins.
Lenneberg, E. H. (1967). Biological foundations of language. New York: Wiley.
Long, M. H. (1990). Maturational constraints on language development. Studies in Second
Language Acquisition, 12, 251–285.
Mack, M. (1989). Consonant and vowel perception and production: Early English-French
bilinguals and English monolinguals. Perception and Psychophysics, 46, 187–200.
Major, R.C. (1992). Losing English as a First Language. The Modern Language Journal, 76(2),
190–208.
36
Meisel, J.M. (2011). First and second language acquisition: Parallels and differences.
Cambridge: Cambridge University Press.
Montrul, S. (2008). Incomplete Acquisition in Bilingualism. Re-examining the age factor.
Amsterdam: Benjamins.
Montrul, S. (2010). Current Issues in Heritage Language Acquisition. Annual Review of Applied
Linguistics, 30, 3-23.
Montrul, S. A. (2012). Is the heritage language like a second language? EUROSLA Yearbook, 12,
1–29.
Moyer, A. (1999). Ultimate attainment in L2 phonology. Studies in Second Language
Acquisition, 21, 81–108.
Moyer, A. (2004). Age, accent, and experience in second language acquisition: an integrated
approach to critical period inquiry. Clevedon, UK: Multilingual Matters.
Oh, J., Jun, S., Knightly, L., & Au, T. (2003). Holding on to childhood language memory.
Cognition, 86, 53–64.
Oyama, S.C. (1976). A sensitive period for the acquisition of a phonological system. Journal of
Psycholinguistic Research, 5, 261–283.
Piske, T., MacKay, I.R.A., & Flege, J. E. (2001). Factors affecting degree of foreign accent in an
L2: a review. Journal of Phonetics, 29, 191–215.
Polinsky, M., & Kagan, O. (2007). Heritage languages: In the ‘wild’ and in the classroom.
Language and Linguistics Compass, 1, 368–395.
Rothman, J. (2009). Understanding the nature of early bilingualism: Romance languages as
heritage languages. International Journal of Bilingualism, 13(2), 155–163.
Schneiderman, E., & Desmarais, C. (1988). The talented language learner. Some preliminary
findings. Second Language Research, 4, 91–109.
37
Scovel, T. (1969). Foreign accents, language acquisition, and cerebral dominance. Language
Learning, 19, 245–254.
Snow, C.E., & Hoefnagel-Höhle, M. (1977). Age differences in the pronunciation of foreign
sounds. Language & Speech, 20, 357–365.
Sundara, M., Polka, L., & Baum, S. (2006). Production of coronal stops by simultaneous
bilingual adults. Bilingualism: Language and Cognition, 9(1), 97–114.
Thompson, I. (1991). Foreign accents revisited: the English pronunciation of Russian immigrants.
Language Learning, 41, 177–204.
Yeni-Komshian, G. H., Flege, J. E., & Liu, S. (2000). Pronunciation proficiency in the first and
second languages of Korean-English bilinguals. Bilingualism: Language and Cognition,
3(2), 131–149.
38
TABLES AND FIGURES
Participant group Language(s) Number Majority lg. during childhood
Monolingual French 5 French
Monolingual German 10 German
Monolingual Italian 5 Italian
L2 French (L1 German) 5 French
L2 German (L1 French) 5 German
L2 German (L1 Italian) 5 German
L2 Italian (L1 German) 5 Italian
2L1 in France German-French 10 French
2L1 in Germany German-French 10 German
2L1 in Germany German-Italian 10 German
2L1 in Italy German-Italian 8 Italian
Table 1. Overview of participants (speakers) in the four experiments
39
Majority language:
Minority/heritage language:
French
German
German
French
German
Italian
Italian
German
Range
1) LP (Language Preference) 1.6 1.1 1.13 1.4 1-3
2) ChFU (Childhood Frequency of
Use)
1.7 2.7 2.38 1.9 1-3
3) CuFU (Current Frequency of Use) 3.9 2.4 2.1 1.63 1-4
4) TSH (Type of Schooling in the
Heritage Language)
1.9 2.6 1.3 2.25 1-3
5) Age 33 27 30 27 18-38
6) LoRHC (Length of Residence
Heritage Country) after 19;0 (in %) 52.4% 17.8% 20.2% 10.8% 0-100%
7) LoRHC after 19;0 (in years) 8.68 0.51 1.96 1.85 0-20 years
8) LoRHC 0-18;11 1.7 2 1.4 1 1-5
Table 2. Differences in preference, use, schooling and residence with respect to the heritage language
40
.
Monolingual
L1ers L2ers
2L1ers speaking the
majority language
2L1er speaking the
minority language Total
French experiment 5 5 10 10 30
German experiment (with
German-French speakers) 5 5 10 10 30
German experiment (with
German-Italian speakers) 5 5 10 8 28
Italian experiment 5 5 8 10 28
Table 3. Distribution of speaker samples in the four experiments
41
majority language:
minority language:
Italian
German
French
German
German
Italian
German
French
2L1ers speaking majority language: 0.96 0.90 0.86 0.99
2L1ers speaking minority language: 0.09 0.36 0.30 0.49
Table 4. Proportion of times speakers were rated as 'native accent' in the two 2L1 groups
42
Italian French German
(with Italian)
German
(with French) Total
as L1 and only
language 5/0/0 5/0/0 4/1/0 3/2/0 17/3/0
as majority language
(for 2L1er) 7/1/0 8/2/0 7/3/0 10/0/0 32/6/0
as minority language
(for 2L1ers) 2/2/6 2/5/3 0/1/7 0/5/5 4/13/21
as L2 0/3/2 0/0/5 0/1/4 0/0/5 0/4/16
Table 5. Number of speakers judged as native/unclear/non-native
43
FA
LP
Ch
FU
Cu
FU
TS
H
Ag
e
LoR
HC
aft
er 1
9;0
rel.
L
oR
HC
aft
er 1
9;0
ab
s.
LoR
HC
0-1
8;1
1 r
el.
FA 1.00 -0.49**
-0.13 -0.41**
-0.09 -0.12 -0.18 -0.10 -0.32*
LP -0.49**
1.00 -0.17 0.52**
-0.21 0.11 0.29 0.29 0.25
ChFU -0.13 -0.17 1.00 -0.28 0.73**
-0.26 -0.32* -0.43
** -0.01
CuFU -0.41**
0.52**
-0.28 1.00 -0.12 0.18 0.71**
0.63**
0.30
TSH -0.09 -0.21 0.73**
-0.12 1.00 -0.20 -0.16 -0.26 0.17
Age -0.12 0.11 -0.26 0.18 -0.20 1.00 0.45**
0.67**
0.08
LoRHC after 19;0 rel. -0.18 0.29 -0.32* 0.71
** -0.16 0.45
** 1.00 0.93
** 0.18
LoRHC after 19;0 abs. -0.10 0.29 -0.43* 0.63
** -0.26 0.67
** 0.93
** 1.00 0.20
LoRHC 0-18;11 rel. -0.32* 0.25 -0.01 0.30 0.17 0.08 0.18 0.20 1.00
Table 6. Correlation matrix, sociolinguistic and affective factors regarding the minority language.
* indicates that the coefficient is significant (** p < .01; *p < .05).
44
Figure 1. Power Point Slides in the experiment (Translations: Bitte bewerte jetzt! “Please judge now!”,
Wie sicher bist Du? Sicher/halbsicher/unsicher. “How certain are you? Certain/semi-certain/uncertain”,
Woran machst Du den Akzent fest? “What is your judgment based on?”, Bleibst Du bei Deinem Urteil?
“Do you want to change your original judgment?”)
45
Figure 2. Proportion of times (%) rated as accented by talker group and languages
0
20
40
60
80
100
Italian speakers French speakers German speakers(with Italian)
German speakers(with French)
L1ers 2L1ers majority language 2L1ers minority language L2ers
46
Figure 3. Proportion of times (%) rated as accented by talker group, all languages combined
0%
20%
40%
60%
80%
100%
L1 2L1 majoritylanguage
2L1 minoritylanguage
L2
native accent
foreign accent
47
Figure 4. Proportion of times (%) individuals were rated with ‘foreign accent’ (MAJ=majority language,
MIN=minority language)
48
Figure 5. Degree of confidence when judging speakers, all languages combined
0%
20%
40%
60%
80%
100%
L1 2L1 majoritylanguage
2L1 minoritylanguage
L2
uncertain
semi-certain
certain