Phonetics and Phonology

Much of the recent research on multilingual first language acquisition concerns the developing perceptual abilities of bilingual infants (for overviews, see Sebastián-Gallés, Reference Sebastián-Gallés2010; Werker, Reference Werker2012; Werker & Byers-Heinlein, Reference Werker and Byers-Heinlein2008), with comparatively few studies on phonological development in early and later childhood (Fabiano-Smith & Barlow, Reference Eichler, Jansen and Müller2010).

One of the first tasks facing children growing up in a multilingual setting is to distinguish and separate the speech input to which they are exposed into two (or more) languages (i.e., language discrimination). A well-established finding from earlier research in this area is that monolingual and bilingual infants show similar patterns of sensitivity to perceptual cues provided by a language's rhythmicityFootnote ¹, that is, they are able to discriminate languages from different rhythmic classes at birth (e.g., Ramus, Hauser, Miller, Morris, & Mehler, Reference Ramus, Hauser, Miller, Morris and Mehler2000) and at around 4 to 5 months, they can also discriminate languages from the same rhythmic class (e.g., Bosch & Sebastián-Gallés, Reference Bosch and Sebastián-Gallés2001). Furthermore, it has been observed that at birth, monolingual infants are better able to discriminate their native language from an unfamiliar language than they are able to discriminate two unfamiliar languages (e.g., Mehler et al., Reference Mehler, Jusczyk, Lambertz, Halsted, Bertonicini and Amiel-Tison1988). In a recent study comparing bilingual (English-Tagalog) and monolingual (English) infants, Byers-Heinlein, Burns, and Werker (Reference Byers-Heinlein, Burns and Werker2010) demonstrated that similar behavior exists for bilinguals. More specifically, in a preferential listening task, Byers-Heinlein et al. observed that newborn infants exposed to bilingual speech in utero demonstrated the ability to discriminate between the two languages spoken by the mother during pregnancy, but they did not show a preference for one over the other, listening to the two equally well (cf. monolinguals, who prefer their native language over an unfamiliar language). As Byers-Heinlein et al. noted, not only do these findings provide evidence that the process of bilingual acquisition has already started at birth, they also show that in the very earliest stages of language development, bilingual and monolingual children make use of the same perceptual biases (i.e., rhythm). Bilingual children have also been found to make use of certain cues used to discriminate language, namely, facial movement, for longer than monolinguals (Sebastián-Gallés, Albareda-Castellot, Weikum, & Werker, Reference Werker2012; Weikum et al., Reference Weikum, Vouloumanos, Navarra, Soto-Faraco, Sebastián-Gallés and Werker2007).

After discriminating between their two languages, children subsequently need to discriminate the relevant phonetic categories within each language. One frequently cited result from earlier research is that when acquiring two highly similar languages, bilingual infants experience a temporary delay around age 8 months, when, unlike monolinguals, they are unable to perceive certain vowel contrasts (e.g., /e/ vs. /ɛ/ in Catalan; Bosch & Sebastián-Gallés, Reference Bosch and Sebastián-Gallés2003). Frequency of input and degree of overlap between phonetic ranges have been put forward as possible explanations for these findings, with the authors suggesting that the considerably higher frequency of the Spanish vowel /e/, which is phonetically very similar to the two Catalan vowels, leads to a temporary inability to make the distinction in Catalan.

More recent research on bilingual discrimination of consonants has, however challenged the generalizability of this finding, with at least two studies showing that bilinguals are able to perceive distinctions or phonemes that are realized in both languages in similar but, crucially, slightly different ways at the same age as monolinguals (Burns, Yoshida, Hill, & Werker, Reference Burns, Yoshida, Hill and Werker2007; Sundara, Polka, & Molnar, Reference Sundara, Polka and Molnar2008). The authors of both of these studies speculated that frequency of occurrence and distribution patterns may play a role in determining the timing of bilingual perceptual reorganization (but see Sebastián-Gallés & Bosch, Reference Sebastián-Gallés, Albareda-Castellot, Weikum and Werker2009). Interestingly, in a more recent article, (Albareda-Castellot, Pons, & Sebastián-Gallés, Reference Albareda-Castellot, Pons and Sebastián-Gallés2011) presented evidence from an anticipatory eye movement paradigm, which suggests that contrary to their previous findings, bilingual Catalan-Spanish children are able to discriminate the /e/-/ ɛ/ contrast in Catalan. That is, the familiarization-preference procedure in previous studies may have concealed bilingual children's real abilities (Albareda-Castellot et al., Reference Albareda-Castellot, Pons and Sebastián-Gallés2011).

We now turn to the phonological development of older bilingual children, an area that has not been widely investigated (but see Holm & Dodd, Reference Holm and Dodd1999; Lleó & Kehoe, Reference Lleó and Kehoe2002). The limited number of recent studies have largely focused on two issues: whether the bilingual acquisition of speech sounds follows the same time course as for monolinguals and whether there is any evidence for transfer or cross-linguistic influence.

In a study on the production of consonants in Spanish and English, Fabiano-Smith and Goldstein (Reference Fabiano-Smith and Goldstein2010) found that while bilingual English-Spanish 3-year-olds showed a comparable rate of development to their monolingual peers in English, in Spanish they were slower. This difference was, however, limited to just a few manner classes, namely trills (e.g., /r/), fricatives (e.g., /s/), and glides (e.g., /w/), with other classes evidencing no differences between bilinguals and monolinguals. The authors speculated that bilingual children may be using their phonological knowledge in one language to facilitate their acquisition in the other, allowing them to acquire the two within the same timeframe as monolinguals. In a study on Russian-English bilingual children, Gildersleeve-Neumann and Wright (Reference Gildersleeve-Neumann and Wright2010) found that bilingual children demonstrated the same level of complexity as monolinguals in their phonetic inventories of consonants (where relevant data are available; see also Fabiano-Smith & Barlow, Reference Eichler, Jansen and Müller2010), but they made more errors with vowels, a pattern that is reminiscent of some of the aforementioned studies concerning bilingual children's early perceptual development.

These studies also found evidence—albeit limited in some cases—of cross-linguistic influence. For example, Gildersleeve-Neumann and Wright (Reference Gildersleeve-Neumann and Wright2010) observed transfer of final devoicing from Russian to English, irrespective of the child's specific bilingual environment, and two of the eight bilingual children in the Fabiano-Smith and Goldstein (Reference Fabiano-Smith and Goldstein2010) study produced de-aspirated stops in English as a result of influence from Spanish (see also Gildersleeve-Neumann, Kester, David, & Peña, Reference Gildersleeve-Neumann, Kester, David and Peña2008, for more extensive evidence of transfer from Spanish to English in Spanish-English bilingual children).

To summarize, recent research into the phonetic and phonological abilities of bilingual children suggests that they are sensitive to language exposure in utero, and that they make use of the same perceptual biases as monolingual children. There is, however, some evidence for a slower rate of differentiation of certain language-specific contrasts among bilingual infants and for cross-linguistic influence in the phonological development of bilingual children. By and large, however, the perceptual and phonological development of bilingual children is similar to that of monolinguals.

Vocabulary

Within the recent research on the multilingual first language acquisition of vocabulary there have broadly speaking been two main areas of interest: (a) the very earliest stages of word learning and the extent to which bilingual children make use of the same types of constraints and phonetic detail to acquire words as monolingual children (see, e.g., Vihman, Thierry, Lum, Keren-Portnoy, & Martin, Reference Vihman, Thierry, Lum, Keren-Portnoy and Martin2007) and (b) children's subsequent vocabulary growth and the factors which modulate this.

A frequently studied constraint in early vocabulary acquisition is mutual exclusivity, the assumption that each object category must have a different label. This word-learning bias essentially means that when presented with a novel word in the context of one familiar and one unfamiliar object, children assume the intended referent to be the unfamiliar object; mutual exclusivity has been observed in monolingual children at age 17 months (Halberda, Reference Halberda2003), but there is little agreement as to its developmental origin (Werker, Reference Werker2012). The results of two recent studies with bilingual and trilingual children suggest that this constraint is not as prominent in multilingual first language acquisition (Byers-Heinlein & Werker, Reference Byers-Heinlein and Werker2009; Houston-Price, Caloghiris, & Raviglione, Reference Houston-Price, Caloghiris and Raviglione2010). More specifically, it has been claimed that mutual exclusivity may be a developmental bias that emerges in monolinguals as the result of experience with one-to-one mappings: children growing up with more than one language are consistently exposed to more than one label for the same referent and consequently fail to develop this bias, or at least do not employ it to the same degree as monolinguals. Interestingly, Byers-Heinlein and Werker (Reference Byers-Heinlein and Werker2009) found the degree to which the multilingual children in their study made use of mutual exclusivity varied according to the number of languages which the child was learning, that is, whereas marginal use of this bias was detected for bilinguals, for trilinguals, it was not observed at all.

A fundamental building block in the development of the lexicon is the ability to associate a word and object. Typically, studies of this kind employ a so-called switch word-learning task, where infants are first habituated to two word-object combinations (Word A–Object A, Word B–Object B) and subsequently tested in a trial where one of these combinations is switched (e.g., Word A–Object B) to determine the extent to which they have learned the relevant associative link. Recent research in this domain has shown that bilinguals develop associative word learning with dissimilar words (e.g., pok and lif) at the same age as monolinguals, around 14 months (Byers-Heinlein, Fennell, & Werker, Reference Byers-Heinlein, Fennell and Werker2013), whereas when similar-sounding words (e.g., bih and dih) are employed, bilingual children successfully complete the task about 3 months later than monolinguals, at 20 months of age (Fennell, Byers-Heinlein, & Werker, Reference Fennell, Byers-Heinlein and Werker2007). This latter finding has been challenged, however: Mattock, Polka, Rvachew, and Krehm (Reference Mattock, Polka, Rvachew and Krehm2010) showed that when tested using stimuli that are representative of their language learning environment (i.e., stimuli which include variants from both languages), bilingual children do, in fact, demonstrate associative word learning with similar-sounding words, illustrating the importance of taking the bilingual child's language learning environment into account not only in the analysis of experimental results but also in the experimental design.

One well-established finding in the literature on bilingual children's lexical development is that while overall vocabulary levels between bilingual preschoolers and their monolingual peers are comparable, when their languages are considered separately, bilingual children typically score below age-appropriate norms for monolingual children on tasks of receptive vocabulary such as the Peabody Picture Vocabulary Test (PPVT; Dunn & Dunn, Reference Dunn and Dunn2007; Pearson, Fernández, Lewedeg, & Oller, Reference Pearson, Fernández, Lewedeg and Oller1997; Pearson, Fernández, & Oller, Reference Pearson, Fernández and Oller1993). This finding has been attributed to the distributed characteristic of bilingual language learning; that is, children typically acquire their two (or more) languages in different contexts, often with one language restricted to use at home, and thus it is unsurprising that the words they know in a given language are generally restricted to the context in which they acquire that language (Oller, Pearson, & Cobo-Lewis, Reference Oller, Pearson and Cobo-Lewis2007).

In light of this, Bialystok, Luk, Peets, and Yang (Reference Bialystok, Luk, Peets and Yang2010) conducted a large-scale study of children between the ages of 3 and 10 years to determine whether the aforementioned bilingual-monolingual differences were also found across a wider age range and a larger number of children (n = 1738). In short, their results indicate that when tested using the PPVT in English, their school language, the bilingual children in all age groups scored significantly lower than monolinguals; however, once scores for school-related vocabulary were analyzed separately from home-related vocabulary items, bilingual and monolingual children's scores become much more comparable, with the difference for home-related items remaining, illustrating the context-specific nature of vocabulary acquisition in a multilingual context. As the authors concluded, “the smaller vocabulary for bilingual children in each language is not an overall disadvantage but rather an empirical description that needs to be taken into account in research designs, especially in tasks that involve verbal ability or lexical processing” (Bialystok et al., Reference Bialystok, Luk, Peets and Yang2010, p. 530).

It is important to bear in mind that the latter study assessed children's vocabulary knowledge in one language only. Replicating earlier findings by Pearson and colleagues (Reference Pearson, Fernández, Lewedeg and Oller1997), other recent studies have shown that once bilingual children's knowledge of both languages is taken into account, their overall vocabularies are indeed on a par with their monolingual peers, and this has been found to hold for children from low socioeconomic status (SES) families (Mancilla-Martinez, Pan, & Vagh, Reference Mancilla-Martinez, Pan and Vagh2011) as well as those from a middle/high SES background (e.g., Marchman & Martinez-Sussman, Reference Marchman and Martinez-Sussman2002). As highlighted in the study by Mancilla-Martinez et al., however, how best to measure children's overall vocabulary is a complex and challenging task requiring further research.

Much of the work on bilingual vocabulary development focuses on young children, often at the preschool or kindergarten age (Barnes & García, in press; David & Wei, Reference David and Wei2008; Dixon, Reference Dixon2011). There are, however, a few studies, in addition to Bialystok et al. (Reference Bialystok, Luk, Peets and Yang2010), that investigate the lexical development of older children. Both Sheng, Lu, and Kan (Reference Sheng, Lu and Kan2011, Mandarin-English) and Gathercole and Thomas (Reference Gathercole and Thomas2009, Welsh-English) found evidence for the stagnation of vocabulary development as a function of reduced input (see below for further discussion concerning effects of input quantity).

One final strand of research on bilingual lexical development concerns the link between vocabulary and other areas, namely, between vocabulary and grammar (Marchman, Martinez-Sussman, & Dale, Reference Marchman, Martinez-Sussman and Dale2004), and between vocabulary and processing (Marchman, Fernald, & Hurtado, Reference Mattock, Polka, Rvachew and Krehm2010). The question of whether there are strong ties between the development of the lexicon and the development of the grammar and/or processing relates to issues such as modularity, the role of domain-specific versus domain-general learning mechanisms, and maturation (Conboy & Thal, Reference Conboy and Thal2006). Studying the relationship between grammatical and lexical development in bilingual children can contribute to the debate about these issues because it allows us to examine both within-language and cross-language relationships while holding maturation constant, that is, the relationship between, for example, grammar and vocabulary can be ascertained within the same child for both languages; if certain developmental patterns are the result of maturation, these should—all things being equal—hold across both languages within the same child. The results of the aforementioned studies suggest that within-language ties are stronger than cross-language ties, and for the relationship between lexicon and grammar, Conboy and Thal (Reference Conboy and Thal2006) concluded that the lack of significant cross-language relationships is incompatible with the idea that these links are the result of maturation.

To summarize, recent research on the multilingual first language acquisition of vocabulary shows that early word learning is affected by context in terms of children's sensitivity to and use of certain constraints found to be relevant to monolingual acquisition. Furthermore, vocabulary is perhaps the one area where significant and persistent bilingual-monolingual differences have been observed, although note that this concerns comparisons involving one of the bilingual's languages only.

Morphosyntax

Most of the research on the multilingual first language acquisition of morphosyntax focuses on oral production, and the results suggest that when any bilingual-monolingual differences are observed, they are generally quantitative in nature rather than qualitative, although qualitative differences have been observed. The target language properties investigated range from more general measures of grammatical ability, such as mean length of utterance (MLU), to specific aspects of the language(s) in question (e.g., clitics or subject-verb agreement).

Several studies report significant differences between bilinguals and monolinguals on general measures of grammatical ability. For example, Hoff et al. (Reference Hoff, Core, Place, Rumiche, Señor and Parra2012) used caregiver-report data (MacArthur-Bates Communicative Development Inventory [CDI]; Fenson et al., Reference Fenson, Dale, Reznick, Thal, Bates, Hartung and Reilly1993) to estimate grammatical complexity and utterance length in a group of English-Spanish toddlers in the United States (1;10 to 2;6) and their age-matched monolingual peers. They found that the monolingual children were significantly more advanced than the bilinguals, but once the bilingual children were divided into dominance groups, based on amount of input as indicated by a detailed parental questionnaire, the bilingual children with the most exposure to the language to question were no longer significantly different from the monolinguals. In a study of four Turkish-Dutch bilingual toddlers, Blom (Reference Blom2010) found significant bilingual-monolingual differences for MLU (in words) in Dutch, but not for Turkish (with similar findings for lexical development); she noted that the results for Turkish may have been due to the considerable variation observed in the monolingual group. This raises the interesting methodological question of what should count as similarity or difference between bilinguals and monolinguals and how to deal with the individual variation that also exists among monolinguals.

In a study on the acquisition of finiteness in bilingual Basque-Spanish children, Austin (Reference Austin2009; see also Austin, in press; Blom, Reference Blom2010) also found that the rate of root infinitives in Basque is higher for bilinguals than their monolingual peers. The term root (or optional) infinitive refers to matrix clauses where the child produces a nonfinite verb instead of the target finite form (e.g., “she eat ice cream” instead of “she eats ice cream.” Austin attributes this finding to the bilinguals’ comparatively limited input, which restricts the rate at which children are able to acquire the target morphologically-specified paradigm (following Blom, Reference Blom2007). These findings contrast with those of Hacohen and Schaeffer (Reference Hacohen and Schaeffer2007), who in a case study of a bilingual child growing up with English and Hebrew, found very low rates of subject-verb agreement errors in Hebrew, comparable with monolinguals (cf. the same child's use of overt subjects, which are regularly pragmatically inappropriate—see also section below on semantic and pragmatics).

Rate of acquisition differences between bilinguals and monolinguals have also been found for the use of word order cues to identify sentential subjects (Gathercole & Thomas, Reference Gathercole and Thomas2009), clitics and direct objects (Larrañaga & Guijarro-Fuentes, Reference Larrañaga and Guijarro-Fuentes2012; Pérez-Leroux, Cuza, & Thomas, Reference Pérez-Leroux, Cuza and Thomas2011; Pérez-Leroux, Pirvulescu, & Roberge, Reference Pérez-Leroux, Pirvulescu and Roberge2009), and grammatical gender (Eichler, Jansen, & Müller, in press; Gathercole & Thomas, Reference Gathercole and Thomas2009; Larrañaga & Guijarro-Fuentes, in press; Nicoladis & Marchak, Reference Nicoladis and Marchak2011; Unsworth, Reference Unsworth2013). In the acquisition of grammatical gender, bilingual children have been observed to make the same type of errors as monolingual children, but for some languages at least (e.g., Welsh and Dutch), their rate of acquisition may be considerably slower than that of monolinguals, at least for those children with relatively little exposure to the language in question. For example, in a study on the acquisition of gender marking on definite determiners and adjectives by English-Dutch bilinguals aged 3 through 17, Unsworth (Reference Unsworth2013) found significant differences between bilinguals and monolinguals when matched on age; however, once matched on their cumulative length of exposure, that is, their amount of exposure over time, these differences disappeared. Once again, this finding raises important methodological questions about how bilinguals and monolinguals can be best compared; while bilinguals are often matched with monolinguals of the same age or with the same MLU, an alternative—and for some purposes, potentially more accurate—basis for comparison might be cumulative length of exposure.

Most of the attested bilingual-monolingual differences in the domain of morphosyntax are quantitative in nature, that is, bilingual children typically make the same developmental errors as their monolingual peers but to a greater—and sometimes lesser—degree. Qualitative differences, however, have been observed, too. For example, in an extensive longitudinal study of six Cantonese-English bilingual children, Yip and Matthews (Reference Yip and Matthews2007) found that under influence from Cantonese, the bilingual children pass through a stage where they consistently use wh-in-situ in English questions, as in This is what? instead of What is this? As Yip and Matthews (Reference Yip and Matthews2007) noted, monolingual children consistently place wh- expressions in wh- questions in the target clause-initial position. The same authors also argued that qualitative differences may in fact underlie what at first appear to be quantitative differences: for example, in the same study of Cantonese-English bilinguals, they suggested that the use of null objects in the children's English results from transfer of the mechanism that licenses sentence topics in Cantonese to English (Yip & Matthews, Reference Yip and Matthews2007).

In line with much of the earlier work on the multilingual first language acquisition of morphosyntax (for reviews, see Genesee & Nicoladis, Reference Genesee, Nicoladis, Hoff and McCardle2007; Meisel, Reference Meisel, Bhatia and Ritchie2004), several recent studies observed similar patterns of development between bilinguals and monolinguals. For example, in a study on the early word order acquisition of bilingual Basque-Spanish children, Barreña and Almgren (in press) observed that bilingual children showed the same word order patterns as monolingual children (i.e., predominantly verb-object [VO] in Spanish and both object-verb [OV] and VO in Basque), replicating earlier findings on bilingual French-German children (Meisel, Reference Meisel, Hyltenstam and Obler1989). Bonnesen (Reference Bonnesen, Guijarro-Fuentes, Larrañaga and Clibbens2008) also observed the target use of verb-second in German and target finite verb placement (i.e., no verb-second) in French in French-German bilingual children (see also Rothweiler, Reference Rothweiler and Lleó2006, for similar findings with Turkish-German bilingual children). Finally, Hoff et al. (Reference Hoff, Core, Place, Rumiche, Señor and Parra2012) showed that when both of a bilingual child's two languages were taken into account, bilinguals were found to produce multiword utterances at a comparable age to monolinguals. This latter study stressed the importance of assessing children in both their languages in order to make an accurate assessment of their linguistic abilities.

To summarize, while there is considerable evidence that the multilingual first language acquisition of morphosyntax by and large follows the same developmental path as monolingual first language acquisition, where differences are observed, recent research continues to show that they are mostly quantitative rather than qualitative in nature. These differences have been attributed to a wide range of factors, including cross-linguistic influence, dominance, and amount and type of input, each of which will be discussed in turn below.

Semantics and Pragmatics

Studies on the multilingual first language acquisition of semantics and/or pragmatics are few and far between. Much of the work in this area has been conducted by Sorace, Serratrice, and colleagues in relation to the interface hypothesis (Sorace, Reference Sorace2011), and has compared bilingual children's acquisition of the syntax-semantics versus syntax-discourse interface; in addition, a number of recent studies have investigated the acquisition of scrambling.

In a study on the acceptability of plural Noun Phrase (NP) subjects in generic and specific context, Serratrice, Sorace, Filiaci, and Baldo (Reference Serratrice, Sorace, Filiaci and Baldo2009) found that English-Italian bilingual children overaccepted (ungrammatical) plural bare plurals in generic contexts in Italian, as in *In genere squali sono pericolosi (“In general sharks are dangerous”; cf. monolinguals who consistently rejected such sentences). A comparable group of Spanish-Italian bilinguals were shown to behave significantly more like the Italian monolinguals, suggesting that typological relatedness (Spanish is more similar to Italian than English) is another factor that should be taken into consideration when assessing bilingual children's language abilities and, more specifically, the likelihood of cross-linguistic influence.

The results concerning the bilingual English-Italian children's sensitivity to specificity and genericity contrast with their use of subject pronouns. In a study with the same children, Sorace, Serratrice, Filiaci, and Baldo (Reference Sorace, Serratrice, Filiaci and Baldo2009) found that both English-Italian and Spanish-Italian children overaccepted overt subject pronouns referring to a pragmatically inappropriate topic subject antecedent in Italian when compared with monolinguals. The bilingual children thus showed similar response patterns, irrespective of their other language. This led Sorace and Serratrice (Reference Sorace and Serratrice2009) to the conclusion that the crucial factor in determining the extent of bilingual-monolingual differences is not typological relatedness, although this clearly plays a role, but whether the linguistic property being acquired involves an internal (syntax-semantics) or external (syntax-discourse) interface (see Sorace, Reference Sorace2011, and commentaries in the same volume for more on this matter). Note, however, that results from a recent study on German-Italian and French-Italian bilinguals do not tally with this claim: While in line with previous findings, Schmitz, Patuto and Müller (Reference Schmitz, Patuto and Müller2012) observed the overuse of subject pronouns in Italian by those children acquiring German as the other language, while the children acquiring French as the other language were similar to monolinguals (i.e., they did not overuse subject pronouns in the same way as the other groups).

Two recent studies have investigated the acquisition of scrambling and, more specifically, the interpretive constraints associated with scrambled indefinite objects.Footnote ² In a study on the acquisition of scrambling in Ukrainian, Mykhaylyk and Ko (Reference Mykhaylyk, Ko, Anderssen, Bentzen and Westergaard2010) showed that bilingual English-Ukrainian children passed through the same stages of development as monolingual children, and that, like monolingual children, their scrambled indefinite objects were restricted to a specific or wide-scope interpretation, as the target adult language required. Similar results were found for English-Dutch bilinguals by Unsworth (Reference Unsworth, Biller, Chung and Kimball2012): Like monolinguals, the vast majority of bilingual children in this study were able to consistently interpret scrambled indefinites in negative sentences as taking a specific or wide-scope interpretation by the age of 6 years old (see Lee, Kwak, Lee, & O'Grady, 2010; O'Grady, Kwak, Lee, & Lee, Reference O'Grady, Kwak, Lee and Lee2011, for related work on scope preferences in Korean-English heritage speakers). As Unsworth (Reference Unsworth, Biller, Chung and Kimball2012) noted, this finding is all the more striking given the paucity of scrambled indefinite objects in the input.

To summarize, while recent research has started to explore the multilingual first language acquisition of semantics and pragmatics, the number of studies is very limited and the results are mixed.

Cross-linguistic Influence

Whereas the focus of earlier research on multilingual first language acquisition was on the question of whether a bilingual child's two languages developed as one or two systems (De Houwer, Reference De Houwer1990; Meisel, Reference Meisel, Hyltenstam and Obler1989; Volterra & Taeschner, Reference Volterra and Taeschner1978), it is generally assumed today that children separate their two languages from very early on, but that some level of interaction between the two may nevertheless occur (Paradis & Genesee, Reference Paradis and Genesee1996). Much of the research over the past 15 years or so has sought to define the linguistic conditions under which such interaction (or cross-linguistic influence, as it is commonly called) manifests itself and what factors modulate its occurrence.

The most prominent proposal concerning the linguistic conditions is that of Hulk and Müller (Reference Hulk and Müller2000; see also Müller & Hulk, Reference Müller and Hulk2001), who proposed that for cross-linguistic influence to occur, two conditions must be fulfilled: (a) The target language property should involve the syntax-pragmatics interface or C-domain, and (b) there should be surface overlap. Thus if language A offers evidence for more than one grammatical analysis for the target language property in question, and language B reinforces one of these analyses, cross-linguistic influence is “probable” (Müller & Hulk, Reference Müller and Hulk2001, p. 2; see Döpke, Reference Döpke1998, for a similar proposal). Research conducted since this proposal was first advanced has shown that condition (a) is too strict, in that cross-linguistic influence has been observed in other areas such as narrow syntax, syntax-morphology, and syntax-semantics-lexicon; (e.g., Argyri & Sorace, Reference Argyri and Sorace2007; Liceras, Fernández Fuertes, & Alba de la Fuente, Reference Liceras, Fernández Fuertes and Alba de la Fuente2011; Pérez-Leroux et al., Reference Pérez-Leroux, Cuza and Thomas2011), but that condition (b) seems to hold (Foroodi-Nejad & Paradis, Reference Foroodi-Nejad and Paradis2009; Hacohen & Schaeffer, Reference Hacohen and Schaeffer2007; see Schmitz et al., Reference Schmitz, Patuto and Müller2012, for a proposed revision to the original formulation of this condition).

An interesting test case for any account of cross-linguistic influence comes from bimodal bilingual acquisition; that is, children who grow up with one spoken and one signed language, as do hearing children of deaf adults, or codas/kodas, as they are often called (see e.g., Petitto et al., Reference Petitto, Katerelos, Levy, Guana, Tétreault and Ferraro2001). Research on coda development is still in its infancy, but in a series of recent studies, Chen Pichler, Lillo-Martin, and colleagues, have started to explore the topic of cross-linguistic influence in this population. For example, in a study on the acquisition of wh- questions in bilingual children acquiring English and American Sign Language (ASL) or Brazilian Portuguese and Brazilian Sign Language (Libras), Lillo-Martin, Koulidobrova, de Quadros, and Chen Pichler (Reference Lillo-Martin, Koulidobrova, de Quadros, Chen Pichler, Biller, Chung and Kimball2012) found evidence for cross-linguistic influence in both directions: in line with the findings for English in the English-Cantonese bilinguals of Yip and Matthews (Reference Yip and Matthews2007), the bilingual toddlers produced significantly more non-sentence-initial wh- questions—grammatical in both of the sign languages—than their monolingual English-speaking and Brazilian Portuguese-speaking peers, and concomitantly, in their sign languages, they produced proportionally more sentence-initial wh- phrases than their monolingual peers, arguably as a result of their simultaneous exposure to English.

Cross-linguistic influence may manifest itself as delay or acceleration (Paradis & Genesee, Reference Paradis and Genesee1996), sometimes simultaneously (Brasileiro, Reference Brasileiro2009). As this review illustrates, most of the studies that found bilingual-monolingual differences have observed delay (i.e., a slower rate of acquisition for bilinguals), although it is important to note that this is for a specific and limited number of properties of language only, not a general bilingual delay (cf. Gathercole, Reference Gathercole2007). There are, however, a handful of studies that found evidence for acceleration, that is, a faster rate of development in bilinguals when compared with their monolingual peers. For example, Kupisch (Reference Kupisch2007) observed that German-Italian bilingual children acquire German determiners faster than most monolingual German children. Her explanation for this finding is was based on the notion of complexity; that is, she argued that the German determiner system is more complex than the Italian one because it contains more types, each type encodes numerous formal features, and the form-function mapping is opaque. Consequently, monolingual German-speaking children typically produce determiners later than monolingual Italian-speaking children. In the case of the bilingual children in her study, the presence of a simpler system, which already had determiners in place, facilitated the development of the more complex determiner system in German. Kupisch also suggested that such cross-linguistic influence only takes place in unbalanced bilingual children when the dominant language, in this case Italian, is the facilitating language.

Another example of acceleration is found in a recent study by Liceras et al. (Reference Liceras, Fernández Fuertes and Alba de la Fuente2011; see also Meroni, Unsworth, & Smeets, Reference Meroni, Unsworth, Smeets, Baiz, Goldman and Hawkes2013), who argued that the presence of two different copulas in Spanish, ser and estar, facilitates the acquisition of the copula in English, such that the bilingual English-Spanish children in their study had lower rates of copula omission than had been observed for monolinguals (Becker, 2004). Finally, on a more general note, in a study on the complexity of phonetic inventories, Fabiano-Smith and Barlow (Reference Fabiano-Smith and Goldstein2010) suggested that the observation that bilingual children achieve the same level of complexity within each of their two languages within the same timeframe as monolingual children could be interpreted as evidence for cross-linguistic influence in the form of accelerated development, that is, the complexity developed in one language may bootstrap development in the other (following Gawlitzek-Maiwald & Tracy, Reference Gawlitzek-Maiwald and Tracy1996).

To summarize, although recent research has helped refine the conditions that drive cross-linguistic influence and further documented how such influence manifests itself, many outstanding issues remain. It is, for example, clear that the conditions on cross-linguistic influence—however they are formulated—are sufficient but not necessary. Not all children exhibit cross-linguistic influence even when the relevant conditions are met (Gathercole, Reference Gathercole2007). The answer to the question of how best to predict which individual children will evidence cross-linguistic influence remains elusive. A recent study by Hauser-Grüdl, Arencibia Guerra, Witzmann, Leray, and Müller (Reference Hauser-Grüdl, Arencibia Guerra, Witzmann, Leray and Müller2010) suggested that fluency, as measured by number of words produced per minute, may also play a role in predicting cross-linguistic influence. Two of the most common predictors discussed in the literature are, however, dominance and relative amount of exposure, sometimes with the latter being used as a proxy for the former.

Dominance

How to measure dominance, broadly defined as “the condition in which bilingual people have greater grammatical proficiency in, more vocabulary in, or greater fluency in one language or simply use one language (i.e., the dominant language) more often” (Genesee et al., Reference Genesee, Paradis and Crago2004, p. 80), remains a controversial issue in the literature on multilingual first language acquisition. While it is clear that most bilingual children are dominant in one of their languages, and that this can and often does change over time (e.g., Yip & Matthews, Reference Yip and Matthews2007), the extent to which dominance can (systematically) predict and explain children's language outcomes remains unclear (for recent discussions, see Bedore et al., Reference Bedore, Pena, Summers, Boerger, Resendiz, Greene and Gillam2012; Cantone, Müller, Schmitz, & Kupisch, Reference Kupisch, Guijarro-Fuentes, Larrañaga and Clibbens2008).

As the definition above indicates, dominance can be defined and subsequently operationalized in terms of children's relative competence or proficiency in their two (or more) languages, or in terms of their relative use (see Yip & Matthews, Reference Yip and Matthews2006, for relevant discussion). Proficiency-based variables commonly used as measures of dominance include MLU, upper bound (i.e., longest utterance[s] in a recording), the number of multimorphemic utterances (Genesee, Nicoladis, & Paradis, Reference Genesee, Nicoladis and Paradis1995), and, more recently, increases in the noun and verb lexicon (Kupisch, Reference Kupisch2007, Reference Kupisch, Guijarro-Fuentes, Larrañaga and Clibbens2008) and vocabulary size (Foroodi-Nejad & Paradis, Reference Foroodi-Nejad and Paradis2009). Measures of dominance based on language use include number of utterances, fluency (based on parental report, Foroodi-Nejad & Paradis, Reference Foroodi-Nejad and Paradis2009; Hauser-Grüdl et al., Reference Hauser-Grüdl, Arencibia Guerra, Witzmann, Leray and Müller2010), and amount of exposure (e.g., Argyri & Sorace, Reference Argyri and Sorace2007).

Once one or more measure have been selected, a further consideration is how it should be used. One approach is to use the children's scores on the selected measure(s) to calculate what is called a balance score (Cantone et al., Reference Cantone, Müller, Schmitz and Kupisch2008), whereby scores from language A are subtracted from the scores from language B, and any differences are assumed to indicate dominance (Bedore et al., Reference Bedore, Pena, Summers, Boerger, Resendiz, Greene and Gillam2012; Pérez-Leroux et al., Reference Pérez-Leroux, Cuza and Thomas2011). The extent to which a given difference can justifiably be taken as an indicator of dominance will, of course, depend on the measure and the range of attested scores; it is, however, sometimes far from clear how large a difference needs to be in order to be able to speak of dominance. One approach is to compare children's scores statistically and only count significant between language differences as indicators of dominance (Yip & Matthews, Reference Yip and Matthews2006), or—when using standardized scores—to see whether children reach age-appropriate monolingual norms for one language, but not the other. Despite these complications, using such differential scores for MLU, as advocated by Yip and Matthews (Reference Yip and Matthews2006), makes it possible to compare children with the same language combination and at the same time circumvent the problem of the cross-linguistic validity of MLU as a measure of morphosyntactic complexity.

As illustrated in the preceding section, dominance has been put forward by some researchers as an explanatory factor for certain instances of cross-linguistic influence (Yip & Matthews, Reference Yip and Matthews2007). Others, however, have demonstrated that cross-linguistic influence can occur independent of language dominance (Hulk & Müller, Reference Hulk and Müller2000), and it has also been claimed that dominance may predict cross-linguistic influence, but the extent to which this is the case is moderated by other factors, such as linguistic complexity (Kupisch, Reference Kupisch2007) and the minority or majority status of the languages in question (Foroodi-Nejad & Paradis, Reference Foroodi-Nejad and Paradis2009). More specifically, Foroodi-Nejad and Paradis investigated the role of dominance as a predictor for cross-linguistic influence in the acquisition of compound nouns in Persian-English bilinguals whose exposure to Persian was restricted to home and who were learning English at daycare. Dominance effects were observed for Persian but not for English, and Foroodi-Nejad & Paradis suggested that this asymmetry may have been due to the majority language status of English and the fact that for all the children, this was an ascending language, whereas Persian was, for some children at least, in the process of stagnating.

Dominance has also been found to correlate with language mixing. That is, bilingual children have been found to use more mixed language utterances in their less dominant language (Kupisch, Reference Kupisch, Guijarro-Fuentes, Larrañaga and Clibbens2008), although again, this effect has been found to interact with other factors, namely sociolinguistic context (Paradis & Nicoladis, Reference Paradis and Nicoladis2007). More specifically, in a study on the language-mixing patterns of French-English bilingual children in a minority-French region of Canada, Paradis and Nicoladis observed that the English-dominant children produced mixed language utterances more frequently in a French context than the French-dominant children did in an English context, reflecting the sociolinguistic characteristics of this community, namely, that almost all French-speaking adults are bilingual, whereas many English-speaking adults are not.

To summarize, while there exists no consensus concerning how best to operationalize dominance, this property of multilingual first language acquisition has been linked to rates of mixing and to the direction of cross-linguistic influence; furthermore, other factors such as sociolinguistic setting have been found to moderate the effects of dominance.

Input Quantity

On the assumption that bilingual children are awake for the same number of hours as monolingual children and that their caregivers vary in talkativeness to the same extent as caregivers of bilingual children, we can safely assume that, on the whole, bilingual children will be exposed to less language input than monolinguals. Furthermore, given the numerous factors affecting bilingual language environments, considerable heterogeneity exists within the bilingual population, too. For example, in a recent study on English-Dutch simultaneous bilinguals growing up in comparable sociolinguistic contexts in the Netherlands, Unsworth (Reference Unsworth2013) found that the average weekly proportion of language exposure in Dutch varied from 8% to 93%. Such variation has been put forward to explain observed differences in rate of acquisition between bilinguals and monolinguals, as well as differences between bilinguals. For example, recent research on multilingual first language acquisition has claimed that differential effect of amount of input on rate of acquisition for general measures of vocabulary and grammar (Barnes & García, in press; David & Wei, Reference David and Wei2008; Hoff et al., Reference Hoff, Core, Place, Rumiche, Señor and Parra2012; Place & Hoff, Reference Place and Hoff2011; Thordardottir, Reference Thordardottir2011), aspects of morphosyntax such as verbal morphology (Blom, Reference Blom2010; Nicoladis, Palmer, & Marentette, Reference Nicoladis, Palmer and Marentette2007; Paradis, Reference Paradis2010; Paradis, Nicoladis, Crago, & Genesee, Reference Nicoladis and Paradis2011) and grammatical gender (Gathercole & Thomas, Reference Gathercole and Thomas2009; Unsworth, Reference Unsworth2013), as well as phonological abilities, including the discrimination of certain phonemes (Sundara, Polka, & Genesee, Reference Sundara, Polka and Genesee2006) and the ability to liaise and elide (Nicoladis & Paradis, Reference Paradis2011).

Studies vary as to whether children's language abilities are measured directly or indirectly, using, for example, parental report measures such as CDI, and whether they concentrate on one domain in depth (e.g., Thordardottir, Reference Thordardottir2011) or compare different domains (Bohman, Bedore, Pena, Mendez-Perez, & Gillam, Reference Bohman, Bedore, Pena, Mendez-Perez and Gillam2010; Gathercole & Thomas, Reference Gathercole and Thomas2009; Hoff et al., Reference Hoff, Core, Place, Rumiche, Señor and Parra2012). Typically, amount of input is estimated on the basis of extensive parental (and sometimes teacher) questionnaires (e.g., Gutiérrez-Clellen & Kreiter, Reference Gutiérrez-Clellen and Kreiter2003; Paradis, Reference Paradis2011) or by using some other measure as a proxy (e.g., vocabulary abilities, see Nicoladis & Paradis, Reference Paradis2011). Other more detailed approaches include asking caregivers to keep a language exposure diary (De Houwer & Bornstein, Reference De Houwer and Bornstein2003; Place & Hoff, Reference Place and Hoff2011) or using new recording and automated analysis technology such as Language Environment Analysis (LENA) (see e.g., Oller, Reference Oller2010, for the use of LENA to measure input in a trilingual setting).Footnote ³

De Houwer (Reference De Houwer2007) assessed the relation between language input patterns and children's language proficiency on a very general level using data from a large-scale survey of family language use in Flanders, Belgium. In the families (n = 1899) examined, De Houwer found that children were most likely to speak both the minority and the majority language when both parents spoke the minority language and at most one parent spoke the majority language at home. Similarly, in a study of the vocabulary development of bilingual kindergarten children in the multilingual context of Singapore, where English is the language of schooling, Dixon (Reference Dixon2011) found that children whose primary caregiver at home spoke English only, or English plus the home language (Mandarin, Malay, or Tamil) scored significantly higher than those children whose primary caregiver spoke the home language only. Children's home language vocabulary was also found to be a significant positive predictor of their English vocabulary, in line with Cummins's (Reference Cummins2000) interdependence hypothesis, the idea that while a child's two languages may have different surface features, they share a common underlying proficiency (see also Scheele, Leseman, & Mayo, Reference Scheele, Leseman and Mayo2010).

One finding replicated in several studies is that differences between monolingual and bilingual children are often restricted to the children's less dominant language; that is, when bilingual children are compared with monolingual children in their dominant language, such differences disappear (Hoff et al., Reference Hoff, Core, Place, Rumiche, Señor and Parra2012; Paradis, Reference Paradis2010; Paradis et al., 2011). In Hoff et al. (Reference Hoff, Core, Place, Rumiche, Señor and Parra2012), for example, bilingual English-Spanish toddlers with at least 70% exposure to English were found not to significantly differ from monolinguals on measures of grammatical complexity and vocabulary (see Barnes & García, in press, for similar findings using CDI data from Basque). Thordardottir (Reference Thordardottir2011), on the other hand, found that the bilingual French-English 5-year-old children in her study reached age-appropriate monolingual norms for receptive vocabulary after having received at least 40% exposure or more to the language in question, whereas for expressive vocabulary, this figure was slightly higher at 60%. Given that these two studies examined different stages of development (toddlers vs. school-aged children) using different methods (parental report vs. standardized tests), it is not clear that their results are directly comparable. One noteworthy aspect of Thordardottir's (Reference Thordardottir2011) study is its novel approach, namely, the use of curve-fitting analysis to detect both linear and nonlinear relationships between input quantity and language outcomes (see also Bedore et al., Reference Bedore, Pena, Summers, Boerger, Resendiz, Greene and Gillam2012).

A factor related to input quantity that is sometimes incorporated into investigations of multilingual first language acquisition is children's output. More in particular, Pearson (Reference Pearson2007) discussed an input-proficiency-use cycle as a means of relating the two, whereby children who hear more input become more proficient in the language in question, and this, in turn, leads to more use, which subsequently invites more input. A number of recent studies have found that output is indeed a significant predictor of bilingual children's language outcomes. For example, Hammer et al. (Reference Hammer, Komaroff, Rodriguez, Lopez, Scarpino and Goldstein2012) observed that both amount of input and output (or usage, in their terms) was significant predictor of bilingual Spanish-English children's vocabulary and narrative skills abilities. These results are broadly in line with those of Bohman et al. (Reference Bohman, Bedore, Pena, Mendez-Perez and Gillam2010), although this study found differential effects on children's input and output across different domains. More specifically, in an assessment of the morphosyntactic and semantic abilities of (a large number of) bilingual children in both their languages (English and Spanish), Bohman et al. observed that input quantity was a significant predictor of whether children scored zero or one or more on a language screening test, which they interpreted as an indicator of initial language learning, whereas output was a significant predictor of higher scores. The authors thus concluded that although input seems to be important in establishing initial knowledge in a language, using the language (i.e., output) is important in adding to that knowledge. Furthermore, different factors were found to affect different domains; for example, initial development on semantics was related to input more than output, whereas morphosyntax was related to both. Bohman et al. suggested that this is due to the need for practice in order to use inflectional morphology productively.

Unsworth (Reference Unsworth2013) also observed differences between different linguistic domains with respect to the role of input were also observed in Unsworth's (Reference Unsworth2013) study on the acquisition of grammatical gender by bilingual English-Dutch children. More specifically, she found that whereas children's scores for gender marking on definite determiners with neuter nouns (gender attribution) were best predicted by amount of input (at time of testing and over time), the only significant predictor for children's scores for gender marking on adjectives (gender agreement) was their scores on definite determiners. Unsworth argued that the effect of input on definite determiners was due to the Dutch gender system, which—as a result of very few systematic cues for neuter—necessitates that children attribute gender more or less on a noun-by-noun basis for neuter, thereby requiring considerable exposure for acquisition to take place, whereas gender agreement, a purely morphosyntactic process, is not (directly) affected by differences in input quantity because once the relevant grammatical features and rules have been acquired, they are applied automatically where required (see also Bianchi, Reference Bianchi2013).

The theoretical importance attributed to the relationship between amount of input and children's language development depends on one's theoretical persuasion. Several studies (Gathercole & Thomas, Reference Gathercole and Thomas2009; Hammer et al., Reference Hammer, Komaroff, Rodriguez, Lopez, Scarpino and Goldstein2012; Paradis, Reference Paradis2010; Paradis et al., Reference Nicoladis and Paradis2011) have argued that input effects in multilingual first language acquisition offer support to the usage-based or constructivist approach to language acquisition (Tomasello, Reference Tomasello2003). Oversimplifying somewhat, the general idea is that if language is acquired in a piecemeal fashion on the basis of the input, as is claimed in this approach, we can expect an effect of relative amount of input for bilingual children for both vocabulary and grammar (see e.g., Gathercole, Reference Gathercole2007). According to these researchers, the observation that such effects exist—and as we have seen in this section, they often do—offers support to this account of language acquisition.

It is, however, important to note that many (but not necessarily all) of the linguistic properties tested in many of these studies, such as verbal morphology and vocabulary, would be expected to involve a significant input component on any approach to language development. In order to systematically investigate the theoretical importance of input in multilingual first language acquisition, and what this means for theories of language acquisition, it is necessary to examine a number of specific linguistic properties from the same and from different linguistic domains within the same children and relate these to particular theories concerning the exact role of amount of input for those specific properties of language. Furthermore, any account that crucially relies on input quantity also needs to be able to explain the similar rates of acquisition observed for bilingual and monolingual children in many cases (Paradis et al., Reference Paradis2011): Even if it is the bilingual children with the most amount of input who are indistinguishable from monolinguals, their total amount of input will still remain significantly lower than that of (many) monolinguals.

In theoretical approaches such as the generative/nativist paradigm, where input is argued to underdetermine the unconscious knowledge children acquire (Chomsky, Reference Chomsky1959), the observation that relative amount of input predicts language outcomes in multilingual first language acquisition presents a challenge. This is not necessarily because there is no role for input on this account—to the contrary, no generativist would deny that input in one (or more) language(s) is crucial for acquisition to take place—but rather, because there is no clear theory on how exactly input interacts with children's innate knowledge such that development occurs.Footnote ⁴

To summarize, faster rates of development in bilingual children often correlate with relative amount of input, and this has been observed for the acquisition of vocabulary and certain aspects of morphosyntax, most typically verbal morphology. This observation has been argued to provide evidence for a constructivist or usage-based theories, but more data from a wider range of linguistic phenomena are needed to test this and other approaches to language acquisition more thoroughly.

Input Quality

Bilingual children's language exposure varies not only in amount but also in type, that is, in terms of quality as well as quantity, and to a certain extent, the two are inextricable (Paradis, Reference Paradis2011). Various factors may contribute to qualitative input differences in multilingual first language acquisition. These include the richness of children's language input, defined as input from a variety of difference sources, (e.g., TV, reading, friends, etc.; Jia & Fuse, Reference Jia and Fuse2007), the variety of speakers providing language input (Place & Hoff, Reference Place and Hoff2011), the types of activities for which the language is used (e.g., Scheele et al., Reference Scheele, Leseman and Mayo2010), whether input-providers speak a standard or nonstandard variety (Cornips & Hulk, Reference Cornips and Hulk2008; Larrañaga & Guijarro-Fuentes, Reference Larrañaga and Guijarro-Fuentes2012), and whether they are native or nonnative speakers (Place & Hoff, Reference Place and Hoff2011), as well as the number and type of literacy-related activities (Scheele et al.). As for input quantity, there are differences between bilingual children that also exist among monolingual children and have been found to affect monolingual language acquisition (e.g., socioeconomic status, Hoff, Reference Hoff2006), and there are also differences that are specific to bilingual acquisition, such as code mixing (Byers-Heinlein, Reference Byers-Heinlein2013). A number of recent studies have started to explore the extent to which variation in bilingual children's input quality affects their language development.

In an investigation of bilingual Spanish-English toddlers using language diary data, Place and Hoff (Reference Place and Hoff2011) examined the relation between children's grammatical and lexical development, as measured by the CDI, and several qualitative input factors, including the proportion of English or Spanish input provided by native speakers, the number of different speakers providing input in a given language. and the number of conversational partners with whom the child exclusively spoke English or Spanish. For Spanish, none of these factors were found to correlate with language outcomes, but as the authors noted, this was likely due to the limited variance in both predictors and outcome variables for this language. For English, all three factors correlated significantly with vocabulary, and importantly, this held when amount of exposure was held constant; for grammatical complexity, there were significant correlations with the number of different speakers and the number of exclusively English conversational partners. Place and Hoff concluded that the finding that variability in speakers is positively related to vocabulary and grammar may reflect the need for “exposure to variability in the signal in order to extract the categories that will support later recognition and production,” as has been suggested for phonological and lexical learning (e.g., Singh, Reference Singh2008, quoted in Place & Hoff, Reference Place and Hoff2011, p. 1847). Furthermore, the findings concerning nonnative input suggest furthermore that “non-native input is less useful to language acquisition than native input” (p. 1847—see also Hammer, Davison, Lawrence, & Miccio, Reference Hammer, Davison, Lawrence and Miccio2009). As the authors noted, however, their study did not identify why this should be the case.Footnote ⁵

The possible effect of nonnative exposure on children's language development has also been investigated for phonemic differentiation. More specifically, Ramón-Casas, Swingley, Sebastián-Gallés, and Bosch (Reference Ramón-Casas, Swingley, Sebastián-Gallés and Bosch2009), in a series of experiments assessing children's ability to detect mispronounced /ɛ/ and /e/ vowels in Catalan, found that whereas monolinguals were sensitive to such mispronunciations, bilingual Catalan-Spanish toddlers and preschoolers—and in particular those with more exposure to Spanish than Catalan—were not. The authors suggested that exposure to similar mispronounced vowels in the accented speech of their nonnative Catalan-speaking parent may mean that bilingual children have “learned to overcome variation in these sounds” (p. 116).

The existence of within-speaker variation is not specific to the bilingual context, and in certain circumstances, exposure to nonnative input may also be a factor to consider in monolingual settings. Exposure to mixed language utterances is, however, almost certainly a characteristic that is exclusive to the multilingual setting. In a recent study, Byers-Heinlein (Reference Byers-Heinlein2013) investigated the relation between exposure to differing degrees of language mixing and children's vocabulary size in English. The findings showed that while controlling for amount of exposure, gender, the child's age, and the balance between a child's two languages in terms of exposure, higher rates of language mixing by parents predicted smaller receptive vocabularies at age 1.5 years and marginally smaller productive vocabularies at age 2. Byers-Heinlein speculated that language mixing may have an adverse effect on children's speech perception abilities, which in turn may mean that new words are less readily detected, which consequently results in a smaller vocabulary size. Whether this explanation holds water upon further investigation remains to be seen, but the Language Mixing Scale that Byers-Heinlein developed and extensively tested as part of this study will surely become a very useful tool for future research into this particular aspect of input quality in multilingual first language acquisition.

To summarize, research on the effects of input quality on multilingual first language acquisition is relatively new and has largely focused on the role of nonnative exposure. The findings available thus far suggest that nonnative input may be less beneficial than native input, and exposure to nonnative input may affect the developmental path which children follow.

Bilingual versus Trilingual First Language Acquisition

Most of the research on multilingual first language acquisition concerns bilinguals rather than trilinguals. The study of trilingual first language acquisition is, however, potentially relevant and informative from both a theoretical and practical perspective. Hoffman (Reference Hoffman2001) distinguished five different types of trilinguals, including children growing up with two languages at home that are different from the language of the wide community, bilingual children who become trilingual via immigration, and third language learners. Given that the topic of this review is multilingual first language acquisition, this section is limited to the first type, (i.e., children who grow up with three languages more or less simultaneously). As we shall see, research on early trilingual development is still in its infancy. Most studies involve observational case studies, often with rather little (linguistic) data (Hoffman, Reference Hoffman2001). Much of the work adopts a comparative approach, contrasting trilingual development with bilingual, and sometimes monolingual, development. Research thus far has focused on two topics: the question of early language differentiation and the potential effect of reduced input on trilingual children's language development.

In a series of articles, Montanari (Reference Montanari2009a, Reference Montanari2009b, Reference Montanari2010, Reference Müller and Hulk2011) provided evidence of early differentiation in the language development of a trilingual Tagalog-Spanish-English child at the pragmatic, phonological, lexical, and syntactic level. The child was exposed to Tagalog by her mother and maternal grandparents, to Spanish by her father and paternal grandparents, and to English from her older sister and indirectly from family conversations, which occurred in English. Her direct exposure to Tagalog (approximately 48%) during the period of study was thus greater than the other two languages (approximately 29% for Spanish and approximately 23% for English). The data in each of these articles concerned her language development up to the age of 2. Montanari (Reference Montanari2009a) examined the child's earliest multiword utterances and found that these followed the appropriate relative argument-predicate ordering in each of her three languages, (i.e., predicate-argument for Tagalog, argument-predicate for English, and both orders for Spanish). Furthermore, the child's mixed utterances are shown to result from vocabulary gaps rather than the use of a unified system, in line with previous research. Quay (Reference Quay2008) also used mixed utterances as a means of showing early language differentiation. Quay found that a trilingual Mandarin-Japanese-English child adapted her rate of mixed utterances and the languages used in those mixed utterances to the language preferences and abilities of her interlocutor. In other words, her different mixing patterns across languages provided evidence of her early language separation (see also Quay, Reference Quay2010).

Montanari (Reference Montanari2010) provided a detailed analysis of the early lexical development of the same trilingual Tagalog-Spanish-English child, showing that translational equivalents were used from early on, as early and at the same rate as has been observed by bilinguals. However, the extent to which the child learned equivalent words in all three of her languages reflected the relative exposure patterns to these languages (i.e., while there were numerous Tagalog-English and Tagalog-Spanish doublets, the number of English-Spanish doublets was limited). Interestingly, the rate at which the child formed a triplet was twice as quick as the rate at which she formed a doublet. This, Montanari noted, is in line with Lanvers's (Reference Lanvers1999) bilingual lexical bootstrapping hypothesis, the idea that knowledge of a concept in one language facilitates the acquisition of the label for that concept in the other. In the case of trilinguals, Montanari argued, familiarity with learning equivalents may have also expedited development of triplet forms. This could be viewed as cross-linguistic influence manifesting itself as acceleration.

Finally, with respect to the child's early phonological development, Montanari (Reference Montanari2011) found that the trilingual child's early word-initial consonants followed the language-specific gestural properties (e.g., appropriate place and manner of articulation) of each of her three languages. Furthermore, the child's consonantal systems were not only differentiated but also fairly advanced when compared with monolinguals. Montanari (Reference Montanari2011) suggested that this may have been due to heightened sensitivity to language and its properties as a result of early multilingual exposure. In a study on the phonological development of a trilingual Spanish-Mandarin-Taiwanese child, Yang and Zhu (Reference Yang and Zhu2010) also observed early differentiation as evidenced by different rates of acquisition in the three languages, sometimes for the same phoneme.

Trilingual children have to divide their time across three languages and thus may have even more reduced input than many bilingual children, at least in one of their three languages. The question of how this comparatively limited input affects trilingual children's language development—as investigated for bilinguals when compared to monolinguals—has also been addressed in the recent literature on early trilingual acquisition. Given that the relative amount of input was not directly reflected in the relative rate of acquisition in the three languages acquired by the child in their study, Yang and Zhu (Reference Yang and Zhu2010) noted that the relationship between amount of input and rate of acquisition is clearly not a direct one. Furthermore, given that the child's development was very much like that of monolingual children with respect to age of acquisition, rate of acquisition, and error patterns, Yang and Zhu stressed that the inevitable reduction in input for trilingual children does necessarily lead to slower acquisition. Similar conclusions were drawn by Montanari (Reference Montanari2009a, Reference Montanari2009b, Reference Montanari2010, Reference Müller and Hulk2011).

In a study on the linguistic development of a bilingual child acquiring Basque, Spanish, and English, Barnes (Reference Barnes2011; see also Barnes, 2006) addressed the role of input in trilingual first language acquisition from a slightly different perspective. She highlighted the fact that many trilingual (and in many cases, bilingual) children receive exposure to one and possibly two of their languages from just one person (i.e., the parent speaking a language that is not spoken outside the home). As such, the child may acquire aspects of the language in question that are marked by the specific characteristics of the person providing the input (e.g., sex and age). In a similar vein to Place and Hoff (Reference Place and Hoff2011), reviewed in the preceding section, Barnes suggested that in order to become fully competent users of the language in question, children may need exposure from a wider variety of sources.

On a more general level, in a large-scale survey on multilingual families De Houwer (Reference De Houwer, Hoffman and Ytsma2004) also analyzed language use in trilingual families (n = 244). In line with the findings for bilingual families, De Houwer found that parental input patterns were related to children's abilities in the two minority languages. More specifically, when neither parent spoke Dutch at home, and both parents spoke both minority languages, there was a three in four chance that at least one of the children in the family spoke all three languages (i.e., the two minority languages plus Dutch). De Houwer's findings for trilingual families are thus in line with those reported above for bilingual families.

The question of whether there is cross-linguistic influence in early trilingual acquisition has received little attention (cf. adult L3 acquisition, Rothman & Halloran, this volume). Interestingly, however, one of the few studies addressing this issue (Kazzazi, Reference Kazzazi2011) examined compound nouns in two Persian-English-German trilingual children and found similar patterns of cross-linguistic influence as observed for bilingual children (Foroodi-Nejad & Paradis, Reference Foroodi-Nejad and Paradis2009). In this particular case, two of the children's three languages (English and German) behave similarly with respect to the linguistic property in question; it is logically possible that more complex situations may arise, where all three languages differ. To what extent such a situation would affect the degree of cross-linguistic influence is unclear, but the variation in language combinations and dominance patterns that the trilingual context offers will surely offer a fruitful area for testing many of the proposals concerning cross-linguistic influence developed on the basis of bilingual data.

To summarize, recent research suggests that trilingual first language acquisition proceeds in much the same fashion as bilingual and monolingual first language acquisition, but the available data are from just a handful of children. As Montanari (Reference Montanari2009a, Reference Montanari2009b, Reference Montanari2010) noted, in order to evaluate the wider validity of the findings and claims observed thus far, more (larger-scale) studies are needed with children learning both typologically related and unrelated languages.