Wednesday, 9 December 2009

Processing two structures simultaneously
Although public opinion is still coming round to the idea, Bilingualism does not, in fact, impede cognitive and linguistic development, but may enhance it. Bilinguals have been shown to be more aware of the use of words in social contexts (Rosenblum and Pinker, 1983), they are better at taking other speaker’s perspectives (Genesee et al., 1975), and better at monitoring the knowledge state of others (Genesee et al., 1996). Recently, Kovacs and Mehler (2009) showed that bilinguals are more flexible at processing linguistic structures.

Kovacs and Mehler (2009) run an eye-tracking experiment on infants. Children had to learn to associate three syllable words with either AAB (e.g. 'babaka') or ABA ('bakaba') structures with a stimulus appearing on either the right or the left side of the screen. Bilinguals successfully learned to associate both structures with the correct side, while monolinguals performed equally well for AAB structures, but could not learn the associations for ABA structures. (See Ed Young's blog for another analysis). This is a bit confusing - it's not just that monolinguals were worse than bilinguals at processing ABA structures, monolinguals actually scored negatively!

Bilinguals have a wider range of input than monolinguals, and so may be more used to different linguistic structures. Bilinguals may be better at processing two linguistic structures simultaneously. Alternatively, monolinguals may have refined their processing to fit their input language.

Part of Kovacs and Mehler’s argument relies on AAB structures being easier to process, referring to Gervain et al. (2008). Presumably it’s true that this is because less memory is needed but Gervain et al. find that ABB structures, not AAB, are easier to process than ABA structures in monolinguals. One question is whether this processing benefit is language-specific. If it is, then the result may reflect bilingual’s experience with ABA strucutres. If it is not, then the result may show either a faster maturation of processing abilities in bilinguals, a slower maturation of processing abilities in monolinguals or a difference in investment of resources in the domain of word learning.

Let's look at some data! Proportions of various structures of 3 syllable words were gathered for English, Dutch, German (CELEX), Mandarin Chinese (CC-CEDICT, taking tone into consideration) and Italian (CoLFIS, although orthographic and automatically parsed for syllables):

The graph above shows that there is not much variation between proportions of word forms with AAB and ABA structures within a language. This suggests that bilinguals do indeed have a processing advantage. However, there is variation in the number of tokens of AAB and ABA across languages, and large variation in the proportions of word forms with ABB structures compared to other structures. If this is the case, then bilingual infants may be exposed to (and therefore be more effcient at processing) a greater range of structures, which may be an additional factor in breaking the Mutual Exclusivity bias.

Furthermore, the proportions of words which conform to either AAB, ABB or ABA structures are very small. Why, then, are these structures used frequently in infant-directed speech (Ferguson, 1983)? A possible answer is a kind of explicit demonstration of linguistic structure. At any rate, Kovacs & Mehler's paper is interesting, as is it's companion paper, described here.

Below, I've split the data from the graph above by counts for Adjectives, Nouns and Verbs. Interestingly, German, Dutch and English have different distributions. For instance, a new word with an ABA structure (e.g. 'bakaba') may be interpreted differently by different speakers (rationally, ignoring heuristics based on morphological cues). An English speaker would assume it was an adjective, a Dutch speaker would assume it was a verb and a German speaker would assume it was a noun:

This would show that different languages have different ways of cuing children into the meanings of their words. Unfortunately, perhaps because of the very low numbers of examples, when I tested some English and Dutch speaking friends on this, they basically chose randomly. Another big, unfounded linguistic theory.

(also, see my followup post here)

Kovacs, A., & Mehler, J. (2009). Flexible Learning of Multiple Speech Structures in Bilingual Infants Science, 325 (5940), 611-612 DOI: 10.1126/science.1173947

No comments:

Post a Comment