Idiom comprehension in L2 learners

[Brian MacWhinney]

Peter,

Peter,

   Thanks for keeping this thread alive.  It appears that the relation of chunking to the L2 idiom comprehension issue can be seen as the tip of a fairly fascinating (and sometimes dangerous) iceberg.  The danger I am thinking about is the idea that a given population (children, L2, aphasics, Williams Syndrome) relies differentially on chunking.  What are my concerns?  They arise from a consideration of the multiplicity of the factors involved here:
1.  There was a discussion in the 1990s about "the importance of starting small" based on Jeff Elman's simulations stimulated by Elissa Newport's analyses.  On a computational level, Rohde and Plaut (1999) reported problems replicating Elman's analyses. On a conceptual level, I was worried that this theory generated the prediction that learners with the smallest working memories would learn language faster.  On an observational level, there was the literature from Peters, MacWhinney, Vihman, and others emphasizing the idea that children often picked up large chunks or amalgams which they later subjected to analysis.  
2.  There is a literature in SLA that emphasizes the role of big chunks.  This idea aligns well with the less is more notion in some regards, but it has its own problems. The classic here is Pawley and Snyder (1983) which pointed to the ways that second language learners use frames productively.  More recently, Ellis has emphasized the role of chunks in second language learning.  It is clear that both L2 and L1 learners use automatized frames such as "I would like to have a ___" .  But these stretches can also be analyzed by both L1 and L2 learners.  So, is there a fundamental difference here?
3.  There are important effects of levels of analysis for phonology.  As Marilyn notes, once children have acquired a phonological system, they can use this to analyze words segmentally.  Eventually, the rise of phonological awareness and morphological awareness shows how extracted units can break up chunks.  Sapir observed this too in his paper on the psychological reality of the phoneme.
4.  As Diana Van Lanckner Sidtis and others have shown so clearly, the brain provides right hemisphere storage for formulaic language and for intonational packaging.  Lesions to the RH lead to problems with intonation, whereas people with nearly global LH damage may still have some formulas encoded in the RH or perhaps the basal ganglia.  However, whether or not this impacts normal populations differentially is unclear.
5.   Michael Ullman notes that estrogen supports declarative memory, whereas basal ganglia processes support procedural memory.  If we want to associate chunking with proceduralization, then there could be some hormonal basis for sex differences.  However, both men and women produce estrogen - probably enough to keep declarative memory functioning.  
6.  People like Paradis have argued that older learners cannot proceduralize, but Avi Karni has shown that even senior citizens can consolidate procedural memories if they are allowed to take a short nap.  In any case, do we want to argue that older L2 learners do more chunking (i.e. proceduralization) or less and for what?
7.  A major problem involves the level on which we want to envision chunking occurring.  Are we talking only about strings of specific lexical items or would be treat constructions as chunks and would we want to claim that they are differentially impacted for different learner groups?
8.  And then there is the issue of the status of compounds and derivational morphology. Are these chunks or analyzed?  Can't they be both?
9.  It could be that L2 learners use more chunking than L1 learners, given their possibly greater working memory (Halford, Halpern and others), it is imaginable that they can store large pieces. But we also know that they often fail to store word groups as well as L1 learners, perhaps because of their excessive use of analysis during comprehension.  For example, they may tend to learn German nouns without linking them tightly to complete nominal phrases including the determiners and adjectives that encode their gender and case.  There is a huge SLA literature on this topic.
10. Marilyn Nippold has studied idiom comprehension mostly with older children, and I have only read a few of her many papers, but it is clear that idiom comprehension is not complete during early childhood. As she noted in her posting, this places some limits on how we want to think about young children's ability to just pick up formulaic language automatically.  Of course, a lot of this can be influenced by literacy, input, and so on.
11.  An then there is the issue of really big chunks such as Homer's memory for the Iliad or the German girl who can give you the sentence before and after any given sentence from Goethe.  Does this type of chunking have anything to do with all the other types mentioned above?

Glancing across this complex territory, it seems to me that one should be careful about imagining that, in general, L2 learners use more chunking than L1 learners and therefore learn idioms more quickly.  Given their possibly greater working memory (Halford, Halpern, and others), it is imaginable that they can store large pieces, but probably these are also analysed. So, when we come to the issue of learning idioms and other frozen forms of the type described by Wray, do we want to think of these as chunks or as non-compositional groupings of independent lexical items? And if we actually get solid data pointing to any population differences, do we then have evidence to attribute this to neurology, as opposed to language support?   And do we want to differentiate any of these possible differences in terms of whether the chunking applies to phonology, lexicon, morphology, syntax, speech acts, or conversational patterns?

The theories of L1, L2, and language disorders all need to deal with these issues, but in doing so, they will need to break up the concept of "chunking" into the many pieces and dimensions of which it is composed.  And they will need richer longitudinal data to study the development of chunking and analysis across all of these linguistic levels and structures in greater detail.

-- Brian MacWhinney


-- 
You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.
To unsubscribe from this group and stop receiving emails from it, send an email to info-childes+unsubscribe at googlegroups.com.
To post to this group, send email to info-childes at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/info-childes/268B2D6A-560F-48A6-8B9D-40A953ED86C0%40cmu.edu.
For more options, visit https://groups.google.com/d/optout.