On Language : in response to Ben Zimmer

Sep 25 2010 Published by under Forget What You've Read!

Last week, Ben Zimmer published an article in the NY Times on chunking in language.  This morning, I received an email in my inbox from one Prof Plum, who was writing to a parent (a good friend of his) to explain what was right -- and what was wrong -- with the article.  I was CC'd, as Plum thought I might enjoy the explanation.  I did!  With permission, I have included it below for any interested readers.  But first, here's Zimmer :

In recent decades, the study of language acquisition and instruction has increasingly focused on “chunking”: how children learn language not so much on a word-by-word basis but in larger “lexical chunks” or meaningful strings of words that are committed to memory. Chunks may consist of fixed idioms or conventional speech routines, but they can also simply be combinations of words that appear together frequently, in patterns that are known as “collocations.” In the 1960s, the linguist Michael Halliday pointed out that we tend to talk of “strong tea” instead of “powerful tea,” even though the phrases make equal sense. Rain, on the other hand, is much more likely to be described as “heavy” than “strong.”

A native speaker picks up thousands of chunks like “heavy rain” or “make yourself at home” in childhood, and psycholinguistic research suggests that these phrases are stored and processed in the brain as individual units. As the University of Nottingham linguist Norbert Schmitt has explained, it is much less taxing cognitively to have a set of ready-made lexical chunks at our disposal than to have to work through all the possibilities of word selection and sequencing every time we open our mouths.

Plum's response covers why word-chunks are better thought of as a skill than an inventory, how we unconsciously change our speech when we talk on cellphones, and how we can learn verbal humor.

Prof Plum, September 25, 2010

It's great that this kind of research is getting attention, but to be honest, the whole set-up is just wrong.  It starts with "storing chunks," as if what kids learn is an inventory, rather than a skill.

Think about tying shoelaces. We all have our normal way of doing it, but we can change how we do it too. It's just that our normal way is practiced, and so we find it easy.  Ditto language.  The chunks aren't "chunks" -- they are strings of sounds / meanings that kids hear more often, and so they have learned that -- given one part -- the rest is more likely to follow.  When you hear "can we just get to the..." the word "point" is really likely, whereas "destination" is less likely (though it could go there too). Just as we find our normal way of tying shoe laces easier than a different way, kids will find more practiced bits of language far easier to work with than less practiced bits.

If you think about it this way, then what kids are doing is not so much storing words / chunks, but rather learning about the likelihood of what they might be hearing / saying, and the likelihood that it has a given meaning.  --Meaning it's hardly surprising that they are better at using and understanding the more practiced bits than the less practiced bits.  And of course, because adults do this too, over time, some "chunks" become more and more likely than others.

Try this: listen to the kinds of things people say on cell phones.  --Because talking on a cell phone is an awful (noisy) communication channel, we tend to use frequently-used, predictable phrasing to make ourselves understood.  Or else, if people haven't figured this out, we end up asking them to repeat themselves a lot, or holding the phone at arms length.  (This is also why it's so much easier to talk to people you know well on the phone -- they are far more predictable, so you can fill in for them more.)

What all this means is that we shouldn't be talking about teaching kids so that they "know" this word or that chunk -- or talking about them "knowing" words or bits of grammar even. Instead, what we should focus on is what they should be practicing so that they can get *better* at their language skills. (My guess is that parents are far more aware that language is an incremental skill than are the people who claim to be experts.)

The highlight of my last week was getting a paper from a guy who's the best math modeler working on language. He just applied our approach to a pretty massive set of data on reading (including data from serbian, which is one of those horrible languages filled with affixes and suffixes that make an English speaker think, "wtf?"). By taking the view that people don't store "words," but instead that they learn about the likelihood of letter collocations predicting meanings (which then predict the "sounds" that we hear in our heads when we read), he was able to make sense of all that data far better than anyone has ever done before.

It's funny -- I was talking to L's younger brother about quips this morning.  I said something amusing and he was impressed, and I was explaining that it was basically something I'd heard before, found funny, and reused because it was appropriate; as far as I can tell, being able to make "spontaneous" funny comments is mostly about noticing when other people say funny things, and borrowing or modifying them later.  Eventually, as your stock of quips grows, it gets easier and easier to say "spontaneous and original things."  I think language works in much the same way -- and if you think about it this way, the article in the Times sets up a dichotomy that doesn't really exist, because how much language kids hear and use is a lot more important than the details of whether it is presented in this way or that way.

This is also why I think all that stuff about not having kids watch TV is so silly.  When A-L watches "Eloise at the Plaza" she gets exposed to a ton of language, and it's pretty clear she understands a lot of it, and that she has come to understand more of it as time has gone on. Which means she is getting practice, and learning, which can only help!  (Though maybe she doesn't need any more practice on the shrieks 😉 ).

Past Posts on this Topic :

Mollies, PokeBalls, and Naked Ladies : A Topsy-Turvy Lesson in Learning Words from Context

Why are Zipfian distributions found in language?

The Long Tail of Language

8 responses so far

  • [...] This post was mentioned on Twitter by ScientopiaBlogs, Melodye. Melodye said: A response to Ben Zimmer's NY Times article - on chunks in language http://bit.ly/b6MKlA [...]

  • [...] the original post here:  On Language : in response to Ben Zimmer | Child's Play By admin | category: University of NOTTINGHAM | tags: compared-the-cases, [...]

  • This makes me think all the more that you guys ought to read up on Construction Grammar, which is very much focussed on co-interpretations of combinations of words, as for example in the 'WXDY' construction illustrated here:

    Diner: Waiter, what's this fly doing in my soup?

    Waiter: Madam, I believe that's the backstroke

    Reference: Kay, Paul and Charles J. Fillmore (1999) 'Grammatical l Constructions and Linguistic Generalizations: The What's X Doing Y? Construction', Language, Vol. 75, pp. 1-33.

  • [...] On Language: in response to Ben Zimmer [...]

  • Coturnix says:

    Of course, as Serbian is my mother tongue, I tend not to think of it as "one of those horrible languages", though I did think that way about English until I got a bit more fluent in it 😉

    • Well, Kayardild is in certain respects like Serbian (and Ancient Greek, Old English, Icelandic etc) on steroids and spike heels.

      So, for a gross first-approximation-like oversimplification the way you say: we will catch fish with that man's good net is:

      we catch-FUT fish-FUT that-POSS-INST-FUT man-POSS-INST-FUT good-INST-FUT net-INST-FUT.

      1. most members of the 'conceptual' verb phrase carries future marking (or past, if it's past, etc, but there's about 11 tenses)

      2. the object also carries the tense, although not with the verbal morphology but with a different one that is not entirely congruent with the verbal one, so fish is really better glossed fish-FUT(NOM)

      3. the instrument carries INSTRumental case, and then the tense

      4. all members of an NP replicate the markings assocated with that NP, so 'good' also gets INST-FUT

      5. a submember of an NP gets one of various cases, here POSS, and then its containing NPs cases get tacked on after that.

      6. by (4), a complex submember sport all of the revelvant cases.

      This is enough to understand the example sentence

      7. for the adverbial cases, such as instrumental, there's usually a choice between a 'nominal' form and 'verbal' form, the former being followed by the affixes for the nominal version of the tense system, the latter by a copy of the regular verbal morphology. The 'verbal' cases are historical verbs stems that somehow got co-opted to serve as case-markers.

      Evans (1995b) below is probably the easiest introduction to this stuff, and the volume it's in has more examples from other parts of the world, an especially exuberant one from a Sudanic language iirc (unfortunately, probably not a place it would safe to visit). Rounds is of course the most thorough and up to date, but would be a bit rugged for linguistics beginners, I suspect.

      Interestingly, coexisting with this amazing interface between morphology and syntax, there are 'morphological sequencing' restrictions that seem to me like the Ramscar lab take on learning would work quite well for. So managing this stuff would be a challenge, but not necessarily a hopeless one.

      Evans (1995a) _Kayardild: The Language of the Bentick Islanders, with
      Elements of a Grammar of Proto-Tangic, Mouton

      Evans (1995b) 'Multiple Case in {K}ayardild: Anti-Iconicity and the
      Diachrnoic Filter' in Plank (ed) _Double Case_: Double Case. Agreement by
      Suffixaufnahme, OUP

      Round (2009) _Kayardid Morphology Phonology and Morphosyntax_ Yale dissertation_, http://www.ling.yale.edu/~erich/Home.html.

      • melodye says:

        Hahah I love that -- "in steroids and spike heels" 🙂 When I get journal access back, will be all over the Evans articles.. + Avery, I think Michael can probably get you a draft copy of the reading model. Would you be interested? (You can email me!)

  • bet365 says:

    I feel you are too good to write Genius!Thanks for posting, maybe we can see more on this.