The question of how much language a child can learn by modelling the patterns she is exposed to in the environment represents an age-old and persistent controversy in language-acquisition research. On a number of occasions researchers felt confident enough to claim that the controversy had been resolved, in favour of their own viewpoint, of course. One such attempt was Skinner's (Reference Skinner1957) behaviourist view of language development, which gave a prominent if not exclusive role to input and experience. After Chomsky's (Reference Chomsky1959) landmark review, however, Skinner's account was left in pieces and was not further pursued by many. A more recent attempt comes from usage-based accounts according to which children directly build linguistic categories and rules from the language they hear around them. Some believe that these accounts have resolved the controversy for good and have ‘overturned’ alternative less experience-driven approaches (e.g., Ibbotson & Tomasello, Reference Ibbotson and Tomasello2016; Dabrowska, Reference Dabrowska2015). Yet, casual inspection of recent research articles reveals that this announcement may be somewhat premature; see, for example, Everaert, Huybregts, Chomsky, Berwick and Bolhuis (Reference Everaert, Huybregts, Chomsky, Berwick and Bolhuis2015), Boxell's (Reference Boxell2016) rebuttal of Dabrowska (Reference Dabrowska2015), and many acquisition studies published in the journal Language Acquisition.

In research on bilingual language acquisition, the question of the role of input and exposure also features prominently, with similar controversies as those mentioned above. While some researchers use measures of linguistic input and exposure to predict a child's proficiency in two or more languages (see, for example, Grüter and Paradis’ (Reference Grüter and Paradis2014) collection of articles), others note that bilingual (like monolingual) children's linguistic knowledge goes beyond what is directly available through input and exposure (e.g., Meisel, Reference Meisel2011). Against this background, our keynote article (Carroll, Reference Carroll2017a) offers a critical review of recent research on the topic. Carroll points out a number of caveats against over-enthusiastic claims on how input and exposure determine bilingual language development. One common type of measure provided in favour of input-based accounts is positive correlations between a bilingual child's linguistic outcomes and corresponding features in her environment, which are supposed to have explanatory or predictive value. Hoff, Welsh, Place and Ribot (Reference Hoff, Welsh, Place, Ribot, Grüter and Paradis2014), for example, claim that differences in the relative quantity and quality of input explain individual differences among bilingual children (our emphasis). Carroll argues that such claims are overstated, because correlations may not necessarily reflect causal links. Another concern is related to the notion of ‘language’ employed in this literature and to the measures used to tap into ‘language’. Carroll argues that the mental system underlying the knowledge of language(s) is complex and consists of distinct subsystems (e.g., lexicon, grammar, phonology, etc), and she reminds us that the commonly used vocabulary measures do not provide measures of ‘language’, but only of vocabulary. Consequently, findings regarding the relevance of input and exposure to vocabulary may not generalize to other subsystems of linguistic knowledge.

As expected, Carroll's keynote article elicited a vivid – even quite fierce – response, from 10 commentators. While many commentators agree with the caveats Carroll pointed out (De Houwer, Reference De Houwer2017; MacWhinney, Reference MacWhinney2017; Pérez-Leroux, Reference Pérez-Leroux2017; Armon-Lotem, Reference Armon-Lotem2017), several other commentators note that current research on input and exposure in bilingual language development already goes beyond these concerns. Grüter (Reference Grüter2017) points to studies showing effects of input and exposure on bilingual grammar development. Likewise, Paradis (Reference Paradis2017) mentions cases in which fine-grained properties of bilingual language use can be predicted from input data, in this case from parental reports. Gathercole (Reference Gathercole2017) criticizes the keynote article for not representing the field accurately, pointing to a number of additional studies to which the caveats do not seem to apply. Weisleder (Reference Weisleder2017) maintains that the amount of exposure has ‘considerable explanatory power’. In addition, a number of commentaries mention other studies that are of interest to the broader topic. Armon-Lotem (Reference Armon-Lotem2017) points to research on bilingual children with SLI, Mougeon and Rehner (Reference Mougeon and Rehner2017) to research on the influence of classroom input versus community exposure on bilingual language development, and Bernardini (Reference Bernardini2017) to studies on the ‘weaker’ language of bilinguals. In her response, Carroll (Reference Carroll2017b) clarifies her points and discusses what she believes are misunderstandings. What remains is the impression that the study of the role of input and experience in bilingual language development is challenging, that broad conclusions – ‘exposure is (not) critical’ – are probably wrong, and that progress can be made by asking more subtle questions, such as how exposure affects different kinds of linguistic knowledge, different kinds of bilinguals, and what the mechanisms are by which the language learner employs information available from the linguistic environment.

We hope our readers will enjoy the keynote article together with the commentaries and the author’s response as well as the interesting regular research articles and research notes presented in the current issue.