clark theoretical comments

Framing in the first chapter. You present the general framing of the two major ‘camps’ in acquisition first (p. 13) in terms of domain-specific nativist proposals vs. domain-general empiricist proposals. But you don’t clearly point out that domain-specific nativist or innatist proposals from the start have focussed only on the acquisition of syntax, while empiricist proposals typically have looked at language as a whole––with a range of studies covering phonology, lexicon, and morphology, in addition to syntax. I wonder whether you might not want to offer a clearer take here, perhaps with the various ‘labels’ that have characterized these two main approaches over the years (since the 1960s) summarized in a small table.

Then you can list the general properties of nativist, Chomskyan, approaches (syntax is innate; principles and parameters; focus on syntax alone; assumptions of poverty of the stimulus + no negative evidence, with an absence of data supporting the last two assumptions), vs. mainly more recent constructivist (constructionist) and emergentist approaches that have been firmly based on data from acquisition and in particular data from adults speaking with children [e.g., Tomasello 2003; Clark 2016, 2018b], so all empirically-based, with much of the research focussed on the acquisition of the lexicon and on how constructions (syntax) are associated with that lexicon, with emphasis on conversational interaction as the context for learning language as a whole. Something like this would be useful for people picking up your book, and not knowing the history of this field.

The demographic effects (sex, birth order, and SES as assessed from mothers’ level of education) line up nicely with other studies of SES x language more generally (Hart & Risley; Hoff; Huttenlocher), so CDI scores offer useful supporting evidence for such effects. One other element that appears to be a birth order effect that might be worth mentioning is found in the acquisition of first and second person pronouns (see Clark 1978; Oshima-Takane 1992, Oshima-Takane et al. 1996): younger siblings ‘get’ the I vs. you contrast earlier than first-borns, in part it seems because they see and hear the shifting reference involved, as they observe their older sib talking to adults, and so don’t make the error of treating I = adult and you = child, for example, when they first start to use these pronouns (at around 20-24 months).

Processing universals. Your emphasis on processing is very important, and I was happy to see how it echoes the distinction we drew in 1977 between process and product in talking about how language is understood, produced, and acquired. But you don’t actually say very much about what processes might be in play when it comes to language acquisition. I think this is in part because you don’t know the phonetic forms of children’s early words in production (cf. Scollon, 1976): all we have with CDI data is parents’ identification of the most probable adult target word for each of their young children’s productions….

I wonder whether it might be appropriate here to refer to Slobin’s work on Operating Principles (1985). Like you, Dan Slobin looked at consistent patterns in acquisition across a variety of languages and extracted the main generalizations that appeared in children’s comprehension and production of each language (but mainly production), at different stages, in order to look for universal processing strategies for dealing with the stream of speech. This might allow you to say a bit more about specific processes in relation to language acquisition as a whole, for instance, in your final chapter (ch. 16).

And while you do allude to social factors affecting acquisition (e.g., the presence of terms for people in very early vocabularies), you don’t actually take up a central question there—namely how speaker and addressee take account of what the other knows when, their common ground and how that common ground is accumulated––actually another shortcoming of CDI data––, you make very little of the cognitive basis for decisions about what words can refer to––the setting up of categories, generalizations from just one or a few instances of a category, say, and hence the fact that children’s words pick out types (rather than individuals). In principle, children could start with the theory that there’s a word for every item…but they don’t.

You also make little of the conversational contexts for language acquisition (Tomasello 2003; Clark 2016), ––of course, this is one of limitations on using CDI data––and how feedback, both positive and negative can inform children, at the level of their having succeeded in understanding someone, or of their getting their own intentions over to someone else. In short, I think it might be worth some acknowledgement of cognitive and social influences on early language acquisition here too.

One other factor that I think could be underlined, especially when you refer to experimental studies and to corpus analyses as complementary to your CDI analyses (see also Monaghan & Rowland 2017), is the actual age or age-range of all the children involved. For instance, in the Conwell & Demuth study of datives that you refer to, the children tested were aged from 3 to 5 years old, so already beyond the scope of the CDI data. And even in the CDI data, age makes a big difference: for example, if you collapse over age, you arrive at your finding for a 24-month-old of around 100 words in production and the start of word combinations. But it’s equally important here to include the age range: for combinations, for instance, the actual range is from about 13-14 months (appearance of the first, rare, ones) to 24 months. So age ranges here are important for making comparisons.

langcog / wordbank-book

clark theoretical comments #62