International Journal of Multilingualism

ISSN: 1479-0718 (Print) 1747-7530 (Online) Journal homepage: [Link]

Code-mixing and language dominance: bilingual,

trilingual and multilingual children compared

Meike Poeste, Natascha Müller & Laia Arnaus Gil

To cite this article: Meike Poeste, Natascha Müller & Laia Arnaus Gil (2019): Code-mixing and
language dominance: bilingual, trilingual and multilingual children compared, International Journal
of Multilingualism, DOI: 10.1080/14790718.2019.1569017

To link to this article: [Link]

Published online: 01 Feb 2019.

Submit your article to this journal

View Crossmark data

Full Terms & Conditions of access and use can be found at

[Link]
INTERNATIONAL JOURNAL OF MULTILINGUALISM
[Link]

Code-mixing and language dominance: bilingual, trilingual

and multilingual children compared
Meike Poestea, Natascha Müllera and Laia Arnaus Gilb
a
Romanische Sprachwissenschaft, Bergische Universität Wuppertal, Wuppertal, Germany; bRomanisches
Seminar, Universität Hamburg, Hamburg, Germany

ABSTRACT ARTICLE HISTORY

Acquisitionists generally assume a relation between code-mixing in Received 13 April 2018
young bilingual and trilingual children and language dominance. In Accepted 4 January 2019
our cross-sectional study we investigated the possible relation
KEYWORDS
between code-mixing and language dominance in 122 children Language dominance;
raised in Spain or Germany. They were bilingual, trilingual or domains and dimensions of
multilingual, the latter acquiring more than three languages. The language dominance; code-
definition of language dominance is grounded on Birdsong’s mixing; trilingualism;
(2014. Dominance in bilingualism: Foundations of measurement, typology
with insights from the study of handedness. In C. Silva-Corvalán, &
J. Treffers-Daller (Eds.), Language Dominance in Bilinguals: Issues of
operationalization and measurement (pp. 85–105). Cambridge:
Cambridge University Press) distinction between domains and
dimensions. The main result of our study is that code-mixing is rare in
a monolingual setting which means that bilingual, trilingual and
multilingual children are able to behave monolingually. Domain-
specific language dominance can explain the relatively high mixing
rate in the Catalan tests but concerning the dimensions of language
dominance no relation was found between the children’s code-
mixing and language (un)balance. A separate analysis of intra- and
intersentential mixing reveals that intersentential mixing is
determined by the typological proximity between the child´s
languages. All instances of intrasentential code-mixing, were insertional.

1. Introduction
It is generally assumed in the literature that bilingual children mix their two languages to
considerable degrees and that children with a weak language do so more often than chil-
dren who are balanced (Bernardini & Schlyter, 2004, Cantone & Müller, 2005 for an over-
view). Apart from the assumption that code-mixing indicates an unbalanced language
development, it is also claimed that language dominance influences the direction of
code-mixing, namely from the strong into the weak language. This assumption has
been relativised for bilingual children by Cantone (2007), Cantone, Kupisch, Müller, &
Schmitz (2008) and Patuto et al. (2014) because, on the one hand, they found a low
mixing rate in the speech of bilingual children investigated during a period from 1;6
until 5 years and on the other hand, it was not always possible to establish a relation

CONTACT Meike Poeste, [Link]@[Link] Romanische Sprachwissenschaft, Bergische Universität Wup-

pertal, Gaußstraße 20, Wuppertal 42119, Germany
© 2019 Informa UK Limited, trading as Taylor & Francis Group
2 M. POESTE ET AL.

between mixing and language dominance. If it comes to the acquisition of more than two
languages from birth, a group of speakers still understudied, most research assumes that
there is always (at least) one weak language (Hoffmann, 2001, p. 5). As a consequence, chil-
dren who acquire more than two languages from birth should mix their languages more
often than bilingual children. In what follows, we will study and describe the relation
between language dominance and mixed utterances in trilingual children and in children
who speak more than three languages from birth and compare them with the mixing
behaviour of children who are early bilinguals. In particular, we will look at the children’s
mixing behaviour in a monolingual speech mode (Grosjean, 2001).

2. Theoretical background
2.1. Code-mixing
In the context of language mixing in bi-, tri- and multilingual children, it is first important to
define the term code-switching.1 According to Müller et al. (2015), code-switching
describes a smooth change of languages, observed in bilinguals who master both
languages very well. Therefore, it is seen as a speech style (MacSwan, 2000, p. 38) and
not produced due to a lack of competence (Müller et al., 2015, p. 24f.). Gumperz (1967),
who coined the word, applied it to describe a discourse strategy which is used by bilingual
speakers (Riehl, 2014, p. 21). Code-switching is one of various possibilities for bilinguals to
express different illocutions and to point out that the listener has to interpret the forth-
coming message somewhat differently (cf. metaphorical code-switching, Chan, 2004).
For a long time, research on CS has focused on the social factors which favour CS (cf. situa-
tional CS, Milroy & Muysken, 1995, among others). Poplack (1980), who analysed bilingual
speakers with English and Spanish in New York, concentrated on a further aspect, namely
the study of the underlying morpho-syntactic machinery of intrasentential CS (Jansen,
Müller, & Müller, 2012, p. 383). This research has led to the distinction between inter-
and intrasentential CS, where the former relates to mixing between sentences or larger
parts of speech and the latter refers to mixing within sentences (Müller et al., 2015, p. 15).
It has to be stressed that CS is a speech style which many linguists conceive as a sign for
a high degree of competence in the respective languages. This competence in both
languages, a prerequisite for CS, distinguishes it from other phenomena of language
contact, for example from code-shifting (Silva-Corvalán, 1983) and borrowing (Müller
et al., 2015, p. 18; Riehl, 2014, p. 22). Another term to clarify is transfer. In contrast to CS,
transfer is a process which is not controlled by the speakers (Müller et al., 2015, p. 22).
Specifically, the term transfer is used if the speaker uses for example the syntax of
language A (her/his mother tongue) and language material from language B (her/his
second, third or x-th language). In contrast, CS is characterised by language material in
terms of vocabulary items of at least two languages. While CS thus requires a high com-
petence in all languages involved, the process of transfer has often been seen as a strategy
of simplification in the context of foreign language learning (Müller et al., 2015, p. 22f.).
In order to illustrate the different functions of CS, see the following example of situa-
tional CS taken from Sivakumar (2017) who analysed CS within the (simultaneous) trilin-
gual child Diego. Diego acquires Spanish, Italian and French from birth.
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 3

(1) Situation: Diego and his mother are playing to give a present to someone. He turns to his
father and gives him a present. While the mother speaks Spanish (SP) with Diego, the
language of the communication between the father and Diego is Italian (IT).

Mother ahíSP noSP estáSP / noSP teSP preocupesSP / ‘it is not there / do not worry’ /
Diego esSP eh – esoSP esoSP eh eh /eccoIT meIT – questoIT perIT teIT / tieneSP ilIT tuoIT legaloIT
(= regaloIT) / quelloIT èIT ilIT mioIT legaloIT (= regaloIT) / quelloIT dueIT èIT legaloIT
(= regaloIT) / ‘it is er - this this er er / here I - this is for you / (he/she/it) has your
present / that is my present / those two is present’ (those two are presents) /
Mother ySP miSP regaloSP ↑ / esSP miSP cumpleañosSP hoySP loSP sabesSP diego ↑ / ‘and my
present ↑ / it is my birthday today you know diego ↑ / (Diego 3;1,132)

Example (1) shows an instance of situational CS. In contrast to situational CS, metaphorical
CS is rather infrequent in children’s productions since it requires a high degree of prag-
matic competence which normally appears later in language production (Jisa, 2000, p.
1366). Nevertheless, Sivakumar (2017) also shows that the majority of mixing in the ana-
lysed child Diego (from 2;8,10 until 4;9,22) is neither situational nor metaphorical CS. There
are often no pragmatic reasons for children to use a language different from the context
language (Sivakumar, 2017, p. 91).
Due to the difficulty to differentiate between code-switching and other phenomena of
language contact, Muysken (1997) distinguishes different processes when languages are
mixed, namely insertion, alternation and congruent lexicalisation. According to
Muysken, congruent lexicalisation is characterised by a shared grammatical structure of
elements from different languages (Muysken, 1997, p. 362). Furthermore, the effect of
going back and forth between the languages suggests that with respect to congruent lex-
icalisation, there are elements from the two languages inserted into a shared structure
(Muysken, 1997, p. 362). Due to the complexity of this process, this phenomenon requires
utterances with more words than the children in these studies normally produce since
they are still rather young.3 With regard to the other two processes, insertion can be
defined by embedding language material from one into the respective other language
(s) of a multilingual speaker (Müller et al., 2015, p. 18), whereas alternation means a
smooth change between languages that involves grammar as well as vocabulary
(Müller et al., 2015, p. 17). Therefore, intrasentential mixing can contain disintegrated
material (insertion) or integrated material (alternation). Alternation is thus assumed to
be more complex. Muysken (1997, p. 362) writes from a psycholinguistic perspective
that the components of both languages are activated to different degrees: in the case
of alternation, activation would switch from one language to the other, in the case of inser-
tion, the activation of one language would be ‘temporarily diminished’. Complexity differ-
ences are also confirmed by the difficulty to distinguish between insertion and borrowing
– a phenomenon that requires less competence in both languages (Müller et al. 2015, p.
18). Having in mind these differences in producing intrasentential code-mixing, a separate
analysis of these phenomena may be appropriate.
Important to add in this context is the distinction between multilingual adults or multi-
lingual children. Arguably, the child is still in the process of building her/his competence of
the languages s/he hears in the input. In other words, the child is on her/his way to
become a competent (native) speaker of the languages. This aspect further complicates
the categorisation of a child’s utterance that contains vocabulary items from language
4 M. POESTE ET AL.

A and language B in terms of CS or an outcome of another contact phenomenon (Meisel,

1994). Because of the difficult task of differentiating between the phenomena of language
contact in the young child, researchers have used the term code-mixing that works as a
cover term in order to express the fact that two or more languages are mixed clause-
internally or across clauses (Müller et al., 2015, p. 24). For this reason, we will use the
term code-mixing in what follows to describe the situation that vocabulary items from
two or more languages have been used and not the term code-switching which would
already indicate that mixing languages is used as a discourse strategy. Therefore, the fol-
lowing two examples 2a (intersentential mixing) and 2b (intrasentential mixing) illustrate
what will constitute mixing in our analysis.
(2a) Situation: The experimenter is carrying out a grammatical test in German (GER) with a
bilingual child who acquires German and Spanish (SP).

Experimenter genauGER / dieGER spielenGER gitarreGER dasGER istGER jaGER lustigGER / ‘right / they
play guitar it is funny’/
Child yoSP voySP aSP tocarSP laSP guitarraSP asíSP / ‘I will (to) play the guitar like that’/
(Olivia, 3;9,27)

(2b) Situation: The experimenter is carrying out a grammatical test in German (GER) with a
bilingual child who acquires German and French (FR).

Experimenter surFR maFR têteFR/ ‘on my head’/

Child surFR leFR baumstammGER / ‘on the trunk’ / (Antoine, 4;10,04)

A detailed analysis of these kinds of code-mixing will follow in the fourth section.

2.2. Language dominance

Language dominance describes the fact that one of the languages of a bilingual child is
mastered better than the other language or is acquired faster (for the two concepts of
language dominance cf. Schmeißer, Hager, Arnaus Gil, Jansen, Geveler, Eichler, Patuto &
Müller, 2016b). Research on bilingualism has also used the term of an unbalanced
language development (Müller et al., 2015, p. 45). In other words, an unbalanced bilingual
child has a strong and a weak language. Strong language refers to the (temporarily) domi-
nant language of an individual while the weak language is the (temporarily) less mastered
or less used language (Müller et al., 2015, p. 46). Let us not forget that language dominance
measures the pace and proficiency degree of language A in relation to language B and
simply assumes that language A and B are not in balance. The pace and proficiency
degree of the weak language in relation to that language in a monolingual child is
another issue (Cantone et al., 2008), not to be confounded with language dominance. It
should be mentioned that some researchers, like Bernardini & Schlyter (2004), have
argued against this distinction and assume that the weak language in a bilingual child
is also weak in a more general sense, i.e. in relation to monolingual peers (but see Gawlit-
zek-Maiwald & Tracy, 1996 for a different view).
Various measures to establish an (un)balanced language development have been pro-
posed (cf. Cantone et al., 2008 and Schmeißer et al., 2016b for an overview). Among the
criteria are the Mean Length of Utterance4, the Upper Bound5, which is based on the
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 5

child’s utterance with the highest number of words/morphemes/syllables, and the lexicon
size. Besides these qualitative criteria and in order to determine language dominance,
there are criteria which measure the speed of speech (measured in terms of the
number of utterances per minute as in Cantone et al., 2008) or the number of words
uttered per minute as in Müller et al., 2015, p. 67) or the speech fluency (number of hesita-
tions as in De Houwer, 1990).
Since the present study of trilingualism is a cross-sectional study, we decided to use the
lexicon size in the different languages of the child as a measure in order to determine
language (un)balance. We carried out the Peabody Picture Vocabulary Test (PPVT, Dunn,
1959; Dunn & Dunn, 1981, 1997), a test to measure the receptive vocabulary, since it is
a standardised test and available for most of the children’s languages.6
Besides, it is important to add Birdsong’s twofold conception (2014) of language do-
minance. According to this idea, domains and dimensions of language dominance have
to be set apart. Domains of language dominance refer to situations, needs, intentions
and social functions, whereas dimensions concern the linguistic competence, the pro-
duction and processing (Birdsong, 2014, p. 86). Hence, dimensions describe language
dominance on an individual level, one measure of which is the size of the lexicon as
measured in the PPVT. Since this test is suitable to measure the receptive vocabulary of
multilingual children, this allows us to make a statement about the linguistic development
at the individual learner level. In Birdsong’s terminology, the individual language develop-
ment of a bilingual (or in this case, a tri- and multilingual) is a relevant aspect when asking
for the language(s) that is/are better mastered or used more frequently by a multilingual
speaker. Sociolinguistic factors are also at play. This means that ‘[l]evels of fluency in a
language will depend on the need for that language and will be extremely domain
specific’ (Birdsong, 2014, p. 68). Therefore, if code-mixing depends on language domi-
nance, there is a need to specify whether domains and/or dimensions are meant.
In order to link the two previous subsections, namely in which way code-mixing and
language dominance correlate or not, we will devote the next section to a brief overview
of the existing research about code-mixing and language dominance in bilingual and tri-
lingual children. Before doing so, the concept of typological proximity has to be intro-
duced as another factor that possibly influences language selection made by
multilingual speakers.

2.3. Typological proximity

Language distance or typological proximity understood as membership to the same
language family – or lack thereof (Liceras & de la Fuente, 2015, p. 329) plays an important
role in second language learning. Since there are reasons to assume that this does not
only apply for successive language acquisition, but also for the simultaneous acquisition
of more than two languages, we intend to apply the concept of typological proximity
also to children who are tri- or multilingual from birth. Investigating L3 syntactic transfer
selectivity and typological determinacy, Rothman (2011) states that ‘[…] child bilinguals
(simultaneous, heritage bilinguals as well as child L2 learners) would be the only set of bilin-
guals of interest […] [since they] have two distinct syntactic systems that could theoretically
be transferred to the L3/Ln initial state’ (Rothman, 2011, p. 108). With regard to syntactic
transfer to the L3, he claims that it comes from either the L1 or the L2 and is conditioned
6 M. POESTE ET AL.

by typology/psychotypology (Rothman, 2011, p. 110). Therefore, transfer in L3 is not always

facilitative, as the Cumulative Enhancement Model (CEM) predicts, but ‘[…] syntactic prop-
erties of the closest (psycho)typological language, either the L1 or the L2, constitute the
initial hypothesis in multilingualism’ (Rothman, 2011, p. 112). Carrying over these results
to the phenomenon of code-mixing, it could be suggested, as an alternative to the
influence of language dominance, that typological proximity plays an important role regard-
ing code-mixing in multilingual children. Paralleling Rothman’s observations concerning L3
syntactic transfer, one could hypothesise that typological proximity between languages
influences the children’s code-mixing in terms of the selectivity of the language used
within mixed utterances. If language dominance does not explain why a child uses language
material from the L1 or the L2 in his/her L3 (in simultaneous trilingual children all languages
have to be labelled as L1s, thus L1a, L1b, L1c), typological proximity between the child’s
languages could be an explanatory factor. As we will see in the following sections, since
our data clearly shows that a causal relationship between language dominance and code-
mixing in bi-, tri- and multilingual children is not tenable, the influence of typology on
code-mixing will be studied as an alternative explanation.
However, Rothman (2011) also emphasises that (psycho)typology only plays the most
decisive role in transfer selection of one of the languages available to the speaker, pro-
vided a certain grouping of languages (Rothman, 2011, p. 113). It does not seem to be
clear which variables determine transfer if typology is not relevant in the particular com-
bination of languages, i.e. ‘where the L3 is equally typologically similar to the L1 and the L2;
or the L3 is typologically not at all similar to either the L1 or the L2’ (Rothman, 2011, p. 122).
Following this assumption, typological proximity can only be a decisive factor for mixing
languages if a relevant combination of languages is given. This will be taken into consider-
ation in our analysis of code-mixing as well.
Although typological proximity will be defined for our analysis as membership to the
same language family, it is noteworthy to mention that Liceras & de la Fuente (2015) differ-
entiate between typological similarity and typological proximity explaining that ‘[…] there
is typological similarity when a typological or formal universal is equally realised in these
two typologically-close languages; otherwise, we will talk about typological proximity’
(Liceras & de la Fuente, 2015, p. 333). The authors use null/overt subjects in French and
Spanish as an example. Although belonging both to the Romance language family,
which could be defined as typological proximity, there is no typological similarity if the
null subject parameter is taken into account since Spanish is classified as a [ + null
subject] language, whereas French is considered as a [-null subject] language (Liceras &
de la Fuente, 2015, p. 339). Therefore, the reason for this differentiation is that languages,
even though they belong to the same language family, may show differences in realising
different options of any given parameter (Liceras & de la Fuente, 2015, p. 335).
Another important term which has emerged from Rothman’s (2011) Typological
Primacy Model is that of ‘psychotypology’, ‘[…] the learner’s notions of the relations
between the L1 and the L2’ (Kellerman, 1983, p. 113). It is claimed that the way multilingual
speakers perceive the typological distance of the languages involved does also influence
transfer but that ‘[i]n many cases, psychotypological and actual typological proximity are
in fact one and the same’ (Rothman, 2011, p. 112). However, the concept of psychotypol-
ogy is not only difficult to define, but also impossible to determine in the case of very
young children because they do not possess the necessary meta-cognitive skills to
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 7

reflect on language as a system and to verbalise these reflections. Therefore, we will focus
in our analysis of code-mixing and typological proximity on typology in the sense of lin-
guistic typology.

3. Previous research about code-mixing and language dominance in

bilingual and trilingual children
After having presented the concepts of code-mixing and language dominance as well as
typological proximity as an alternative factor which might have an influence on code-
mixing, we now have a look at the relation between code-mixing and language domi-
nance as it has already been described in previous studies. Regarding the comparison
of bilingual and trilingual children, it has to be said that early child trilingualism is still
in its infancy (cf. Unsworth, 2013, p. 41f., Quay, 2001, p. 149, 2011a, 2011b; Hoffmann,
1999, p. 16, 2001, p. 13; Barnes, 2006, p. 28) and that the comparison between early trilin-
guals and bilinguals is completely unstudied (Barron-Hauwaert, 2000, p. 2). According to
Hoffmann (2000) amongst others there is a need to study early child trilingualism in its
own right since ‘[t]rilingualism is obviously placed somewhere between bilingualism
and multilingualism, but one should not assume it to be simply an extension of bilingual-
ism. It probably shares features with both, while at the same time retaining characteristics
of its own.’ (p. 84) Therefore, it is important to consider the phenomenon of trilingualism as
an independent area of research because the simple transfer of the results of the studies of
bilingualism to the field of trilingualism does not seem to be satisfactory. Nevertheless,
according to Hoffmann (2000), bilinguals and trilinguals share the ability ‘[…] to move
between different languages, switching, mixing and borrowing […]’ (p. 88) with the differ-
ence that trilinguals could involve three instead of two language systems.
Concerning bilingual children, some main observations will be described before com-
paring those results to the few existing studies about trilingual children and their mixing
behaviour. Bilingual children studied longitudinally (balanced as well as unbalanced) do
not show a high (intrasentential) mixing rate when observed in a monolingual mode
(Patuto et al., 2014, p. 197). The same can be observed for cross-sectional studies on bilin-
gual children, balanced and unbalanced ones (Müller et al., 2015, p. 104). Therefore, it can
be supposed that mixing languages intrasententially in early bilingualism is not directly
related to language dominance. However, intersentential mixing can be observed more
frequently in some bilingual children whereas intrasentential mixing is much less frequent
(Schmeißer et al. 2016a, p. 256f.). In accordance with Schmeißer, Eichler, Arnaus Gil, &
Müller (2016a), if one defines the use of the non-context language (the use of the language
which is not required by the linguistic context and not desired by the adult interacting
with the child) as intersentential mixing, this kind of mixing is related to language domi-
nance since mixing occurs in the children’s weak language.
The few existing studies of code-mixing in trilingual children are all longitudinal studies.
They show that especially intrasentential mixing is rather infrequent (Quay, 2001, 2008;
Chevalier, 2015; Hoffmann & Stavans, 2007), whereas intersentential mixing can often
be observed in the dominant language, mostly the community language (Sivakumar,
2017). Trilingual children are usually considered to develop one language as their domi-
nant language, which is generally the community language (Chevalier, 2015; Quay,
2001, 2008 for example). Since it is assumed that, generally speaking children with a
8 M. POESTE ET AL.

weak language mix their languages more frequently, it follows that trilingual children
should mix their languages more often than bilingual children, but a systematic study
which compares bilingual and trilingual children and takes into account language domi-
nance is still missing (with the exception of Mieszkowska, Luniewska, Kolak, & Kacprzak,
2017, a study which looks at the size of the receptive vocabulary in bilingual and trilingual
children between the age of 4;5 and 6;7). From the literature on bilingual children we may
deduce that dominance is not the determining factor for the frequency of intrasentential
mixing (Patuto et al., 2014) but it does matter if whole sentences, utterances or larger
pieces of discourse like turns are taken into account (Schmeißer et al., 2016a). Furthermore,
in Müller et al. (2015), it is claimed that, although it occurs rarely, it is possible that the non-
community language is dominant in the bilingual child (before the age of 5 which is the
time when the investigation in Müller et al. (2015) has stopped). In other words, it is poss-
ible that trilingual children may develop an unbalanced trilingualism with the non-com-
munity language(s) as dominant language(s). Since the studies on trilingual children are
scarce and they have not received much room in the previous literature, we will briefly
present the results of those studies which present quantifiable data.

3.1. Chevalier (2015)

One of the most detailed studies on trilingual language development is Chevalier (2015).
She investigates the language use of two trilingual children, focusing on contextual factors
that influence active trilingualism in early childhood. During the analysed period (2;1–3;1)
and with respect to the language use of the two trilingual children Lina (with a Swiss-
German mother, a Belgian father who speaks French and an American aunt in the
German part of Switzerland) and Elliot (with a Swiss-German father, an English-speaking
mother, attending French day-care in the French-speaking part of Switzerland), we can
observe that both children use an extremely small amount of intrasententially mixed utter-
ances: In both children, the mixing rate lies under the 5% line, concretely with a mixing
rate of 2% for Lina and 3% for Elliot. This result is similar to longitudinal studies of bilingual
children where Patuto et al. (2014, p. 197) report a mixing rate of 2.42% during a period
between 1;6 to 4. The same can be observed in the present cross-sectional study which
will be analysed and discussed in detail in section four. Concerning intersentential
mixing, Lina produces a relatively high number of intersentential mixing with 1144 utter-
ances, whereas Elliot nearly always sticks to the interlocutor’s language and there are only
210 cases of intersentential code-mixing to be observed. Figure 1 has been produced on
the basis of the numbers found in Chevalier (2015).
In order to relate the results of the children’s language production to their language
dominance in Chevalier (2015), the proficiency in production is measured on the basis
of the MLU and the Upper Bound. It is important to note that these are measures for
language dominance which rely on production, while we have measured language dom-
inance on the basis of the size of the receptive vocabulary in our cross-sectional research
study. Furthermore, it is important to add, with regard to the definition of language do-
minance in Chevalier (2015), that domain specific language dominance is extremely unli-
kely for two-year-olds because their sphere of activities is not sufficiently specialised yet.
However, according to De Houwer & Bornstein (2016), older children are possibly domi-
nant in different linguistic domains (Chevalier, 2015, p. 95). For the present study which
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 9

Figure 1. Lina’s and Elliot’s language use in the context of German, French and English (based on Che-
valier, 2015).

looks at children with a mean age of 58 months (4;8), Birdsong’s distinction between the
domains and dimensions of language dominance will be maintained and considered to be
relevant for the description of their language (un)balance. Returning to the trilingual chil-
dren Lina and Elliot, Lina is clearly most proficient in Swiss German and slightly more profi-
cient in English compared to French according to MLU and UB (Chevalier, 2015, p. 108),
whereas Elliot is most proficient in French and more proficient in English than in Swiss
German at the beginning of the study. At the end of the study, the order of English and
French is not clear anymore if MLU and UB are compared in the two languages. Relating
the information about language dominance and code-mixing and taking into account that
Chevalier (2015) defines a base language for code-mixing which is the context language,
directionality and quantity of mixing and language dominance go hand in hand with the
previous literature: Lina mainly uses her dominant language (Swiss German), not only with
her German mother, but also with her English aunt and her French father. This applies to
intersentential mixing (cf. also Schmeißer et al., 2016a where it is argued that these cases
are probably not instances of adult code-switching but instances of the child’s choice of
the wrong or non-desired language, where language choice is previously negotiated
between the adult and the child). Intrasentential mixing is extremely scarce and, since
the child is an unbalanced trilingual child, arguably not related to language dominance;
if there was a relation between the two, we would expect the reverse, namely that intra-
sentential code-mixing is high(est) in children with a language dominance and (much)
higher in the non-dominant language. Notwithstanding, there is no French or English
10 M. POESTE ET AL.

language material mixed intrasententially into Swiss German (her dominant language),
but there are cases of Swiss German words produced in the context of French or
English (Chevalier, 2015, p. 119). Unlike for Lina, there are also examples of intrasentential
code-mixing from Elliot’s non-dominant language in the context of his stronger languages.
Chevalier (2015, p. 125) explains the difference in the two children by the fact that the
language development in Lina is extremely unbalanced while Eliot’s trilingual language
development is more even, in the sense that it is perhaps exaggerated to speak of a domi-
nant language in Eliot’s case. Interestingly, Eliot’s MLU is rather low in Swiss German at the
end of the study which suggests that Eliot’s mixing behaviour is best described with respect
to different acquisition stages. As with Lina, intrasentential mixing is scarce. Notice that, in
contrast to Lina, this is also true for intersentential mixing since Eliot mostly sticks to the
language choice of each of his interlocutors (Chevalier, 2015, p. 99f.). This is so, although
Eliot’s father mixes nearly as frequently the languages (in other words he does not stick
to the one person – one language method) as Lina’s French father.
Taking the results of the two children together, Lina’s case seems to confirm what has
previously been observed by most researchers for bilingual children (if intersentential
mixing is concerned). For Elliot it can even be confirmed that not only intra- but also inter-
sentential mixing is scarce. Although he develops one language as his weak language, this
does not influence the direction of code-mixing.

3.2. Hoffmann and Stavans (2007)

Hoffmann and Stavans (2007) studied mixed utterances of two early trilingual children at
different times of their life. Looking at the amount of code-switching (CS) and code-
mixing (CM) over time, the results show that the production of CM increases as the chil-
dren grow older. It is noteworthy that the authors define CM as using language material
from different languages within a sentence and CS as mixing elements from two or
more languages across sentences (Hoffmann & Stavans, 2007, p. 57). In other words,
the difference between CM and CS corresponds to the distinction between intra- and
intersentential code-mixing respectively made in the present study. The increasing
number of CM (intrasentential mixing) is explained by the also increasing dominance
in the three languages since CMs ‘[…] require not only a deeper understanding of
the formal aspects such as lexicon, the morphology and the syntax in each language
but also the pragmatic and the functional appropriateness of these forms’ (Hoffmann
& Stavans, 2007, p. 61). The authors argue that the fewer switches produced by the chil-
dren at a younger age can be explained by the fact that the three language systems
have not been fully acquired yet. Growing older, the child’s contact with the language
systems increases and her/his general linguistic knowledge is more advanced (Hoffmann
& Stavans, 2007, p. 61). Nevertheless, this assumption seems to refer only to intrasenten-
tial mixing (i.e. CM) since the rate of intersentential mixing (i.e. CS) is higher at the first
time of investigation. This goes in line with the results obtained in Chevalier (2015),
namely that intrasentential mixing, in contrast to intersentential mixing, is scarce in
young trilingual speakers since it requires an advanced language competence.
However, CS decreases in both children over the time, as the two measurement times
have shown. In order to explain these observations, the authors hypothesise that ‘[…]
as children grow older, their dominance of the languages increases as evidenced by
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 11

the production of fewer switches that require less linguistic knowledge’ (CS) (Hoffmann
& Stavans, 2007, p. 61). In conclusion, the authors assume that sociolinguistic factors
weigh more than psycholinguistic or linguistic ones. Creating a model of developing
multilingual competence, the authors regard CS (intersentential mixing) as a basic multi-
lingual competence which is influenced by sociolinguistic factors and occurs at the level
of utterances or turns, while the mixing of smaller linguistic units is seen as an indicator
for an advanced multilingual competence (Hoffmann & Stavans, 2007, p. 70). Regarding
the relation between language dominance and mixing, which is the focus of the present
paper, it is shown that first there is no alternation away from English, but extensive intra-
and intersentential mixing into English which could be explained by the dominance of
English at that time (Hoffmann & Stavans, 2007, p. 62). Later alternations from and to all
three languages seem more balanced in the older subject which is explained by the
introduction of schooling which enhances trilingual language development and main-
tenance (Hoffmann & Stavans, 2007, p. 63). In this sense, the study does not indicate
whether language dominance causes code-mixing but suggests an influence of
language dominance on the directionality of mixing, as also observed in Chevalier
(2015).

3.3. Montanari (2010) and Quay (2008)

Comparing the above results with the studies by Montanari (2010) and Quay (2008), we
may deduce that age is a relevant factor when it comes to code-mixing in very young tri-
lingual children. Whereas there seems to be a difference between intra- and intersenten-
tial mixing in older children (in the sense that intrasentential mixing is rare while
intersentential mixing is relatively frequent and related to language dominance), very
young children may use intrasentential mixing relatively more frequently due to lexical
need. Montanari (2010) for example shows that Kathryn, a child raised in the U.S. with
the community her languages English by a Tagalog-speaking mother and a Spanish-
speaking father, mixes her languages rather frequently (intrasententially). However,
mixing is mostly due to lexical gaps so that children below the age of two possibly
have a restricted lexicon in all three languages and mix their languages for reasons of
lexical need (Montanari, 2010, p. 79). Quay (2008) states the same for the child Xiaoxiao
in Japan with an English-speaking father and a (Mandarin) Chinese-speaking mother.
However, she also makes clear that one fifth of a child’s input is enough to mainly use
English with her father and not mix more in her weakest languages, which is English. In
Quay (2001), the language use of the trilingual child Freddy, raised in Japan, is analysed
from age 1;1 to age 2;0. It is shown that the child uses many English words in the
English recording (and not in the German recording) and German words in the German
recording (and not in the English recording), although German, the father’s native
language, is the child’s weak language. Therefore, it has to be questioned if language dom-
inance always influences the direction of code-mixing.

3.4. Hypotheses
From the review of the literature in previous sections, we can state that intrasentential
mixing in bi- and trilingual children is rather infrequent (cf. Montanari, 2010; Quay,
12 M. POESTE ET AL.

2008). With respect to the causal relation between language dominance within the indi-
vidual and the use of code-mixing, previous studies do not seem to show clear-cut
results (for bilingual children and intersentential mixing cf. Schmeißer et al., 2016a, for tri-
lingual children and inter- as well as intrasentential mixing cf. Chevalier, 2015 and
Hoffmann & Stavans, 2007). Finally, the study of Quay (2008) seems to suggest a (possible)
influence of language dominance on the direction of code-mixing. Having summarised
these findings, we can now present the following hypothesis for our research study on
code-mixing in bi-, tri- and multilingual children:
Hypothesis 1: Intrasentential mixing in bi-, tri- and multilingual children will be as scarce as
already observed both longitudinally and cross-sectionally in previous studies.

Hypothesis 2: There is no causal relation between language dominance and code-mixing in

early bi-, tri- and multilingualism.

Hypothesis 3: Language dominance will not influence the directionality of code-mixing. We

would like to assume that the different results observed in the literature can be explained in
terms of Birdsong’s (2014) distinction between domains and dimensions of language domi-
nance, that is, by taking societal multilingualism into consideration, by differentiating
between intra- and intersentential code-mixing and by analyzing insertional and alternational
mixing separately.

After having presented the pertinent literature on code-mixing in bilingual and trilingual
children with a focus on language dominance and having pointed out the relevant
hypothesis, we will now proceed with the findings of our own study.

4. Our own research

4.1. Participants
In our study we examined in total 122 children7 with more than one language from birth in
order to describe their grammatical proficiency in their different L1s. There are 51 bilinguals,
62 trilinguals and 9 multilinguals with different language combinations (cf. Figure 3) and an
average age of 58 months (4;8). The bilingual group has an average age of 59 months (4;9),
the trilinguals 57 months (4;7), and the multilinguals 55 months (4;5). Therefore, the groups
differ in their mean age but the differences are not statistically significant.8 The age range is
from 28 months (2;4) for the youngest child (trilingual) to 127 months (10;5) for the oldest
one (bilingual). Figure 2 illustrates the age distribution in the sample. The total amount of
121 children is due to the fact that information about age is missing in one child.
Considering all children in our study, most of them are between 48 and 59 months
which means 4 years old (n = 38), almost the same number of children is between 36
and 47 months (n = 34), namely at an age of 3 years, and there are 28 children between
60 and 71 months, i.e. at an age of 5 years. There are only few children younger than 3
years (n = 5) or older than 5 years old (n = 19). Taking into account only the bilingual chil-
dren, the same age distribution can be observed, namely most of them are 4 years old (n =
19), fewer are 3 or 5 years old (n = 11 or n = 12, respectively) and a small number is younger
than 3 (n = 2) or older than 5 years old (n = 9). Regarding the tri- and multilingual children,
their age distribution differs in that most of the children are 3 years old. It is important to
mention that the group of multilinguals contains only 9 children which makes it difficult to
draw any significant conclusions.
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 13

Figure 2. Age distribution of all children in the sample and divided into the bi-, tri- and multilingual
groups.

The children who were attending kindergarten were tested in different cities in
Germany and in Palma de Mallorca (Spain) at the respective institution. The institutions
were bilingual with the community language (German in Germany or Spanish and
Catalan in the Balearic Islands) and one of the minority languages of the children. In
Figure 3, the different language combinations for the bi- and trilingual children are pre-
sented. The multilingual children had the following language combinations: German-
French-Spanish-English (1), German-French-Spanish-English-Arabic (1), German-French-
Spanish-Catalan (2) German-French-Spanish-Dutch (1), German-Spanish-English-Hebrew (1),
German-Spanish-Catalan-English (2) and French-Spanish Catalan-Galician (1).

4.2. Methodology
The study consisted of several grammatical tests in order to elicit the bi-, tri- and multilin-
gual children’s placement of subjects in French (Arnaus Gil & Müller, 2018b), adjective pla-
cement in Spanish and French (Arnaus Gil, Zimmermann, Tirado Espinosa, & Müller, 2019),
the placement of finite verbs in German (Arnaus Gil & Müller, 2018a) and the use of ser and
estar in Spanish and Catalan (Arnaus Gil, Jiménez Gaspar, & Müller, 2018; Kleineberg,
Arnaus Gil, & Müller, 2019). The tests were designed as production and comprehension
tasks including a picture story or performing an interactive game with the children.
Before starting the grammatical tests, the experimenter always talked a few minutes
with the child in order to determine/negotiate the language of the following test and
to ask the children implicitly to behave monolingually. Therefore, code-mixing could be
observed when, during the tests, the child used another language than the one deter-
mined at the beginning of the test phase. As can be seen, the aim of the study was not
to elicit code-mixing9 but to focus on the grammatical proficiency of bi-, tri- and multilin-
gual children. Therefore, the children were expected not only to use the respective
14 M. POESTE ET AL.

Figure 3. Language combinations: bilingual and trilingual children.

language of each grammatical test but also not to mix their languages. A deviation from
this behaviour can thus be analysed in terms of code-mixing. Despite the monolingual
setting, some relevant factors which might have influenced the appearance of code-
mixing can be therefore investigated.
All 122 children were investigated regarding their mixing behaviour in all linguistic con-
texts. The mean mixing rate is a result of correlating the observed code-mixing with the
entire number of utterances of all children. Notice, however, that the information on
the children’s language dominance was only available for 107 children.
As mentioned above, the children were implicitly asked to behave monolingually
during the tests, i.e. mixing languages was not desired. Initially, the amount of code-
mixing was studied in order to prove to what extent the tests worked. However, code-
mixing can also be analysed in its own right in order to find out about the influencing
factors, more precisely, under which circumstances code-mixing occurred despite the
monolingual setting. One of these influencing factors is language dominance, both its
domains and dimensions.

4.3. Language dominance

In order to determine the children’s language dominance, we measured language profi-
ciency with the PPVT (Peabody Picture Vocabulary Test, Dunn, 1959; Dunn & Dunn,
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 15

1981, 1997) from which the size of the receptive (hearing) vocabulary can be inferred. The
examiner presents a page with four pictures to the child. S/he produces a word and asks
the child to point at the picture that describes the word. The (in)correct response is noted
on a record form (Figure 4).
The French version of the test was designed by Dunn, Thériault-Whalen, & Dunn (1993)
and covers a broad range of French-language content words and syntactic categories
(adjectives, adverbs, verbs, nouns). The test is suitable for children between the age of
2;6 and 18;0. It consists of 170 items equally distributed across item-sets (1 item-set
equals 12 items). Administration should take 8 to 15 min. The French PPVT was designed
on the basis of a representative sample of 2038 children in Canada. The children came
from regions in Canada, where English is spoken in addition to French (especially
Québec and Ontario).
The PPVT is a standardised test designed for typically developing French-speaking chil-
dren to analyse their vocabulary development and in clinical assessment in order to detect
language impairments and school readiness. The test allows a comparison with children in
several age groups who speak French as their native language (control groups). In the
present case, it was used to measure the linguistic competence in the different L1s of chil-
dren who acquire more than one language from birth. To describe the linguistic compe-
tence a raw score is transformed into an age-dependent t-value which is linguistically
described as:
Extremely high “excellent” (IQ point above 130)

Moderately high “bon” (IQ-points 115–130)

High average “moyen” (IQ-points 100–115)

Figure 4. Administration of PPVT ([Link]

16 M. POESTE ET AL.

Low average “moyen” (IQ-points 85–100)

Moderately low “médiocre” (IQ-points 70–85)

Extremely low “faible” (IQ-points below 70)

The German version of the PPVT has been available since 2015. It was designed by
Lenhard, Lenhard, Segerer, & Suggate (2015). The German test covers a broad range of
German-language words (adjectives, verbs, nouns). Children between the age of 3;0 and
16;11 can be tested. The test consists of 228 items equally distributed across 19 item-
sets (1 item-set equals 12 items). Administration took 20 min. across all children. The
German PPVT was designed on the basis of a representative norming sample of 3555 chil-
dren in Germany, 29% of whom had one parent born outside Germany. The Spanish
version was designed by Dunn, Padilla, Lugo, & Dunn (1986). It consists of 125 items
and the norming sample consisted of 1219 children from Mexico and 1488 children
from Puerto Rico. In the case of the Catalan children, we used a translation of the
French and the Spanish version of the PPVT since no standardised test version was avail-
able.10 All other languages that the bi-, tri- and multilingual children in our study acquired
were not tested because the focus of the research project lay on the grammatical devel-
opment of the languages German, French, Spanish and Catalan.
In order to determine the language (un)balance of the tested children, a balanced tri-
lingual child had to be categorised in all her/his languages alike, low average to give an
example (Sivakumar, Sette, Müller, & Arnaus Gil, 2019). Remember that the term language
dominance expresses the fact that the languages are not in balance, in other words the
level of accuracy reached in the languages is not important; it simply matters whether
there was a difference between the languages of the speaker or not. The following
table 1 is taken from Sivakumar et al. (2019). It shows in more detail the different sub-
categories for the labels ‘balanced’ and ‘unbalanced’ in our data.
Once the test has been carried out, a raw number is obtained. The test allows us to
convert these raw numbers (which, unfortunately, differ from language to language)
into IQ-points, by which it is possible to compare the children’s results in the
different languages. The difference of IQ-points was taken to divide the bi-, tri- and
multilingual children into different groups with regard to their language dominance
or balance. Bilingual children can be balanced or unbalanced in their two languages.
Trilingual children can be balanced in their three languages, balanced with two
languages or balanced with two languages together with the possibility that there is
no PPVT available in the third language. With respect to an unbalanced language
development, the trilingual children can be unbalanced in three languages or with
only two languages, while the PPVT was not carried out in the third language. The chil-
dren with more than three languages in the present study, the multilingual children,
are balanced with two languages, while there was no PPVT carried out in the third
and fourth language, or unbalanced with two languages, while there are no results
for the PPVT in the other two languages. Furthermore, there is one multilingual child
balanced in all the four languages.
The following Figure 5 shows the status of (un)balance of the children tested in our
study. It distinguishes between balanced and unbalanced children, abstracting from the
subcategories mentioned in Table 1.
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 17

As can be seen, there are in general more balanced than unbalanced children (n
= 67 and n = 40, respectively). Taking into consideration only the children with data
available concerning their language dominance, 63% of the children are balanced
and 37% are unbalanced.11 For the different groups of bi-, tri- or multilingual chil-
dren a similar proportion of balance and unbalance can be observed. There are 24
balanced bilinguals (57%) and 18 unbalanced bilinguals (43%), 38 balanced trilin-
guals (67%) and 19 unbalanced trilinguals (33%), 5 balanced multilinguals (62,5%)
and 3 unbalanced multilinguals (37,5%). The balanced trilinguals in our study
were mainly tested in Germany and were balanced in the community language
and in one of their other L1s.12 Therefore, we can deduce from these observations
that children with more than two languages do not tend to develop the commu-
nity language as their dominant language, as claimed in the literature. It has to be
admitted that the children tested in this study are below the age of compulsory
education in Germany. Thus, it is possible that they have not yet developed a
language dominance with the community language, which usually comes after
the children start formal education (Wang, 2008). However, we would like to
point out that Sivakumar et al. (2019) showed that German as the community
language does not influence the results in the Peabody Picture Vocabulary Test.
Language proficiency measured by the receptive vocabulary is even slightly
better for the children acquiring German in a non German-speaking community
than for the ones living in Germany (Sivakumar et al., 2019, p. 26). Therefore, the
assumption that the community language tends to be the dominant language
cannot be confirmed, at least for children at kindergarten age. What has to be
proved in our following analyses is the proportion of (un)balance for those children
who mix their languages in the study in order to describe the relation between
code-mixing and language dominance.

Figure 5. Number of bilingual, trilingual and multilingual (un)balanced children.

18 M. POESTE ET AL.

Table 1. Categories to capture language (un)balance in bi-, tri- and multilingual children, taken from
Sivakumar et al. (2019).
BIL TRIL MULTI
Balanced balanced with 2 balanced with 2 languages (3rd lg. balanced with 2 languages (3rd and 4th
languages not measured) lg. not measured)
balanced with 3 languages balanced with 3 languages (4th lg. not
measured)
balanced with 4 languages
Unbalanced unbalanced with 2 unbalanced with 3 languages unbalanced with 2 languages (3rd and 4th
languages lg. not measured)
unbalanced with 2 languages (3rd lg. unbalanced with 4 languages
not measured)

4.4. General results

The first general result is that all children rarely mix their languages. This means that the
children are aware of the monolingual setting and use only in 134 out of 1913 cases (utter-
ances) a language other than the requested in the experimental situation. Inter- and intra-
sentential code-mixing is collapsed in the ‘entire’ column in the following Figure 6. This
corresponds to a mean mixing rate of 7%. The highest mixing rates can be observed by
tri- and multilingual children with 8.78% and 9.3%, respectively, while bilingual children
show an extremely low mean mixing rate of 4.32%. These percentages are calculated
with respect to the whole number of utterances of each group (bilinguals: 741, trilinguals:
957; multilinguals: 215). Of all 122 children, only 45 children mix their languages at all.13
Since the percentage values noted before are calculated for all children participating in
the study, it seems interesting to have a closer look at the proportion of mixing for the
children who code-mix. In other words, the children who code-mix (n = 45) produce 134
mixed utterances out of 792 realised items in the tests in total. This corresponds to a
mean mixing rate of 16.92%. Obviously, their proportion of mixing is higher than the

Figure 6. Amount of code-mixing (entire group).

 INTERNATIONAL JOURNAL OF MULTILINGUALISM 19

general rate of mixing. In other words, although only a few children code-mix at all, if they
do, they code-mix to a degree which cannot be neglected. Regarding the groups of bi-, tri-
and multilingual children they behave much alike with respect to their mixing rates. The
bilingual children who code-mix show a mean mixing rate of 16.24%, the trilinguals
17.46% and the multilinguals 17.54%. We will discuss the results qualitatively in sections
4.5 and 4.6.

4.4.1. Results in relation to societal language dominance

If we consider the children’s language combinations, a clear picture emerges with the chil-
dren who acquire Catalan. The mean mixing rate calculated for all children in the sample
decreases from 7% to 4.7%, once the testing in Catalan is excluded14, which means that
only 83 out of 1767 utterances are mixed. The trilingual’s mean mixing rate decreases
to 5.1% and the multilingual’s mean mixing rate to 4.44%. The percentage of code-
mixing within the bilingual children stays the same with 4.32% because there are no bilin-
gual children with the combination Catalan + X who were tested in both languages. There-
fore, the mean mixing rate is below 5% and classified as extremely low. Regarding the
number of languages of a child, the trilinguals show the highest mixing rate with 5.1%
and the bilinguals the lowest one with 4.32% (cf. Figure 7).15
Interestingly, the children who use Spanish in the Catalan testing did not mix their
languages when they had to speak Spanish. Furthermore, they mixed Spanish linguistic
material into Catalan (and not from their other languages). The reason for the relatively
high mixing rate in Catalan is argued to be related to the linguistic situation in Spain.
All Catalan speaking children were tested in Palma de Mallorca, a region that benefits
from societal bilingualism, a fact which might have fostered mixing. Boix-Fuster & Sanz
(2008) claim that societal bilingualism in Catalonia is asymmetrical in favour of Spanish
in most situations. If we can extend their results to the Balearic Islands, the children
might have assumed that all people who are able to speak Catalan are also able to

Figure 7. Amount of code-mixing excluding the testing in Catalan.

20 M. POESTE ET AL.

speak Spanish but not vice versa. According to this assumption, it seems appropriate to
apply the distinction between domains and dimensions of language dominance. Regard-
ing the Catalan children’s proficiency in the PPVT, the majority of the children who mix
Spanish into Catalan reached high scores (‘moderately high’, ‘extremely high’) in
Catalan. Therefore, the dimension of language dominance, that is, the linguistic compe-
tence within the individual, cannot explain the relatively high mixing rate in Catalan.
With respect to Birdsong’s (2014) concept of the domains of language dominance, the
results in Catalan indicate that Spanish is the preferred language in the school context
in Palma de Mallorca where the children were tested.
Apart from the possibility that language dominance in the sense of the domains of
language dominance might have fostered code-mixing in the Catalan tests, the
influence of the individual language dominance, in Birdsong’s conception, the dimension
of language dominance, will be statistically investigated in what follows.

4.4.2. Results in relation to individual language dominance

As mentioned before, we used the results of the PPVT to determine language dominance
within the individual. According to this classification, 21 out of 43 children who mix their
languages are classified as balanced, while the other 22 children are unbalanced. There-
fore, 48.84% of the children who produce code-mixing are balanced and 51.16% are unba-
lanced. Thus, it cannot be confirmed that children who mix their languages show an
unbalanced language development (Eichler, 2011). The reverse is also unlikely, namely
that unbalanced children mix more often their languages, as is generally assumed in
the research literature on bilingual children. In total, there are 77 children who do
not mix their languages, for 63 of them data concerning language dominance is avail-
able. These children are classified as balanced (63.49%) and as unbalanced (36.51%). In
other words, we could suppose that it is less likely for balanced children to mix their
languages since there are slightly more balanced than unbalanced children within
the group of children who do not mix their languages. However, there are more
balanced children (63%) than unbalanced ones (37%) in the whole sample (see
Figure 2). Put differently, the slightly higher number of balanced children in the
group of children who do not mix their languages could be linked to the overall
higher number of balanced children. In what follows, we will show on the basis of stat-
istical analyses that there is no relation between language dominance and code-mixing
in bi-, tri- and multilingual children.
Among the bilingual, trilingual and multilingual children, there were balanced as
well as unbalanced children (cf. Figures 8 and 9). Is there a relation between the children’s
code-mixing and language dominance within the bilingual, trilingual or multilingual
group?
A chi-squared test revealed no statistically significant relation assuming a random prob-
ability under 5% (X2 = 2,751, df(1), p = 0,97). Notice, however, that the tri- and multilingual
children have been combined in one group in order to obtain a reliable result since the
multilingual group consisted of less than 5 individuals.16 The same result is obtained if
the children who mix their languages in the Catalan tests are excluded: There is no stat-
istically significant difference between the bilinguals and the children with more
than two languages regarding code mixing and language dominance (X2 = 1.551, df(1),
p = 0.213). Nevertheless, it is important to analyse not only the relation between
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 21

Figure 8. Number of bilingual, trilingual and multilingual (un)balanced children who mix their
languages (entire group).

code-mixing and language dominance on the one hand and the relation between code-
mixing and the number of languages acquired by the children on the other hand but also
to compare the mean proportion of code-mixing between groups (bi-, tri- or multilinguals)
split by language dominance. Therefore, we carried out a two-way ANOVA to check
whether there is a significant interaction between language dominance and language

Figure 9. Number of bilingual, trilingual and multilingual (un)balanced children who mix their
languages (without Catalan tests).
22 M. POESTE ET AL.

group with respect to the amount of code-mixing. The percentages of code-mixing were
considered as the dependent variable whereas language dominance and language group
were treated as independent variables, also called factors. The statistical analysis revealed
no significance, in other words, the amount of code-mixing is unrelated to language dom-
inance and to the number of languages known (F(4,37) = 0.677, p = 0.514).17
Furthermore, we can add information on the family attitudes towards code-switching
since these may vary largely and probably influence the children’s amount of mixing. The
attitudes towards CS may differ for example in allowing, encouraging, discouraging or
banning CS in the participating families. For this reason, we use the results of the input ques-
tionnaires which were given to the parents of the children.18 These questionnaires are avail-
able for 57 of 122 children and contain – among other things – information about the
language/s spoken in the family and about the language/s the mother and the father
respectively speak with their child. Out of these 57 questionnaires, there are 19 from children
who produce mixed utterances and 38 from children who do not mix their languages. With
regard to the children who code-mix, there are only six parents (32%) who indicate having
only one family language whereas 13 parents (68%) indicate having more than one family
language. In contrast, the relation is in balance if it comes to the children who do not
code-mix. 19 of the families (50%) indicate having one family language and 19 of the families
(50%) indicate having more than one family language. Thus, having more than one family
language does not cause code-mixing but children who mix their languages do often
speak more than one language in their family communication. Furthermore, for seven of
the children who code-mix (37%) the parents indicate that that each parent speaks only
one language with the child. Regarding the parental questionnaires of the children who
do not produce mixed utterances, it can be observed that 22 parents (58%) indicate that
each parent speaks only one language with the child. Therefore, the situation that each
parent only speaks one language with the child is less frequent for children who mix their
languages. Due to the scarce amount of data, we may only deduce very carefully that
having more than one family language and addressing the child in more than one language
in adult–child interaction demonstrates a more tolerant handling of code-mixing on the part
of the parents and thus in the child´s input. A liberal attitude towards code-mixing on part of
the parents will possibly give rise to a liberal use on part of the child as well. The following
figures illustrate the described information from the parental questionnaires for the children
who code-mix (Figure 10) and for the children who do not mix their languages (Figure 11).
Until now, we focussed on the children who mix their languages. In what follows, we will
investigate the relation between code-mixing and language dominance on the basis of the
134 instances of inter- and intrasentential code-mixing. Information regarding the chil-
dren’s language dominance is available in 129 cases. The fact that 43 children produce
134 mixed utterances shows that several children change their languages more than
one time between or within utterances. An independent two-sample t-test reveals that
there is no statistically significant19 difference between the amount of code-mixing that
was produced by a balanced child and the amount of code-mixing that was produced
by an unbalanced child (t = 1,162, df(39), p = 0,252). 72 out of 114 instances of code-
mixing can be assigned to balanced children and 57 to unbalanced ones. The difference
between code-mixing produced by balanced and by unbalanced children does not
reach statistical significance, not even if the Catalan tests are excluded (t = 0,861, df(19),
p = 0,4). Therefore, language dominance does not seem to determine code-mixing.20
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 23

Figure 10. Information about the family languages of the children who mix their languages.

4.5. Analysis of code-mixing: intersentential

Our review of the literature has suggested a difference between inter- and intrasentential
mixing. Until now, we have collapsed the two. In what follows, we will study inter- and
intrasentential mixing separately.
111 out of 134 instances of code-mixing can be classified as inter-sentential mixing –
106 with data available regarding language dominance. Hence, there is a clear imbalance
in favour of intersentential mixing. This result coincides with the results in Schmeißer et al.

Figure 11. Information about the family languages of the children who do not mix their languages.
24 M. POESTE ET AL.

(2016a, p. 261) according to which intersentential mixing is generally more frequent than
intrasentential mixing.
57 out of 106 instances of code-mixing were produced by balanced children and 49 by
unbalanced ones. An independent two-sample t-test reveals that there is no statistically
significant difference among the groups (t = 0,118, df(32), p = 0,907). In other words,
language dominance and intersentential mixing are unrelated. 58 instead of 106 instances
of code-mixing are observed if we exclude the Catalan tests, due to the reasons already
mentioned above. The difference between code-mixing produced by balanced (n = 26)
and by unbalanced children (n = 32) is not statistically significant either (t = −0,393, df
(16), 0,699).21 This result stands in sharp contrast to Schmeißer et al. (2016a) who observed
mainly intersentential mixing in the weak languages of four (German-French) bilingual
children. Notice, however, that the children of the study by Schmeißer et al. (2016a)
were younger than the children of the present study and they were also analysed
longitudinally.

4.6. Analysis of code-mixing: intrasentential

On the basis of 23 cases of intrasentential code-mixing out of 1913 utterances in total, we
may only outline some tendencies concerning intrasentential mixing in multilingual
children.
While 13 cases of code-mixing were produced by seven balanced children, ten mixed
utterances can be observed in eight unbalanced children. This balance according to
language dominance and code-mixing is confirmed by an independent two-sample t-
test which reveals no statistically significant difference (t = 1,245, df(8), p = 0,248). The
same results are obtained if we exclude the Catalan tests. The absolute number of intra-
sentential code-mixing produced by balanced children is eleven, whereas the number
of code-mixing produced by unbalanced children is nine. The difference is not statistically
significant (t = 1,892, df(4), p = 0,146). Although the number of intrasentential mixing is
rather small, it can be deduced that language dominance is not a determining factor
when it comes to mixing languages within one utterance.

5. An alternative analysis
From the literature on (successive) third language acquisition (Rothman, 2011, 2015), we
know that (psycho-) typology is an important concept in order to explain cross-linguistic
influence. With regard to intersentential mixing in the analysed children, in 51 cases the
language used is typologically closer to the context language of the test (cf. Figure 12).
As already mentioned, ‘typological proximity is understood as membership to the same
language family’ (Liceras & de la Fuente, 2015, p. 329). For the present dataset, we can
mainly distinguish between the Germanic and the Romance language family. Further-
more, Spanish and Catalan can be seen as typologically closer to one another than
French (for example, the first two are null-subject languages whereas the latter is not
pro-drop). In 51 cases of intersentential mixing, the typologically closer language is
involved. This can be explained, among other things, by the higher number of Spanish
utterances in the Catalan tests. Only in eleven cases does the switch come from the typo-
logically more distant language, i.e. from a language of a different language family,
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 25

Figure 12. Directionality of intersentential mixing in relation to typological proximity.

although the child would have had access to a language from the same family as the
context language. An independent two-sample t-test reveals that the difference is statisti-
cally significant (t = 3,13, df(13), p = 0,008). In other words, the proximity between the
languages of tri- and multilingual children influences the direction of intersentential
mixing. There are 23 cases of intersentential mixing for which it cannot be decided
which language is the typologically close one. For 21 of these 23 cases, the language of
the mixed utterance is the community language. Put differently, if there is no typologically
closer language which can clearly be identified, the children seem to use the language
spoken in their community. Concerning the bilinguals’ 26 cases of intersentential code-
mixing, nothing can be said with respect to typological proximity since the bilingual
child has no choice, i.e. s/he cannot decide between two languages LB or LC to mix in LA.
Let us turn to the children’s intrasentential code-mixing. While in ten cases, the typolo-
gically closer language is used, in seven cases it is the typologically more distant one.
Nothing can be said with respect to the six bilingual children. We are therefore led to
believe that typological proximity plays no role in intrasentential code-mixing. As reported
earlier, an influence of typological proximity on intersentential code-mixing could be
proved. Having a closer look at the different processes of intrasentential mixing, namely
insertion and alternation, the former involves disintegrated language material and is,
therefore, more similar to intersentential mixing than the process of alternation. Conse-
quently, we could assume that, just like with intersentential mixing, instances of intrasen-
tential code-mixing which can be classified as insertion are related to typological proximity
while the ‘integrated’ mixing (alternation) is not.
The introduction of language material from a bilingual’s respective other language or
from one of the other languages of a multilingual speaker is one characteristic of the
process of insertion. Insertion denotes a change between languages introducing language
material of language A into language B (Muysken, 1997, p. 363). Insertion always needs a
base language into which material from another language is inserted. In contrast,
26 M. POESTE ET AL.

alternation means the change between languages that comprises grammar as well as the
lexicon of the two (or more) languages (Muysken, 1997, p. 362f.).
Based on the assumption that CS-processes can be studied empirically, Muysken pro-
poses different criteria to determine the processes underlying mixed utterances. The diag-
nostic criteria constituency, peripherality, nestedness, selectivity, linear equivalence,
length and complexity, bidirectionality, embedding in discourse, structural position,
flagging, category and adaptation are shown in Table 2.22 Besides, Table 2 includes all
three CS-processes (insertion, alternation and congruent lexicalisation). The plus sign
means that the criterion is an indicator of the particular CS process.23
We will illustrate the application of some of the features of Table 2 to intrasentential
code-mixing by using example (3) from our study.

(3) äh dieGER vispaSP ( = avispaSP) / ´er the wasp´

Since this instance of intrasentential mixing occurs during testing in German, we have to
consider ‘vispa’ as the element mixed into an otherwise German utterance. According to
the features which determine the CS processes, it can be classified as a single constituent,
indicating therefore insertion as the underlying process. Furthermore, the mixed element
is not situated in a peripheral position. Since a peripheral position indicates alternation, the
non-occurrence of this feature results in the respective other CS process, namely insertion.
With regard to nestedness, no result is obtained because there is no element following the
mixed element ´vispa but it can be classified as ‘selected element’ which is indicative of
insertion. However, the linear equivalence between the German and Spanish structure
of a noun phrase shows characteristics of alternation. With respect to features 6 and 7,
the mixed element is neither considered as long (utterances with more than 3 words
are defined as long) nor as complex since complexity is given if the mixed element is a
subordinate clause. Musyken describes the feature of embedding in discourse as
another possibility to indicate a type of switching. ‘Consider a mixed clause starting in
language A and, ending in language B. If the preceding utterance is in A, and the following
clause is in B, alternation is a plausible analysis’ (Muysken, 1997, p. 371). Due to the fact

Table 2. Diagnostic features of the three processes of code-switching (Modified according to Muysken,
1997, p. 373).
Insertion Alternation Congruent lexicalisation
⎫
⎬ single +
1. number of constituents several +
⎭
non +
2. peripheral  +
3. nestedness nested a b a +
non-nested a b a + +
4. selected element + +
5. linear equivalence + +
6. length +
7. complexity +
8. embedding in discourse +
9. major clause boundary +
10. flagging ⎫ +
⎬ lexical category +
11. type of category function word +
⎭
adverb, conjunction +
12. adaptation + +
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 27

that in the example there is no utterance of the child which either precedes or follows the
mixed one, the negation of this feature makes insertion more plausible. Mixing within a
syntactic phrase, like it is observed in example (3), indicates insertion, while the lack of
flagging before the switch, given by a pause or a particle such as ´euh´, makes insertion
less likely. Concerning the type of category, the mixed element is a noun and thus a
lexical category which indicates insertion whereas absence of adaptation of the mixed
element to the base language of the utterances is a sign for alternation. In conclusion,
nine features are indicative of the process of insertion and only two of the process of alter-
nation. For one feature no result is obtained.
After having presented one example, we will now describe the results of the analysis for
all instances of intrasentential code-mixing in the present study. As can be seen in Figure
13 below, after having applied Muysken’s classification to the children’s productions of
intrasentential mixing, there are two cases in which seven features indicate insertion,
two cases in which eight features point to the process of insertion, 17 cases in which
nine features correspond to insertion and one case in which eleven features indicate
insertion.
If we postulate that insertion is present, if more than half of the features ( = 7) are in
favour of this CS-process, then 23 out of 23 cases of intrasentential code-mixing can be
classified as such. An important result of this analysis is the fact that the processes of inser-
tion and alternation constitute a continuum. This result is described in Poeste (2017).
Muysken himself characterises the features used to distinguish the CS-processes as fea-
tures which make a certain process more or less likely (Muysken, 1997, p. 365ff.).
In the present dataset, 13 cases of intrasentential code-mixing classified as insertion are
produced by balanced children and five by unbalanced children. For four children, the
relation cannot be investigated since information regarding language dominance is
missing. The data is too scarce for further analysis. Notwithstanding, we believe that the

Figure 13. Assignment of intrasentential mixing to CS-processes by means of the features of insertion.
28 M. POESTE ET AL.

result that all cases of intrasentential code-mixing are cases of insertion is interesting as
such and should be pursued in future research.

6. Discussion
A central result of our study is that code-mixing is rare in a monolingual setting (mean
mixing rate of 7.27% and 4.81% when excluding Catalan tests). Therefore, not only bilin-
gual but also trilingual and multilingual children are able to behave monolingually in a
monolingual setting. Furthermore, the assumption formulated in the literature that an
unbalanced language development exercises an influence on code-mixing can only
be partly confirmed. Considering the distinction between domains and dimensions of
language dominance (Birdsong, 2014), we can presume domain-specific language do-
minance in the children who speak Catalan. Domain-specific language dominance
could explain the relatively high mixing rate in the Catalan tests. However, if it comes
to the dimensions of language dominance, i.e. the relation of the languages measured
in the individual, no relation was found between the children’s code-mixing
and language (un)balance. The result remains if a subdivision is made between inter-
and intrasentential mixing. Consequently, an unbalanced language development does
not cause the appearance of code-mixing. The study of intersentential code-mixing
has suggested an influence of the child’s typologically close language. With respect to
intrasentential mixing, all instances could be identified as instances of insertional
code-mixing.
How can we account for the relevance of typological proximity and the process of inser-
tion? It seems plausible to assume that it is easier to process typologically related
languages, due to their structural similarities, than typologically distant languages. Not-
withstanding, it is also probably more difficult to keep typologically closer languages
apart, or, to speak with Grosjean’s (2001) terminology, to always behave in a monolingual
mode and suppressing the non-desired language. Furthermore, since insertion always
demands a base language (a feature which insertion shares with borrowing, a process
which can also occur in otherwise monolinguals), it is arguably less complex than alterna-
tion. But here are exactly the limits of our research study. We do not know why the children
code-mixed at all, nor do we know how they behave if a multilingual setting is conveyed
(but see Patuto et al., 2014).
Our results stand in sharp contrast with existing studies in the literature on trilingual
and bilingual children with regard to the relevance of language dominance. We believe
that this is the case because we started to apply the distinction between domains and
dimensions of language dominance to our dataset. We propose to consider societal multi-
lingualism as a factor that influences the children’s mixing behaviour as well as a distinc-
tion between the different CS-processes since it makes a difference mixing integrated or
disintegrated language material. Possibly, the relation between language dominance and
code-mixing differs depending on the level of complexity of mixing. In the case of early
child trilingualism, all existing information comes from longitudinal studies. Hopefully,
future research will include a fine-grained conception of language dominance, a shift in
the methodology and multilingual children who acquire more than two languages and,
therefore, have a choice!
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 29

Notes
1. Abbreviated as CS.
2. Years;months,days.
3. A detailed description of the participants in our study will follow in section 4.1.
4. Abbreviated as MLU.
5. Abbreviated as UB.
6. Although language dominance is generally determined by measuring MLU or Upper Bound,
we decided to use a measure of receptive vocabulary (size of receptive vocabulary, as
measured by PPVT) for the following reasons: First, the children’s results in the Peabody
Picture Vocabulary Test describe not only language performance but also language compe-
tence which is not clear when using the MLU to determine language dominance (De
Houwer, 1990, p. 15). Second, it was not possible to measure the MLU for all children
because at least 100 utterances in each of the children’s languages would be needed. This
would have required a spontaneous recording of around 20 min. per language.
7. As a matter of fact, the study counts 126 children but there are no transcriptions available for
three of them and one child was not predisposed to speak. The latter was also the youngest
with 22 months (1;8). Therefore, in what follows, the total number of children will be 122.
8. An independent two-sample t-test reveals that there is no statistically significant difference
between the mean age of the bilingual and the trilingual group assuming a random pro-
bability under 5%. (t = 0,85, df(102), p = 0,398). There is also no statistically significant differ-
ence between the mean age of the trilingual and the multilingual group (t = 0,23, df(9), p =
0,82) and between the bilingual and the multilingual group (t = 0,81, df(12), p = 0,433).
9. It is important to note that code-mixing is rather difficult to elicit. In cross-sectional studies
bilingual children mix rarely their languages if addressed in one of their languages (mean
mixing rate of 0.94% in Müller et al., 2015, p. 109). Patuto et al. (2014) show that mixing is
difficult, if not impossible to elicit in production, since it is generally assumed that bilingual
speakers mix their languages spontaneously. In the study of Patuto et al. (2014), bilingual chil-
dren had to repeat a sentence with code-mixed material but, although they were able to
repeat the sentence, they ‘corrected’ the mixed material and produced the sentence with
lexical material from one of their languages. When repeating the sentence, they tended to
choose the language of the last word of the adults’ utterance.
10. Admittedly, it is problematic to use a translation of the French and the Spanish PPVT in order
to examine the children’s vocabulary size in Catalan. Such practice may result in obtaining
items which are close in meaning but do not necessarily correspond in respect to other cri-
teria, e.g. structural word complexity, cultural interpretation, familiarity, or frequency of occur-
rence (Peña, 2007). Nevertheless, looking at the French and Spanish PPVT in detail, even
though the norming sample and the total number of test items differ, the distribution of
the standard scores and the linguistic categories are the same (cf. Dunn et al., 1986; p. 40
and Dunn et al., 1993, p. 37). Since no other solution was available at the time of testing,
the Catalan translation of the French and the Spanish PPVT were thus implemented in
order to measure the Catalan receptive vocabulary.
11. The children’s language dominance is sometimes unknown due to two reasons. First, for some
children there is no PPVT available. Second, difficulties arise when comparing the results of the
PPVT and the results of the Wechsler Intelligence Test, which was used during a time of testing
when the German version of the PPVT was still not available (Petermann, 20112). There are
children who were too young or too old, making it impossible to convert the age-based
scaled scores (with a mean of 10) of the Wechsler Intelligence Test into IQ-points (Lenhard
et al., 2015), a necessary step in order to determine the children’s language dominance.
12. This result corresponds to 36 of the 38 balanced trilingual children (the other minority
language, namely English, was not tested). Notice that only two of the 38 balanced trilingual
children were tested in all three languages and the latter show a balanced relation among
their three L1s. More research is needed which measures all three languages in the trilingual
children.
30 M. POESTE ET AL.

13. This is a relevant fact. From the 45 children who code-mixed, we will only refer to a total
amount of 43 since there is data available on language dominance for only these 43 children.
14. There were no bilingual Catalan-Spanish children tested in both languages. Therefore, we do
not know whether there is a difference in the mixing behaviour of bilingual and trilingual chil-
dren. We carried out two quantitative analyses, one including the Catalan tests and one
excluding them, in order to find out whether societal bilingualism has an impact on the
mixing behaviour of the tri- and multilingual children.
15. The illustration of the data in figure 7 parallels the one in figure 6.
16. We are aware of the fact that the group of children who mixed their languages is relatively
small. Nevertheless, this shows that the children behaved according to the monolingual
setting as outlined in the study.
17. It is noteworthy that the statistical analysis was carried out with the percentages of mixing for
the corresponding children who code-mix and not with the absolute number of code-mixing
of each child.
18. For a detailed analysis of the parental questionnaires see Arnaus Gil, Müller, Sette & Hüppop
(2019).
19. Statistical relevance is given by a random probability under 5%.
20. In addition to measuring language dominance based on the receptive vocabulary of the child,
a reviewer suggested exploring the correlation between the children’s general verbosity and
the proportion of CS in future research. In this respect, it is possible to assume that children
who talk more (e.g. produce longer speech samples) also code-switch more.
21. The statistical programme calculated a negative value for t because the mean value of the first
group (balanced)) was lower (2,17) than the one of the second group (unbalanced, 2,46).
22. Regarding the criteria of constituency, Muysken (1997) proposes that when the switched
element is a single, well-defined constituent it is likely that insertion is the underlying
process. For peripherality, he claims that a mixed element at the periphery of an utterance indi-
cates the process of alternation. The term nestedness means that there exists a structural
relation between the mixed element and the preceding and following elements in the respect-
ive other language. Given this structural relation, insertion is probable. Concerning selectivity, it
can be said that if the mixed element is selected by an element in the other language, insertion
seems to be the underlying process. Linear equivalence between the languages involved, for
example the same word order, indicates alternation or congruent lexicalisation. With regard to
length and complexity, Muysken suggests that the more words a mixed fragment contains, the
more likely it is alternation. The same applies for complexity. Bidirectionality refers to the fact
that no matrix language can be determined within a mixed utterance indicating, therefore, a
shared structure. This criterion might help to identify the process of congruent lexicalisation.
The way the mixed utterance is embedded in discourse can also be indicative of the CS
process. This is explained in more detail in the analysis of example (3). Looking at the structural
position of the mixed element, a switch at a major clause boundary has to be distinguished
from one internal to a phrase. The latter is more likely when talking about insertion or con-
gruent lexicalisation. If a change between two languages is flagged by a pause or a particle
like ‘euh’ alternation between codes seems more probable. With single mixed words it is
also important to consider their lexical category. Whereas content words are likely to be inser-
tions, mixing discourse particles and adverbs may indicate the process of alternation. Adapting
means that a mixed fragment is adapted morphologically or syntactically to the other
language. Given this modification, insertion or congruent lexicalisation is more likely.
23. The negation of features 2, 4–10 and 12 results in the respective other CS process, for example
selected element = insertion/congruent lexicalisation and non selected element = alternation
(feature 4).

Disclosure statement
No potential conflict of interest was reported by the authors.
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 31

Funding
This work was supported by Deutsche Forschungsgemeinschaft: [grant number 232285006 (Laia
Arnaus Gil and Natascha Müller)].

References
Arnaus Gil, L., Zimmermann, K., Tirado Espinosa, M., & Müller, N. (2019). The acquisition of French
adjective placement in monolingual, bilingual, trilingual and multilingual children: A robust syn-
tactic domain. Manuscript, Bergische Universität Wuppertal.
Arnaus Gil, L., Jiménez Gaspar, A., & Müller, N. (2018). The acquisition of Spanish SER and ESTAR in
bilingual and trilingual children: Delay and enhancement. In A. Cuza & P. Guijarro-Fuentes
(Eds.), Language acquisition and contact in the Iberian Peninsula (pp. 91–124). Berlin: de Gruyter.
Arnaus Gil, L., & Müller, N. (2018a). Acceleration and delay in bilingual, trilingual and multilingual
German-Romance children: Finite verb placement in German. Linguistic Approaches to
Bilingualism, 1–29. Published online. doi:10.1075/[Link]
Arnaus Gil, L., & Müller, N. (2018b). French postverbal subjects: A comparison of monolingual, bilin-
gual, trilingual and multilingual French. Languages, 3(3), 1–28. doi:10.3390/languages3030029
Arnaus Gil, L., Müller, N., Sette, N., & Hüppop, M. (2019). Active bi- and trilingualism and its influencing
factors. In International Multilingual Research Journal. Submitted.
Barnes, J. (2006). Early trilingualism. A focus on questions. Clevedon: Multilingual Matters.
Barron-Hauwaert, S. (2000). Issues surrounding trilingual families: Children with simultaneous
exposure to three languages. Zeitschrift für Interkulturellen Fremdsprachenunterricht, 5(1), 1–13.
Bernardini, P., & Schlyter, S. (2004). Growing syntactic structure and code-mixing in the weaker
language: The ivy hypothesis. Bilingualism: Language and Cognition, 7(1), 49–69.
Birdsong, D. (2014). Dominance in bilingualism: Foundations of measurement, with insights from the
study of handedness. In C. Silva-Corvalán & J. Treffers-Daller (Eds.), Language dominance in bilinguals:
Issues of operationalization and measurement (pp. 85–105). Cambridge: Cambridge University Press.
Boix-Fuster, E., & Sanz, C. (2008). Language and identity in Catalonia. In M. Niño-Murcia (Ed.),
Bilingualism and Identity: Spanish at the crossroads with other languages (pp. 87–106). Amsterdam:
Benjamins.
Cantone, K. F. (2007). Code-switching in bilingual children. Dordrecht: Springer.
Cantone, K. F., Kupisch, T., Müller, N., & Schmitz, K. (2008). Rethinking language dominance in bilin-
gual children. Linguistische Berichte, 215, 307–343.
Cantone, K. F., & Müller, N. (2005). Code-switching at the interface of language-specific lexicons and
the computational system. International Journal of Bilingualism, 9(2), 205–225.
Chan, B. (2004). Beyond “contextualization”: Code-switching as a textualization cue. Journal of
Language and Social Psychology, 23(1), 7–27.
Chevalier, S. (2015). Trilingual language acquisition. Contextual factors influencing active trilingualism
in early childhood. Amsterdam/Philadelphia: Benjamins.
De Houwer, A. (1990). The acquisition of two languages from birth: A case study. Cambridge:
Cambridge University Press.
De Houwer, A., & Bornstein, M. (2016). Balance patterns in early bilingual acquisition: A longitudinal
study of word comprehension and production. In C. Silva-Corvalán & J. Treffers-Daller (Eds.),
Language dominance in bilinguals: Issues of operationalization and measurement (pp. 134–155).
Cambridge: Cambridge University Press.
Dunn, L. M. (1959). Peabody picture vocabulary test. Circle Pines, MN: American Guidance Service.
Dunn, L. M., & Dunn, L. M. (1981). Peabody picture vocabulary test – revised. Circle Pines, MN: American
Guidance Service.
Dunn, L. M., & Dunn, L. M. (1997). Peabody picture vocabulary test (3rd ed.). Circle Pines, MN: American
Guidance Service.
Dunn, L. M., Padilla, E. R., Lugo, D. E., & Dunn, L. M. (1986). Test de Vocabulario en Imágenes Peabody.
Adaptación Hispanoamericana. Minneapolis, MN: Pearson.
32 M. POESTE ET AL.

Dunn, L. M., Thériault-Whalen, C. M., & Dunn, L. M. (1993). Échelle de vocabulaire en images Peabody.
Toronto: Pearson.
Eichler, N. (2011). Code-Switching bei bilingual aufwachsenden Kindern. Eine Analyse der gemischtspra-
chlichen Nominalphrasen unter besonderer Berücksichtigung des Genus. Tübingen: Narr. (Tübinger
Beiträge zur Linguistik 528).
Gawlitzek-Maiwald, I., & Tracy, R. (1996). Bilingual Bootstrapping. Linguistics, 34, 901–926.
Grosjean, F. (2001). The bilinguaĺs language modes. In J. Nicol (Ed.), One mind, two languages:
Bilingual language processing (pp. 1–22). Oxford: Blackwell.
Gumperz, J. J. (1967). On the linguistic markers of bilingual communication. The Journal of Social
Issues, 23(2), 137–153.
Hoffmann, C. (1999). Trilingual competence: Linguistic and cognitive issues. Applied Linguistic Studies
in Central Europe, 3, 16–26.
Hoffmann, C. (2000). Bilingual and trilingual competence: Problems of description and differen-
tiation. Estudios de Sociolingüística, 1, 83–92.
Hoffmann, C. (2001). Towards a description of trilingual competence. The International Journal of
Bilingualism, 5(1), 1–17.
Hoffmann, C., & Stavans, A. (2007). The evolution of trilingual codeswitching from infancy to school
age: The shaping of trilingual competence through dynamic language dominance. International
Journal of Bilingualism, 11(1), 55–72.
Jansen, V., Müller, J., & Müller, N. (2012). Code-switching between an OV and a VO language: Evidence
from German-Italian, German-French and German-Spanish children. Linguistic Approaches to
Bilingualism, 2(4), 337–378.
Jisa, H. (2000). Language mixing in the weak language: Evidence from two children. Journal of
Pragmatics, 32, 1363–1386.
Kellerman, E. (1983). Now you see it, now you don’t. In S. Gass & L. Selinker (Eds.), Language transfer in
language learning (pp. 112–134). Rowley, MA: Newbury House.
Kleineberg, D., Arnaus Gil, L., & Müller, N. (2019). The acquisition of the Spanish copula system in
mono-, bi- and trilingual children. Submitted.
Lenhard, A., Lenhard, W., Segerer, R., & Suggate, S. (2015). Peabody picture vocabulary test (4th ed.).
Frankfurt am Main: Pearson.
Liceras, J., & de la Fuente, A. A. (2015). Typological proximity in L2 acquisition. The Spanish non-native
grammar of French speakers. In T. Judy, & S. Perpiñán (Eds.), The acquisition of Spanish in under-
studied language pairings (pp. 329–358). Amsterdam: Benjamins.
MacSwan, J. (2000). The architecture of the bilingual language faculty: Evidence from codeswitching.
Bilingualism: Language and Cognition, 3(1), 37–54.
Meisel, J. M. (1994). Code-switching in young bilingual children. The acquisition of grammatical con-
straints. Studies in Second Language Acquisition, 16(4), 413–439.
Mieszkowska, K., Luniewska, M., Kolak, J., & Kacprzak, A. (2017). Home language will take care of itself:
Vocabulary knowledge in trilingual children in the United Kingdom. Frontiers in Psychology, 8, 1–11.
Milroy, L., & Muysken, P. (Eds.). (1995). One speaker, two languages. Cross-disciplinary perspectives on
code-switching. Cambridge: Cambridge University Press.
Montanari, S. (2010). Language differentiation in early trilingual development. Evidence from a case
study. Saarbrücken, Germany: Verlag Dr. Müller.
Müller, N., Arnaus Gil, L., Eichler, N., Geveler, J., Hager, M., Jansen, V., Patuto, M., Repetto, V. &
Schmeißer, A. (2015). Code-Switching: Spanisch, Italienisch, Französisch. Eine Einführung.
Tübingen: Narr.
Muysken, P. (1997). Code-Switching processes. Alternation, insertion, congruent lexicalization. In M.
Pütz (Ed.), Language Choices: Conditions, constraints, and consequences (pp. 361–380). Amsterdam:
Benjamins.
Patuto, M., Hager, M., Arnaus Gil, L., Eichler, N., Jansen, V., Schmeißer, A., & Müller, N. (2014). Child-
external and -internal factors in bilingual code-switching: Spanish, Italian, French and German.
In A. Koll-Stobbe & S. Knospe (Eds.), Language contact around the globe. Proceedings of the
LCTG3 Conference (pp. 191–209). Frankfurt am Main: Lang.
 INTERNATIONAL JOURNAL OF MULTILINGUALISM 33

Peña, E. (2007). Lost in translation: Methodological considerations in cross-cultural research. Child

Development, 78(4), 1255–1264.
Petermann, F. (Ed.). (20112). Wechsler preschool and primary scale of intelligence-III (WPPSI-III).
Frankfurt am Main: Pearson.
Poeste, M. (2017). Code-Switching als Insertion oder Alternation: Eine empirische Studie bilingualer
Kinder mit Deutsch und einer romanischen Sprache. Unpublished bachelor thesis, Bergische
Universität Wuppertal.
Poplack, S. (1980). Sometimes I’ll start a sentence in Spanish y termino en español: Toward a typology
of code-switching. Linguistics, 18(7), 581–618. Retrieved from [Link]
downloadpdf/j/[Link]-7-8/ling.1980.18.7-8.581/[Link]
Quay, S. (2001). Managing linguistic boundaries in early trilingual development. In J. Cenoz & F.
Genesee (Eds.), Trends in bilingual acquisition (pp. 149–199). Amsterdam: Benjamins.
Quay, S. (2008). Dinner conversations with a trilingual two-year-old: Language socialization in a multi-
lingual context. First Language, 28(1), 5–33.
Quay, S. (2011a). Introduction: Data-driven insights from trilingual children in the making.
International Journal of Multilingualism, 8(1), 1–4.
Quay, S. (2011b). Trilingual toddlers at daycare centres: The role of caregivers and peers in language
development. International Journal of Multilingualism, 8(1), 22–41.
Riehl, C. (2014). Sprachkontaktforschung. Eine Einführung. Tübingen: Narr.
Rothman, J. (2011). L3 syntactic transfer selectivity and typological determinacy: The typological
primacy model. Second Language Research, 27(1), 107–127.
Rothman, J. (2015). Linguistic and cognitive motivations for the typological primacy model (TPM) of
third language (L3) transfer: Timing of acquisition and proficiency considered. Bilingualism:
Language and Cognition, 18(2), 179–190.
Schmeißer, A., Eichler, N., Arnaus Gil, L., & Müller, N. (2016a). Mélanges interpropositionnels chez les
enfants bilingues français-allemands: est-ce vraiment du code-switching? Language, Interaction
and Acquisition LIA, 7(2), 238–274.
Schmeißer, A., Hager, M., Arnaus Gil, L., Jansen, V., Geveler, J., Eichler, N., Patuto, M., & Mü ller, N.
(2016b). Related but different: The two concepts of language dominance and language profi-
ciency. In C. Silva-Corvalán & J. Treffers-Daller (Eds.), Language dominance in bilinguals: Issues of
operationalization and measurement (pp. 36–65). Cambridge: Cambridge University Press.
Silva-Corvalán, C. (1983). Code-Shifting Patterns in Chicano Spanish. In L. Elías-Olivares (Ed.), Spanish
in the U.S setting: beyond the Southwest (pp. 69–87). Rosslyn, VA: National Clearinghouse for
Bilingual Education.
Sivakumar, A. (2017). Die soziolinguistischen Faktoren des Code-Switching bei trilingual aufwachsen-
den Kindern: Eine empirische Untersuchung eines trilingual spanisch-französisch-italienischen
aufwachsenden Kindes. Unpublished master’s thesis, Bergische Universität Wuppertal.
Sivakumar, A., Sette, N., Müller, N., & Arnaus Gil, L. (2019). Die Entwicklung des rezeptiven
Wortschatzes bei bi-, tri- und multilingual aufwachsenden Kindern. Submitted.
Unsworth, S. (2013). Current issues in multilingual first language acquisition. Annual Review of Applied
Linguistics, 33, 21–50.
Wang, X. (2008). Growing up with three languages. Birth to eleven. parents’ and teachers’ guides. Bristol:
Multilingual Matters.