What can adult speech tell us about child language acquisition?

1. Frequency information in corpora and its use in acquisition studies

This contribution explores a methodological problem in language acquisition studies. Much research in language acquisition has shown that children use statistical learning as a strategy in the acquisition of their native language (Saffran et al. 1996 and many others). Frequency of occurrence is also believed to determine the order of acquisition of phonological structures in the construction of the grammar (Boersma and Levelt 2000, Levelt et al. 2000, van de Weijer and Sloos 2013). How do we obtain the relevant frequency information for acquisition studies?

Ideally, we should take into account children’s speech or child-directed speech (CDS), depending on the purposes of the investigation. Investigations into the construction of the lexicon and acquisition of the grammar depend on the input, the perception, and the lexical storage of the child, and therefore, frequency data on CDS seem most desirable. CDS has been investigated to study, for instance, (i) the effect of frequency on the acquisition of verbs and verbal inflection in English (Cameron-Faulkner et al. 2003), (ii) phonological variation in Tyneside English (Foulkes et al. 2005), and (iii) the joint effects of frequency and markedness on language acquisition (Stites et al. 2004). Unfortunately, frequency counts of CDS are not widely available. Although this gap is slowly being filled (for example as a result of the CHILDES database project (MacWhinney 2000)), many languages are still underrepresented, and some data are collected only for specific purposes. This has led researchers to consult other sources, like adult-directed speech or text corpora. This paper addresses the question of whether it is methodologically adequate to use adult-directed speech (ADS) in such language acquisition studies.

Of course, it very much depends on the topic of investigation whether ADS can be used in L1 acquisition research. As is well known, CDS differs considerably from ADS. For example, nouns are used more often in CDS than in ADS and the reverse holds for verbs (that is, nouns are acquired before verbs (Gentner 1982)); prosodic differences are exaggerated; sentences are shorter (sometimes only isolated words are used); and new words tend to be located in final position (Dominey and Dodane 2004, Ferguson 1964). If we are interested in the lexico-semantic, syntactic, and morphosyntactic aspects of acquisition, a CDS corpus therefore seems indispensable. This is supported by the fact that Goodman et al. (2008) found only weak correlations between the age of acquisition of lexical items and frequency in ADS corpora. However, we would like to argue that in research of the acquisition of phonological structures (such as natural classes of sounds, syllable structure, and tone), an ADS corpus may serve as a good substitute if no (adequate) CDS corpora are available. The reason for this hypothesis is that the frequency of occurrence of different phonological structures is not likely to differ very much between different speech styles (specifically, ADS vs. CDS). Consider, for instance, languages with complex onset clusters or nasal vowels, which are usually assumed to be “marked” kinds of structures and less common than single-consonant onsets and oral vowels. If these occur in 20% and 10% of the words in the lexicon, respectively, we expect, roughly, the same relative occurrence for CDS as for ADS. That is, although in CDS, some words may be simplified, we do not expect consistent simplification of phonotactics in CDS.

We are not aware of studies that investigated the relation between CDS and ADS for methodological purposes. The comparability of CDS and ADS is, however, addressed in Levelt et al. (2000), who investigated the order of acquisition of syllable structure (CV, CVC, CCVC, VC) in Dutch, assuming a crucial role for frequency in the order of acquisition of the different syllable structures. They first based the frequency counts of the different syllable types on a written database (CELEX; Baayen et al. (1993)) and found that the order of acquisition did not fully match the relative frequency of occurrence of the syllable structures. When the authors compared the order of acquisition with the frequency of...