INTRODUCTION: Recent developments in early bilingualism
This Special Issue (SI) is dedicated to early bilinguals, who acquire two languages during early childhood, before age 6, simultaneously (2L1 bilinguals) or sequentially (cL2 bilinguals).Footnote 1 Recently, the notion of heritage speakers (HSs) − bilinguals who grow up speaking a minority language at home − has become prominent in this context. HS research has typically targeted bilinguals at a mature (adult) state, but early developing bilinguals may of course be HSs too, though not uniformly labelled as such. HSs of a moribund language or variety are another type of early bilingual, representing the final or penultimate, often 4th or 5th, generation of speakers. Unfortunately, ‘deficiency’ or ‘incompleteness’ is a common thread linking much HS research − despite a wealth of evidence demonstrating HSs’ maintenance of complexity in many grammatical domains along with differences to monolinguals (see, e.g., Kupisch & Rothman, Reference Kupisch and Rothman2016; Putnam & Sánchez, Reference Putnam and Sánchez2013). This SI brings together studies on how and why the morphosyntax and phonology of early bilinguals might differ from that of monolinguals.Footnote 2
While it seems uncontroversial that (2)L1 and eL2 children tend to outperform L2 adults, it remains controversial whether such age effects are due to a sensitive period (e.g., Muñoz & Singleton, Reference Muñoz and Singleton2011). Relevant research has typically compared early and late bilinguals during adulthood and resulted in various proposals as to the age at which the ability to acquire an L2 to a native-like degree declines (e.g., Abrahamson & Hyltenstam, Reference Abrahamson and Hyltenstam2009; Johnson & Newport, Reference Johnson and Newport1989). Relatively fewer studies investigated age effects during child development, some claiming that children with an AoO after 4;0 will acquire their L2 in a similar fashion to L2 adults but fundamentally differently from children with an earlier AoO (e.g., Meisel, Reference Meisel, Haznedar and Gavruseva2008, Reference Meisel2009; Rothweiler, Reference Rothweiler and Lleó2006). Sometimes, these findings are argued to be consistent with the notion of a sensitive period after which native-like acquisition gradually becomes impossible. Meisel's (Reference Meisel2018) study is representative of this view. He studies similarities and differences between (2)L1 and cL2 acquisition, claiming that AoO is crucial in determining the course and result of early bilingualism. cL2 learners count as qualitatively different from (2)L1 learners if (i) producing constructions absent from the speech of (2)L1ers, (ii) proceeding through distinct developmental sequences, or (iii) ultimately attaining a different grammatical knowledge system. Evidence is provided from grammatical gender in cL2 French children (L1 German), where children with an AoO before 3;6 resemble L1 learners and children with a later AoO behave like adult L2 learners.
However, it has long been noted that other factors may co-vary with AoO, including proficiency and language dominance (e.g., Treffers-Daller & Silva Corvalán, Reference Treffers-Daller and Silva Corvalán2015), relative L1/L2 use (e.g., Flege, Frieda & Nozawa, Reference Flege, Frieda and Nozawa1997) or differential amounts of input (e.g., Unsworth, Reference Unsworth2016). These factors are closely related, since language dominance is often operationalized based on proficiency, and the dominant language coincides with the language used and/or heard more frequently. Many studies on mature HSs have indeed shown that adult HSs pattern with L1 monolinguals in their dominant language, while often differing from monolinguals in their heritage language, which they use less frequently (e.g., Bianchi, Reference Bianchi2013; Kupisch, Reference Kupisch2012, Reference Kupisch2014). Only few studies have examined both the role of input/exposure and that of AoO. Granfeldt's (Reference Granfeldt2018) study provides one example. Like Meisel, he focuses on French gender in cL2 acquisition, but argues that the process is not determined by AoO alone. He compares (2)L1 and cL2 learners with an AoO below and above 4;0 in terms of three aspects of gender: (i) discovery of the abstract gender feature, (ii) gender assignment, (iii) article-adjective concord. The cL2 learners discover the abstract gender feature quickly, but gender concord is affected by AoO, as the 2L1ers perform monolingual-like, while the cL2ers produce qualitatively different errors. For gender assignment, however, more input leads to earlier acquisition in both eL2ers and 2L1ers. Thus, AoO and input affect linguistic properties selectively (see also Unsworth, Reference Unsworth, Grüter and Paradis2014; Unsworth, Argyri, Cornips, Hulk, Sorace & Tsimpli, Reference Unsworth, Grüter and Paradis2014).
While Meisel and Granfeldt focus on AoO, all remaining contributions investigate the role of cross-linguistic influence (CLI). CLI is typically seen as implying a mutual rather than unidirectional influence of two languages (Sharwood-Smith & Kellerman, Reference Sharwood-Smith, Kellerman, Kellerman and Sharwood-Smith1986). Müller (Reference Müller1998) paved the way for much research on CLI, proposing that transfer does not necessarily operate from dominant to weaker language, but may be a relief strategy for learners to cope with structurally ambiguous input: If language A has two possible structures for one phenomenon and language B shares one of them, the structure that is not shared will be abandoned, leading to more similarity between the two languages (cf. Döpke, Reference Döpke1998). Hulk and Müller (Reference Hulk and Müller2000) further confined the necessary conditions for CLI, proposing that, besides structural ambiguity, the syntax-pragmatics interface must be involved. They viewed CLI as quantitative, leading to accelerated or delayed development compared to monolinguals. Some current research, elaborating on Paradis and Genesee (Reference Paradis and Genesee1996), also interprets acceleration and delay as quantitative effects, while reserving the notion of transfer for qualitative effects.
The remaining four contributions partially build on these ideas. Egger, Hulk and Tsimpli (Reference Egger, Hulk and Tsimpli2018) also study grammatical gender, but addressing 2L1ers and taking both languages (Dutch, Greek) into account. Greek and Dutch represent extreme opposites in the frequency and transparency of gender marking cues: as the morpho-phonology of Greek nouns provides clear cues and gender is early acquired, while cues in Dutch are limited and gender late acquired. The study shows that in Dutch, Greek–Dutch 2L1ers assign gender on par with age-matched Dutch monolinguals, while displaying an advantage over monolinguals with adjective concord. The authors argue that Greek, having a more transparent gender system, accelerates gender discovery in Dutch. Thus, CLI is determined by language specific properties. Lleó (Reference Lleó2018) studies the phonological development of Spanish in German–Spanish 2L1ers, focusing on spirantization and place assimilation in nasal codas − both non-existent in German. When the data collection stops, the 2L1ers have not yet acquired these phenomena, resorting to gap-filling transfer from German instead, e.g., producing stops instead of continuants. The acquisition of the Spanish phonological system thus appears to be incomplete, but, as the author stresses, incompleteness need not be a final state. Kehoe (Reference Kehoe2018) examines the acquisition of /r/ in both languages of German–Spanish 2L1ers. The /r/-sound is a uvular approximant [ʁ] in German, while Spanish has an alveolar tap [ɾ] and a trill [r]. The acquisition of /r/-sounds is phonetically and phonologically complex, making delays or transfer in bilinguals expectable. The 2L1ers acquired the German uvular /r/ akin to monolinguals, but with a slight delay in clusters. In Spanish, the 2L1ers generally outperformed monolinguals, despite the occurrence of some substitutions. The findings thus show CLI resulting in an approximation of the two phonological systems.
Kehoe's study nicely illustrates that bilingual children may maintain or enlarge contrasts between categories in their two languages or lose the contrast; Flege's (Reference Flege and Strange1995) notions of merging and deflecting depict this well, while acceleration and delay are more relevant when studying the languages of a bilingual individually. The idea of deflecting resonates in Anderssen, Lundquist and Westergaard (Reference Anderssen, Lundquist and Westergaard2018), whose point of departure is that CLI is likely to occur when there is structural ambiguity. Such a scenario is represented by possessive DPs in English-dominant HSs of Norwegian. English only allows Poss-N structures, while Norwegian allows both Poss-N and N-Poss, thus foreshadowing overuse of Poss-N. However, Norwegian HSs overuse N-Poss rather than Poss-N, although the latter is perfectly possible in Norwegian, thereby ‘over-inhibiting’ the structure that is similar in their two languages. The ability to inhibit the dominant language appears to depend on the speaker's proficiency, with inhibition of dominant-language influence being harder for less proficient HSs.
Taken together, contributions to this Special Issue paint a complex picture, pinpointing the factors minimally at play when two languages are in contact, though comparing individual trajectories suggests we are far from predicting what will happen in each individual case.
