Does the empirical evidence support domain general learning mechanisms in first language acquisition?

2018 edition


It is fascinating that children learn their native language in the space of a very few years, at an age when they are unable to master mathematics or other cognitive skills of comparable complexity. There is no doubt that a normal child from Spain raised in China would readily learn Chinese, or vice versa, regardless of individual differences in intelligence and other talents. Thus, the question “how do children acquire their native language?” has prompted a lively theoretical debate and a great deal of empirical works. Probably, language is acquired via a general mechanism of gradual generalization with memorialization on the basis of input as assumed by some usage-based approaches (e.g., Abbot-Smith, Lieven, & Tomasello, 2001; Tomasello, 2000).

However, if we adopt this kind of view, we should face the problem of explaining the uniformity of developmental stages in child language, namely, why all the children in the world acquire their first language in similar manners: infants babble around 6 months and they know most of the grammar before they are 5 years; and if input is the only factor that matters, we should be able to explain why a language cannot be acquired simply by looking at TV programs (Kuhl, Tso, & Liu, 2003), or why children can acquire a structure at a very early stage even the structure is scarce (Belletti, 2017; Friedmann & Costa, 2011; Gavarró, Lewandowski, & Markova, 2010; Zhu & Gavarró, 2017). Besides, many works have pointed out the ability of humans to “create” a language as the creation of the Nicaragua Sign Language by children in the course of only several years (Senghas, 1995).

Moreover, although children make mistakes, those mistakes are far from random. Children will never say the phrase like ‘John the washed car’, in which the determiner ‘the’ is separated by the verb. Catalan-speaking children tend to omit object clitics, while omission is very low in Spanish-speaking children even they are considered as two closely related languages (Gavarró, Torrens, & Wexler, 2010). Deaf children who received no conventional linguistic input could converge to a personal communication system that exhibited a high degree of regularity and language-like structure (Goldin-Meadow, 2005; Coppola & Newport, 2005). Everything seems to indicate that there may exist some principles which are domain-specific that guide child language acquisition.

In this talk, I will present some evidence using eye-tracking methods to show that infants, even those are younger than 18 months of age are able to ‘know’ that sentences they are acquiring have a hierarchical structure and there is an abstract parametric grammatical knowledge, independently of the lexicon, which permits children to set correctly their grammar on the basis of limited input, rather than progressively on a lexical basis.