Here’s a little test. Read this (authentic) text, and identify the grammar. You have one minute, starting now:
A girl was taking her little brother for a walk in the park. ‘Can I go and run along the top of that wall?’ he asked her.
‘No,’ said the sister.
‘Go on,’ insisted the little boy.
‘Well, OK,’ she said, ‘but if you fall off and break both your legs, don’t come running to me.’[1]
Ask most EFL teachers what the grammar is in that text and they will probably home in on the past continuous (was taking), modal auxiliary verbs in an inverted question-form (Can I…?), the past simple (asked, insisted, said) and some kind of conditional construction: ‘if …’. They might also pick up on the phrasal verbs (go on, fall off), although they might not be sure as to whether these are grammar or vocabulary, strictly speaking.
These are all items that are prominent in any coursebook grammar syllabus.
But if grammar is defined as something like ‘generative multi-morpheme patterns’, and if we understand ‘pattern’ to mean any sequence that recurs with more than chance frequency, a quick Google search, or, more scientifically, a nearly-as-quick corpus search, will throw up many more patterns in this text than your standard grammar syllabus accounts for.
For example:
- take a/the [noun] for a/the [noun] – there are over 100 instances in the British National Corpus (BNC), according to StringNet, of which round 20 are some form of take the dog for a walk
- a walk in the [noun] – 44 occurrences in the BNC
- a [noun] in the [noun] – 10,000 occurrences
- [verb] and [verb], as in go and run – 82,000 occurrences, of which over 5000 start with some form of go
- [preposition] the top of [noun phrase] as in along the top of that wall
- [prep] the top of the [singular N] = 1665 instances in the BNC
- [prep] the [sing N] of the [sing N] = 60,000 occurrences
- [personal pronoun] + [verb] + [personal pronoun], as in he asked her – over 220,000 occurrences, of which 3169 involve the verb ask
- [verb] + [subject], as in said the sister, insisted the little boy – too difficult to count, but very common, especially in fiction
- both + [possessive pronoun] + [plural noun] (as in both your legs): 423 examples
- come/came etc running – 174 examples
- don’t come running to me (a Google search returned a figure of approximately 579,000 results for this complete utterance)
This doesn’t exhaust the frequently occurring patterns by any means, but it’s enough to give you an idea of how intensely and intricately patterned that text is. Moreover, many of the patterns in my list are just as frequent – if not more so – as the relatively narrow range of patterns that form traditional coursebook grammar. There are as many instances of the pattern [preposition] the [noun] of the [noun] (as in along the top of the wall) per million words of running text as there are examples of the past continuous, for example.
The range and heterogeneity of these patterns also challenges the traditional division between grammar and vocabulary, such that some grammarians have opted for the vaguer, but perhaps more accurate, term constructions. As Nick Ellis (2011, p. 656) puts it:
Adult language knowledge consists of a continuum of linguistic constructions of different levels of complexity and abstraction. Constructions can comprise concrete and particular items (as in words and idioms), more abstract classes of items (as in word classes and abstract constructions), or complex combinations of concrete and abstract pieces of language (as mixed constructions). Consequently, no rigid separation is postulated to exist between lexis and grammar.
Note that, according to this view, the pattern go and [verb] is a construction, and so is the idiom don’t come running to me, since both have a semantic and syntactic integrity that has become routinised in the speech community and entrenched in the minds of that community’s speakers. Given the first couple of words of each construction we can make a good guess as to how it will continue.
In this sense, predictive writing tools, like Google Scribe, that draw on a vast data-base to predict the next most likely word in a string, are replicating what speakers do when they speak, and what listeners do when they listen. Rather than mapping individual words on to a pre-specified grammatical ‘architecture’ (as in a Chomskyan, generative grammar view), speakers construct utterances out of these routinised sequences – the operative word being construct. As one linguist put it, “when it comes to sentences, there are no architects, there are only carpenters” (O’Grady, 2005, p. 2).
And it is out of these constructions that a speakers ‘grammar’ is gradually assembled. Nick Ellis again: “The acquisition of grammar is the piecemeal learning of many thousands of constructions and the frequency-biased abstraction of regularities within them” (2003, p. 67).
If this is true, what are the implications for the teaching of a second language, I wonder? Where do learners encounter these ‘many thousands of constructions’? How do they ‘abstract regularities’ out of them?
References:
Ellis, N. 2003. Constructions, Chunking, and Connectionism. In Doughty, C J, & Long, M H (eds) The Handbook of Second Language Acquisition Oxford: Blackwell.
Ellis, N. 2011. The emergence of language as a complex adaptive system. In Simpson, J. (ed.) The Routledge Handbook of Applied Linguistics. London: Routledge.
O’Grady, W. 2005. Syntactic Carpentry: An Emergentist Approach to Syntax. Mahwah, NJ: Lawrence Erlbaum.
Illustrations from Goldschmidt, T. 1923. English by Intuition and Pictures. Leipzig: Hirt & Sohn.
[1] Girling, B. 1990. The Great Puffin Joke Directory. London: Puffin Books.
An A-Z of ELT 
Another stimulating Sunday morning piece has arrived from Scott in which he quotes Nick Ellis: ““The acquisition of grammar is the piecemeal learning of many thousands of constructions and the frequency-biased abstraction of regularities within them” (2003, p. 67).” Perhaps not for the first time on this blog I quote the Dutch linguist, Rens Bod (1998 Preface xi-xiii) “…the productive units of natural language cannot be defined in terms of a minimal set of rules (or constraints or principles) as is usually attempted in linguistic theory….The knowledge of a speaker/hearer cannot be understood as a grammar, but as a statistical ensemble of language experiences that change slightly every time a new utterance is processed.”
The Ellis quote follows on from some detailed examination of language use with raw material from Google and BNC and concludingly comes the question: “What are the implications for the teaching of a second language, I wonder? ”
Surely the implication is that detailed information about the intricate use of the language as enabled by modern linguistic insights and resources like BNC is of use and concern to creators of language learning materials and may well interest teachers wearing their “Applied Linguist” hats, but the teacher in the classroom is going to be primarily concerned with determining the linguistic needs of the learner and facilitating the appearance of as many of those thousands of constructions (Ellis) as possible. I would simply add that I would personally be happier if the teacher were also aiming at the acquisition of the language, not just the grammar.
Thanks Dennis – I never tire of reading that Rens Bod quote, and the notion of ‘a statistical ensemble of language experiences’ fits very comfortably with the notion of a construction grammar.
Morning Scott
A very thoughtful piece. I think that you have made a very powerful case for orienting adult learning around fluency and communication – warts and all. Although not all language teachers are comfortable with the idea of tolerating errors in student utterance….
Aren’t we talking about lexis here, though? I use the word ‘lexis’ here by way of differentiating it from both ‘vocabulary’ and ‘grammar’.
I see lexis as different from vocabulary, in that the ‘v’ word tends to conjure up lists of words, and as we all know lexis includes much more than just words (a dictionary definition gave me ‘the total stock of words and idiomatic combinations of them in a language’).
But more importantly here, the expressions (mostly semi-fixed) that you outline above are different from grammar, I think. Grammar is more systematic. Take the ‘present simple’ for example (neither ‘present’ nor ‘simple’, by the way) – you can apply this to hundreds of different lexical items e.g. I run every day, she takes the dog for a walk in the morning, cheese is nice, etc. But if we look at one of the semi-fixed expressions above, e.g. ‘take a/the (noun) for a/the (noun), we can’t apply this to other words. To say ‘grab these/those (noun) on these/those (noun)’ has a completely different sense, function and use. This is why what you’re referring to above is lexis, rather than grammar, I think.
I think you’re referring to coligation here?
Yep, in the example used, anyway. But not grammar.
Damian asks: ‘Aren’t we talking about lexis here, though?’
To a certain extent, perhaps. Many of the constructions I identified in that passage are – as you suggest – patterns of collocation with variable slots. But isn’t that what ‘grammar’ is, too? If ‘come -ing’ (as in come running) is a collocation, isn’t ‘was + -ing’ (aka as the past progressive) just another one, although admittedly (as you point out) capable of accommodating a wider – but not necessarily an unlimited – range of verbs? Or, turned around, if ‘was + -ing’ is a generative multi-morpheme string, isn’t ‘come + -ing’ another one?
The history of the way certain lexical collocations become metaphorical and thence grammaticised (as in the case of ‘going to’ or ‘used to’) would seem to suggest – again – that the border between grammar and lexis is a fluid one.
And then there’s the case of those constructions (beloved by proponents of construction grammar) that elude any rigid classification, and seem to have meaning irrespective of the words that fill their slots, such as ‘the -er, the -er’ (the more the merrier, the stronger the better, the older. the wiser etc) and the construction that generates sentences like My mother run for Parliament! You must be joking‘ (cf: Me learn Chinese! The Queen ride a bicycle! etc).
Let’s leave the last word to one such proponent, Ronald Langacker:
Langacker,R. W. 2008. Cognitive Grammar as a basis for language instruction’. In Robinson, P., & Ellis, N.(eds.) Handbook of Cognitive Linguistics and Second Languge Acquisition. London: Routledge p. 57)
Thanks for such a well thought out and informed reply. And while that was such an eloquently put last word by Langacker, I still think there’s (yet another) distinction to make here. On one side (and I think this is what you were getting at and I agree with wholeheartedly), we’re considering what’s best for learners of English. And in terms of learners, I agree that the distinction between lexical and grammatical structures has, as Langacker puts it, ‘no clear line’.
On the other hand, I’m also approaching this as a way of analysing language (well, English at any rate) purely for it’s own sake. And this is the pedant in me, I know, but still. In this way, I think it’s worth bringing another language system into the fray, that of supra-segmental phonology. While I’m instantly aware that this is the very thing warned against in the above quote, I also feel that systems in a language are just as integrated as skills (such as the lexical/grammatical continuum referred to).
When I’m teaching listening skills, one of the strategies which I find helps learners who have problems adopting a top-down approach is that of focusing on sentence stress in stress-timed varieties of English. I demonstrate its relevance with an example utterance e.g. ‘By night I’m a drinker, by day I’m a teacher’ (purely for medicinal purposes, you understand). If we look at the stressed syllables (night, drink, day, teach), we can still gain some semblance of meaning. The unstressed syllables, (By, I’m a, er, by, I’m a, er), we don’t even get an inkling of what the person is trying to say. In other words, the stressed syllables carry semantic meaning while the unstressed words provide the ‘fine-tuning’, or ‘grammar’ of the utterance.
With this in mind, I think ‘be + -ing’ doesn’t carry any stress, while ‘come + -ing’ does on the ‘come’, in the same way that in the ‘go for a + noun (drink, meal, run, bike ride, etc.) carries stress on the ‘go’ and ‘noun’. This would indicate to me that there’s still a lexical/grammatical difference between these types of constructions.
However, as I mentioned above, I’m being completely pedantic here by analysing language for the sake of it (I really need to get out more), and I’m not sure that’s going to help learners much. If you’re really saying (as I think you were in the original post, but correct me if I’m wrong) that making such a distinction is of no real use to learners, then I completely agree. I do feel that even today there exists somewhat of an obsession with traditionally defined ‘grammar’, and that blurring the line between ‘lexical’ constructions and the traditionally grammatical ones can only be of benefit in the big picture.
Another interesting post, and again lots to think about.
I think we all need to become a bit more aware of the language – teachers and students alike. I don’t think – and I don’t think you’re necessarily saying – that the grammar in coursebooks is not useful (as long as it’s accurate) but it’s probably not enough.
I also don’t think that we should keep students in the dark about grammar, thinking we should focus on communication rather than grammar. To be able to communicate effectively (and accurately), students need, and usually want, an understanding of how we organise language into patterns to communicate particular meanings.
So, the more aware we become of how language works, identifying patterns ourselves and thinking about why we use particular constructions and not simply relying on what we have learned about grammar from coursebooks, the more able we’ll be to guide students to notice the different ways there are of achieving communicative goals. We shouldn’t only be looking for texts that have good examples of the past perfect etc, but also looking at texts to see what patterns emerge that might be useful to bring to students’ attention.
Blog posts (articles & books) like this help to raise our awareness, as do people like Dave Willis who I’ve seen doing interesting workshops & talks on this kind of thing.
Thanks,
Carol
“…the more aware we become of how language works, identifying patterns ourselves and thinking about why we use particular constructions and not simply relying on what we have learned about grammar from coursebooks, the more able we’ll be to guide students to notice the different ways there are of achieving communicative goals. ”
I couldn’t have put it better, Carol! As a trainer myself, with a special interest in language analysis, I see it as one of my main roles to both broaden and narrow the focus of study – broaden it so as to capture all the other language patterns that conventional, pedagogical grammar ignores, and narrow it so as to take a much closer look at individual itmes by using, for example, corpus tools.
OK, as a sort of first read reaction, I kind of find myself wondering if there really is such a difference between Ellis’s view and Chomskian ideas of grammar. At least I can see how a Chomskian would simply say that Ellis’s ideas might be valid, but that the inborn hardwiring which allows for language acquisition and the underlying grammar or deep grammar is what these constructions are mapped onto and how we as humans can even begin to make sense of the overwhelming statistical data in regards to proper usage. So I’m not so sure the ideas are incompatible, at least they do not seem so to me. But the idea that routinised sequences are crucial to language learning and that they provide a schema which helps make understanding possible seems very true. And the question of how this theoretical understanding can be used to influence our classrooms seems very important. Although I do wonder if conscientious teachers are not already doing much of what is possible to make use of this information. When we do book floods or extensive reading programs, we are in part hoping to bring students into multiple contact with these naturally occurring constructs to help to build up the living grammar that is always more than just lexi or form alone. When we require our students to transcribe their own conversations and those of the people around them, we hope to expose them to the spontaneity of language, the way it can work, like a carpenter, with the limited supplies and time that he/she has on hand. In some ways though, it seems to me that to “assemble” a grammar, there needs to be a recognition of how the pieces fit together. And I think this requires not only multiple exposures to a “construct” of language, but a chance to familiarize yourself with the contours so that when the time comes, it can be slotted into the proper place smoothly. So even in the face of the staggering task of helping students build up a grammar from a huge number of component “constructs,” I think we as teachers cannot get overly anxious and push our students too hard and too often towards new language. Regardless of the terms we use, students will always need time to work on fluency and accuracy with the “constructs” that have already begun to emerge in their own language development.
Thanks for the comment, Kevin. I think that the ‘cognitive linguists’ (of the likes of Ellis) would argue: why hypothesise a (still unlocated) ‘innate’ and ‘universal’ grammar, when ordinary, and observable, learning capacities – such as associative memory, pattern recognition, and chunking – can account for language acquisition perfectly well? As Michael Tomasello (2003) argues, ” There is no need to posit a specific genetic adaptation for grammar because processes of grammaticialization and syntacticization can actually create grammatical structures out of concrete utterances — and grammaticalization and syntacticzation are cultural-historical processes, not biological ones. Thus, it is a historical fact that the specific items and constructions of a given language are not invented all at once, but rather they emerge, evolve, and accumulate modifications over historical time as human beings use them with one another and adapt them to changing communicative circumstances” (Constructing a Language, pp 13-14).
However, I agree with the rest of your post, especially the point that “students will always need time to work on fluency and accuracy with the “constructs” that have already begun to emerge in their own language development.”
Thanks Scott and everyone else who suggested articles. I’ve been busily reading articles on construction and language emergence since I posted this comment and feel kind of sheepish. It really is a whole different way of looking at how language works. Not sure how it will impact what I do in class yet, but just having a new way to understand how language might work makes each class feel like a potential answer to a half-formed question, and that can’t be bad, right?
Another question would be; how do we pass the message on to the teachers of many EFL contexts? In countries like Japan, where the textbook is only 50 pages thick, with 12 units (fragmented by grammar points) and only a short A-B-A conversation introducing a grammar point. Moreover, many teachers do not like to use vocabulary that students ‘have not learned’ yet. It is extremely difficult to give the students an opportunity to encounter these ‘many thousands of constructions’. When will teachers in many EFL contexts stop having lesson in which students just dissect the language?
Good question. The syllabus in most Japanese high English classes is geared towards providing students with the grammar points and lexi they need to pass their university entrance exam. Most teachers feel an inordinate amount of pressure to make sure students can recognize (not necessarily use) the material that is most often included on these tests. I teach in a high school in Japan and unfortunately feel the same pressure (although my school is much more focused on TOEIC scores…another problem in and of itself). Usually I introduce my students to COCA and BNC in conjunction with the General Service List and the Academic Word list. I encourage students to focus on the words that will be most useful for them and teach students how to use vocabulary cards in a way that will help facilitate the functional use of language. Most of my students vocabulary cards include meaning, part of speech and a few collocations as well as one yes/no question and a W question so that they can not only familiarize themselves with how the language is actually used, but use the language to quiz each other when they have time. Still, this requires a student with a high level of motivation and I have had students who were overwhelmed by the whole endeavor. If anyone has other ideas for how to broaden the way students interact with and study vocabulary, especially in Japan, I would love to hear it.
Hi Scott, in answer to you question, I would say that we as ELT practitioners can use these considerations when dealing with emerging language as it comes up in the class.
Looking at the ‘go’ construct, we can look at ‘go go go!’, ‘have a go!’, ‘go on’ as well as ‘get going’, ‘it comes and goes’ as well as the famous ‘we are go for…’ although they may not really need to ‘use’ the last one unless they work for a space agency such as NASA : )
Luckily, we don’t need to have a full functioning ‘declarative’ (as differentiated from our unconscious) map of all the most common combinations in the English language. This is because we should already have a sense, from the LNA, of what the learner(s) need from our learning programme.
So for example, when a learner working in a business context and learning business English starts telling us that they are working on a spreadsheet that will outline how much they need to spend next year, we can feed in the word ‘budget’ and the patterns that go with it like “draft/prepare/revise a budget” as well as “under/over budget” “to budget for something” and “put something in the budget” etc.
The potential danger is that the teacher gets carried away with all these possibilities and the class turns into a lecture about the “life and times of the word budget”.
Thanks for the comment, Chris. I agree that there is a real danger of adopting an approach that becomes fixated on the minutiae at the expense of the bigger picture (e.g. how all these smaller pieces join up), but – as you say – if the lexicogrammatical choices that become the object of focus are those determined by an analysis of the learners’ needs (whether pre-determined or emergent), then there may be less risk of lexical overkill. But it’s a real challenge.
Scott,
I’ve read this post over and over, trying to draw some sort of line between the concept of constructions (which, to some extent, I’ve been just been introduced to – thank you!) and other lexical chunks in general, though I’m not even sure why I set out to do that and whether the distinction is at all relevant.
Could it be – and I’m really guessing here, so please correct me if I’m wrong – that “constructions” is the broader (but hazier) construct (oops), under which we could somehow subsume generative grammar, more routinized lexical chunks and the interface between them?
I got the sense that the constructs you’re referring to in this post are the pieces of language that don’t really “jump off” the page at first glance – the stuff that stands out from the rest of the input the least. So, as a course designer, if I were to highlight and extract a certain amount of language from, say, an article, the “grammar” would be fairly easy to identify (not least because they have already been pre-packaged and labeled in a way) and, to a lesser degree perhaps, so would the lexical chunks, especially the ones that are pragmatically / socially “relevant” – the stuff that would enable students to get something done with the language, as it were.
The constructs you’re referring to are harder to pin down, aren’t they? They strike me as harder to package, harder to label and harder to “pre-notice” for the students – and yet, it’s still raw material for acquisition.
VERY interesting stuff, Scott. Thank you.
Thanks, Luiz, for those ruminations! I’m not sure if my response to Damian’s comment (above) addresses some of your concerns.To return to the literature, here’s another definition:
Taylor, J R 2008 Prototypes in Cognitive Linguistics, in Robinson & Ellis, op. cit, p. 55.
Yes, they do, Scott. Thank you!
Hi Scott I basically agree with everything in this thought-provoking post, however, ‘What are the implications for the teaching of a second language?’ could be supplemented by ‘What are the implications for the learning of a second leanguage?’
I make full use of the Lexical Approach (as others have mentioned or implied there seems to be a lot of similarites) and still a lot of my students are overly-concerned with third conditional. It can take a whole term to get them to start thinking about lexis, chunks, collocations and the like, and then they disappear off to another teacher and I have to start again (I sometimes feel sorry for the teacher who gets them after me).
Not only are the ideas here going to take a lot of teacher education, but also student education. Even after years of innovative research many students still think that ‘traditional’ grammar is the most important thing to learn.
Yes, Stephen, I agree that learners’ perceptions of the goals of second language learning are often at odds with our own or with the researchers’. However, if we can show them that the communicative uptake on acquiring a ‘generative phraseology’ is worth the initial investment in memorisation, and, even, in old-fashioned pattern practice, we may be able to win them over, just as students were once won over by audiolingualism, and its often dogged refusal to ‘label the parts’.
Very thoughtful as usual.
In response to Damian’s comment above, I think depending what perspective you’re coming from you can refer to what Scott talks about as lexis (ELT methodologists, e.g. Lewis), pattern grammar (grammarians, e.g. Willis) or constructions (cognitive linguists, e.g. Ellis). I’ve also come across the terms lexico-grammar and collostructions. Scott, please correct me if I am wrong.
I think the things Scott highlighted do not “jump off the page”, as Luiz put it, because as teachers we’ve been programmed to notice the features that have been traditionally (over)emphasised by pedagogical grammar and received labels such as “Past continuous” and “conditionals”.
I was recently trying to help a student of mine with the use of articles and I drew his attention to the pattern the [NP] of a [NP]: “the opportunity of a lifetime”, “the birth of a child”, “the beginning of a new era” “at the push of a button”, “at the click of a mouse”. Naturally I came up with these examples myself or extracted from the corpus as you can’t find it in traditional ELT materials.
Thanks for the comment, Leo. I hadn’t heard of ‘collostructions’. But I can make sense of the blend. Good point, too, about the labelling of grammar: we tend to prioritise constructions that have names (past continuous, etc) over those that do not. The conditionals are a good case in point, with many combinations that resist labelling as first, second or third, falling through the cracks.
Stefanowitsch, A. and Gries, S. (2003). Collostructions: investigating the interaction of words and constructions. International Journal of Corpus Linguistics 8:2 , 209-243.
Good afternoon Scott.
I read your entry about viewing grammar from the perspective of constructions with interest as I am in the middle of reading S. Wulff and S. TH. Gries (2011) Corpus-driven methods for assessing accuracy in learner development. In P. Robinson (Ed.) Second Language Task Complexity. Amsterdam: John Benjamins. Wulff and Gries’ chapter is about using corpus linguistics to examine Robinson’s Cognition Hypothesis connection to accuracy, as described by Goldberg (1995) construction approach to grammar. Readers interested in Scott’s entry, may find Wulff and Gries chapter of interest.
Cheers from rainy Yokohama.
Colin
“What are the implications for the teaching of a second language, I wonder?”
Currently I’m experimenting with associative-learning mechanisms such as constructions and chunking as the teaching rationale for a more systematic work on pronunciation, which is of extraordinary preoccupation for EFL students whose L1 is more syllable-timed (if the stress-timed/syllable-timed taxonomy is still in order these days). Pronunciation does not seem to be taking care of itself for the majority of L2 learners as some applied linguists suggest.
My main goal is to move students to a comfort zone of mutual intelligibility. I think the first island of reliability to be conquered is articulatory fluency presented as the first step towards lexical and semantic fluency in terms of communicative effectiveness, rather than a matter of native-like sound production ability.
I was a bit puzzled about how this neo-structuralism and its regularities might overkill idiolects, the language instinct, or personal feelings of language ownership, but I felt much better after reading Ellis, N. 2003. Constructions, Chunking, and Connectionism referenced in this post, and finding that Ellis acknowledges that a linguistic system “comprise[s] a conspiracy of constructions” (p.68).
“I was a bit puzzled about how this neo-structuralism and its regularities might overkill idiolects, the language instinct, or personal feelings of language ownership, but I felt much better after reading Ellis, N. 2003.”
Almagro, maybe it’s precisely this sort of exposure to “ready-made” language that will give students the raw material they need to map out the core linguistic elements they need to start playing with the language and creating their own novel utterances.
“The acquisition of grammar is the piecemeal learning of many thousands of constructions and the frequency-biased abstraction of regularities within them” (Ellis, 2003, p. 67).
“If this is true, what are the implications for the teaching of a second language, I wonder?”
One obvious implication, indirectly stated above, is that the form(s) we focus on as teachers become more like recipes and less like what Scott calls ‘grammar McNuggets’, which means we share with learners how key ingredients must be used (eggs in a quiche/verbs in a [verb] and [verb] construction like “go and play”, while noting how other ingredients can be substituted (mushrooms for broccoli, cheddar for swiss, etc./’run’ for ‘play, or ‘live’ and ‘learn’ for both verbs). And, while there’s no accounting for taste, we must consider how a formal dinner is different from a casual lunch; how presentation, the way things are laid out, matters. Hence, polite constructions are usually more ornate (complex patterns).
Where do learners encounter these ‘many thousands of constructions’?”
The general aim of such a language course would be to provide resources by which an ecology of affordances might evolve. Social occasions can serve this purpose, especially in an ESL context (NB: I consider extensive reading a social occasion). As mentioned above, texts of the sort that interest learners and relate to their interests are also essential. Part of the teachers job then becomes judicious recycling of relevant texts and attendant constructions, along with avid exploration of techniques and strategies that enhance learners exposure to, memorization and practice of, these constructions. One way to perform these duties simultaneously, in my opinion, is to scaffold emergent constructions as conversations unfold among students engaged in meaningful dialogue with each other or as they carry out a task.
“How do they [language learners] ‘abstract regularities’ out of them [constructions]?”
That I can’t answer, but we might look for evidence by examining what works and what doesn’t: which techniques seem effective for which learners, and has our approach been helpful in this regard? By keeping journals and holistically assessing our work, teachers and learners might gain insight into how we do this. As a teachers, I would be less interested in mechanical operations of the brain – not that you’ve implied you are thus interested, Scott – and more enthralled with what serves the needs and interests of the people in the room with me.
I hope I haven’t stated the obvious or repeated too much of the comments above, all of which I read with enthusiasm. Thank you for another stimulating post, Scott.
Rob
Rob wrote “One obvious implication,… is that the form(s) we focus on as teachers become more like recipes and less like what Scott calls ‘grammar McNuggets’”
And less like vocabulary McNuggets, too – at least that’s the line that proponents of cognitive linguistics (which encompasses a construction grammar view) might take. That is where the CL position parts company with the lexical approach. Constructions are not simply ‘chunks’. And they should not, therefore, be learned as chunks – i.e. as invariant units, ‘swallowed whole’. Rather they should be unpacked, their ‘innards’ revealed, and then re-assembled and used as templates for creative experimentation. It’s all about deconstructing and reconstructing the constructions.
Thank you for responding to my comment, Scott and for providing the reference, Almagro – yes, that’s the one. To be honest I don’t know why they had to call it “collostructions”. Don’t we have enough terms for this phenomenon already?
As regards conditionals, I agree that the pedagogic treatment does not account for how “would” functions at all with more cases falling through the cracks than the actual cases that can be correctly labelled as 1st, 2nd and 3rd. I am actually planning to write my own blog post entitled “The fourth conditional”
Thinking on Scott’s last question, maybe it is not necessary (or possible) on a conscious cognitive level to abstract regularities out of the constructions. Have you noticed how native speakers throw out expressions which contain words that on closer questioning they don’t appear to know the meaning of? In spite of this, they have usually correctly matched the construction’s use with the intended meaning in the context. Does this point to these ‘chunks’ playing a greater part of first language acquisition than the individual word learning that is commonly used in second language teaching gives credit for? Wherever the balance lies, there are processes at work that we probably will never fully understand and it may be difficult to adequately formulate an approach that marries the subconscious child-like acquisition with the conscious adult desire to learn.
Fascinating post and comments.
“Have you noticed how native speakers throw out expressions which contain words that on closer questioning they don’t appear to know the meaning of?” I’m not sure if what follows relates to what you are getting at, James, but I’ve been reading more about constructions, and one of the features of certain constructions is that they impose a meaning on words within them by virtue of the meaning of the construction itself. Or, put another way, the words ‘inherit’ the sense of the construction. A much-quoted example is the verb with two objects construction, meaning ‘ transfer of possession, and whose prototypical verb is ‘give’, as in ‘She gave him a sweater’. There are lots of other verbs that fit into the verb slot, such as ‘loaned’, ‘bequeathed’, ‘sent’, ‘posted’, ‘fedexed’ etc. What’s interesting is that when verbs that don’t normally mean ‘transfer of possession’ are slotted into this structure, they inherit the meaning of transfer of possession, as in ‘She knitted him a sweater’ or ‘She baked him a cake’. (Apologies for the rather sexist examples).
It’s this capacity of ‘unfilled’ constructions to encode meaning, irrespective of the words that we fill them with, that allows us to invent phrasal verbs like ‘I’m conferenced out’ or to understand sentences like ‘They blogged their way to the summit’, and to make sense of how countable nouns can take on uncountable meaning, as in: “In an exciting second half The Magpies took the game to Clonlara. Ronan Garvey and Adam Cassidy supplied a lot of ball to their forwards, …”
So, coming back to your point, James, I read this to mean that the individual words in many constructions are less important than the construction itself, so it doesn’t really matter what they are. This is often the case with idioms or quotations that retain some long outdated usage, as in ‘hoist with his own petard’ or ‘gone to rack and ruin’ or ‘as is my wont’ or ‘I fain would lie down’ and so on.
Yes Scott, that was what I was searching for, and you have given much more as well, thanks.
It raises some psycholinguistic questions for me concerning the balance of variable and systematic variation (see Ellis (1999) -the balance between item and system learning for first language learners and whether this is the same for second language learners or available to them in the same way; and if the self-organizing process in first language development is the same in SLA.
Do you know of any literature that might help?
Hi James, I’m not sure I can answer this one, except to say that Ann Peters (1983, The Units of Language Acquisition) was one of the first to distinguish between ‘holistic learners’ and ‘analytic learners’ in first language acquisition, a distinction that seems to carry over into SLA, although there is less evidence that in SLA those who start by learning ‘holophrases’ are capable of abstracting their internal structure in ways that lead to subsequent creativity. Alison Wray (2002, Formulaic Language and the Lexicon) quotes a number of scholars “who believe that ‘there does not seem to be a direct line from prefabs to creative language” (Granger 1998: 157), so that teaching formulaic sequences for this purpose will have no point”. She adds that “in contrast, Nattinger and DeCarrico (1992) work very much on the basis that adult learners do use formulaic sequences as input for their analysis of the language out of which they will derive grammatical and morphological rules” (p. 193).
The last word she gives to Dave Willis, and this is worth quoting in full, since it has a bearing on the whole discussion about constructions:
Wray A. 2002. Formulaic Language and the Lexicon. Cambridge: Cambridge University Press.
Sparks a book idea, Scott: Invisible Grammar (subtitle here).
Rob
In relation to the words ‘inheriting’ the sense of the construction, another quoted (see first reference or Google search ‘”conduit metaphor” in education’) example is the verb ‘teach’ which, when incorporated into the two objects construction, again conveys the ‘transfer of possession’ idea, unwittingly signing users of the construction up to a transmission metaphor/theory of education and, perhaps in a Sapir-Whorfian manner, encouraging users of the construction to construe a particular view of language teaching as a result.
On the labelling of parts and the tendency to prioritise constructions that have names, my surmise is that this feeds into the adult learner’s conscious learning strategy (and and the teacher’s teaching strategy) to ‘Name the problem so we can fix it’; however, the problem then arises of “if all you have is a hammer, everything looks like a nail”, in other words, as was expressed above if all you have is named structures, then what isn’t named “falls through the cracks”.
[The next bit is looooong – so skip to end for a summary if you like].
One way to deal with the issue of the countless nameless patterns/constructions/colligations/collocations is to this may be to develop what I would call the twin counterpart to the learner’s internal constructor of constructions (!): the parser. This is “the mental program that analyzes sentence structure during language comprehension.” (Pinker, The Language Instinct, p. 198).
Parsing is the ability, studied by psycholinguists, that language users have to, among other things, sort out sentences like ‘The horse raced past the barn fell.’ This parser does not depend on grammar alone to figure out meaning but uses a predictive approach similar to Google Instant, (a feature that displays suggested results while the user types); Pinker spells out the evidence for this in Chapter 7 thus:
‘How do people home in on the sensible analysis of a sentence, without tarrying over all the grammatically legitimate but bizarre alternatives? There are two possibilities.
1. One is that our brains are like computer parsers, computing dozens of doomed tree fragments in the background, and the unlikely ones are somehow filtered out before they reach consciousness (“breadth-first search”).
2. The other is that the human parser somehow gambles at each step about the alternative most likely to be true and then plows ahead with that single interpretation as far as possible – the “depth-first search.” ‘ (i.e. Google Scribe approach).
Experiments show that people do not take the “breadth-first search” for two reasons. ‘One is that many sensible
ambiguities are simply never recognized” (Professor gives talk on moon). The second reason is “(people) stubbornly fail to find the only tree that is consistent with a sentence ….e.g. The horse raced past the barn fell.’
Pinker (ibid. p 214-216) goes on to say that: “A depth-first parser (like ours) must use some criterion to pick one tree (or a small number) and run with it—ideally the tree most likely to be correct.” Experiments show that the criteria used are”:
(a) some real-world knowledge
(b) knowledge of constructions (“The most common entry for a verb seems to pressure the mental parser to find the role-players it wants”.
(c) priming (“Words can also help by suggesting to the parser exactly which other words they tend to appear with inside a given kind of phrase. Though word-by-word transition probabilities are not enough to understand a sentence, they could be helpful; a parser armed with good statistics, when deciding between two possible trees allowed by a grammar, can opt for the tree that was most likely to have been spoken. The human parser seems to be somewhat sensitive to word pair probabilities.”) and
(d) by favoring (syntactic) trees with certain shapes, a kind of mental topiary. (some Gricean-analogous strategies are mentioned such as ‘thrift’ and ‘momentum’) [Judges have similar tactics when interpreting ambiguous laws interestingly called “canons of construction”].
(e) The parser also uses grammatical islands of certainty: “Languages have “bounding” restrictions that turn some phrases, like the complex noun phrase the rumor that Mary likes him, into “islands” from which no words can escape”. These islands are similar to the way chunks such as multi-word units can be exploited as islands of certainty by learners.
Parsing of course is not the whole story in processing a sentence. In written text, like transcriptions of conversations, Pinker notes “the intonation and timing that delineate phrases is lost”; and prosody is how people sort out a lot of ambiguity in real-life (such as in defining vs. non-defining relative clauses).
All of the above are still not sufficient to sort out meanings of fragmentary speech – you need the Gricean maxims of mutual expectation, knowledge of discourse patterns like problem-solution etc.
TL:DR – the ability to use constructions is the flipside of the ability to parse. Parsing is not just “knowing grammar” but also hinges on things like real-world knowledge, a feel for constructions in terms of SVO etc, lots of priming and activation of this priming, psycholinguistic shortcuts to trim unwieldy Chomskyan trees down to memory-managable size, prosody, Gricean maxims, shared knowledge and ways discourse c reates text coherence.
Getting to the point, what are the implications for teaching? What techniques develop this internal constructor-parser?
– processing instruction ?
– problematizing ?
– text reconstruction ?
– vast exposure / extensive reading-listening ?
– collocational competence training ?
References:
Richard G Tiberius, Metaphors Underlying the Improvement of Teaching and Learning, British Journal of Educational Technology Volume 17, Issue 2, pages 144–156, May 1986 (abstract: http://onlinelibrary.wiley.com/doi/10.1111/j.1467-8535.1986.tb00504.x/abstract)
Click to access reddy_1979_conduit_metaphor.pdf
http://en.wikipedia.org/wiki/Law_of_the_instrument
http://en.wikipedia.org/wiki/Garden_path_sentence
Steven Pinker, The Language Instinct (1994) [Chapter 7], ISBN 978-0060976514
http://www.qwantz.com/index.php?comic=204
Exploiting multi-word units in statistical parsing and generation
Cafferkey, Conor (2008) Exploiting multi-word units in statistical parsing and generation. Master of Science thesis, Dublin City University. at http://doras.dcu.ie/615/
On priming (although not parsing) I just happened on this piece by Chris Tribble, who writes regularly on corpus issues for the TEFL section of the Guardian Weekly: http://www.guardian.co.uk/education/mind-your-language/2011/jul/05/first-words-weekly-chris-tribble
He looks at the most frequent word combinations that begin Guardian articles. For example, “with It is a we find: measure of, matter of, shame that, tribute to and kind of; and This is a combines most frequently with: matter of, book about, man who, very important and moment of….”
Useful information for students embarking on a discursive essay?
On the ‘garden path’ phenomenon, i.e. sentences that lead to a dead end, of the type ‘The horse raced past the barn fell’, O’Grady (in the book Syntactic Carpentry that I have already referenced, and in which he argues that utterances are constructed in a linear fashion, based on an efficiency principle (‘resolve dependencies as soon as possible’)) argues that we tend to process these ‘garden path’ sentences using the same efficiency, or ‘least effort’, principle.”In each case of a garden path effect, the processor encounters a dead-end because of its haste to incorporate new material into the sentence by associating it with an already available element… The sort of missteps associated with garden paths is just what one would expect of an efficiency driven processor that has to deal with complex material under severe time constraints” (p. 165). Such a view does not presuppose ‘trees’ as such, but rather a computational system that “seeks to complete its work as quickly as possible” (p.7).
Thanks for the comments on my overlong (unedited) spiel. I add the disclaimer that I am certainly no Chomskyan syntax-tree-hugger but Pinker (his MIT mate) writes very clearly and was my best source for the state-of-the-art for parsing.
The constructor-parser theory I propounded was at best an ill-informed guess at the inner processing of language; but, I’m heartened to see that O’Grady agrees in one respect: “I take the computational system to be identical in the relevant respects for both production and comprehension.” (p. 7) though he goes on to say: “This is not to say that production and comprehension proceed in exactly the same way—clearly they do not”. O’Grady later reiterates this view: “The key claim of this book is simply that the brain processes sentences in certain ways (linearly, efficiently, and so forth), not that there has to be a single identifiable processing mechanism in a particular place in the brain”. (p.210) Still, I think he might be being coy here.
He also says that “The suggestion that those operations apply in a linear manner is less common of course, but certainly not unprecedented,… “, and it brought to mind David Brazil’s “grammar of speech” which I could never quite get my head around. Any comparison to be made ?
http://books.google.pl/books/about/A_grammar_of_speech.html?id=ADDF6jzEZMAC&redir_esc=y
O’Grady talks a lot about the language user detecting “arguments” as part of the computation (metaphor alert !) and I’m wondering if there would be any mileage to be had with using garden-path sentences with students of a certain level to test the processor’s efficiency – I remember Mario Rinvolucri had an exercise called “Hinged Sentences” in which things like “Where are my scissors are there on the table” [More Grammar Games p 52] were puzzled out by students – a type he called cognitive games, from the Silent Way, and suitable for academically-minded students (he says in the intro).
Apart from parsing such freakish examples, another possible avenue might be to help adult L2 learners get their processor chopping up the sentence (or growing it at the right node), by having them recognise arguments more readily, and to this end perhaps more attention could be paid to the structure of Noun Phrases (many of which occur in the joke you quote) and Adverbial phrases, as well as common constructions around verbs [as seen in Collins Cobuild Grammar Patterns: Verbs (1996)] – the latter a much-underrated tome.
Thanks, Declan…. re Brazil, I have his name scratched in the margin against several passages in the O’Grady book. They seem to be saying very similar things, although coming from totally different linguistic paradigms (generative grammar in O’Grady’s case, and functional grammar in the case of Brazil). Take this quote, from Brazil (1995) for example:
And later:
I think what’s missing from Brazil’s account (and also from O Grady’s) is recognition of the distinction that Sinclair makes between the open-choice principle and the idiom principle. That is to say, speakers do not “necessarily say one word and follow it with another”, rather they retrieve several words at once, in the form of a chunk, more often than not.
I like the idea of “parsing freakish examples”, and on teacher training courses (but not with real students) I’ve sometimes challenged students to say why these apparent constructions are not in fact constructions, although they do occur in real sentences:
e.g. ‘What I like to do is go shopping at midnight”; “the person I was talking to was to sing at my wedding”, “A car I once had had had ten different owners” etc. Not sure what good it does, but it’s a great game!
Dan Jurafsky has an interesting parallel to make between comprehension by the parser and production. I quote in full: “… all three of these roles (exist) for probability in comprehension: access, disambiguation, and processing. The claim that human language processing is probabilistic also has implications for production. Probability may play a role in accessing structures from the mental lexicon or grammar. High-probability structures may be accessed faster, or more easily, or simply with more confidence. Disambiguation, itself a phenomenon of comprehension, has a correlate in production: choice.” (p.40)
Dan Jurafsky, Chapter 3, Probabilistic Modeling in Psycholinguistics: Linguistic Comprehension and ProductionProbabilistic linguistics, editors: Rens Bod, Jennifer Hay, Stefanie Jannedy. 2003 (Bradford Books)
“Probability may play a role in accessing structures from the mental lexicon or grammar. High-probability structures may be accessed faster, or more easily, or simply with more confidence.”
This connects interestingly with Michael Lewis’ assertion that “Grammar becomes Lexis as the event becomes more probable”, doesn’t it?
For teachers I think the notion of constructions provides very real food for thought about how we ‘do grammar’. The assumption in CG, as in other emergentist-friendly views (Hopper; Bybee; Larsen-Freeman; N. Ellis) is that grammar is an effect, not a cause – the result of those “frequency-biased abstraction of regularities” Scott mentions above. In this view there are no ‘rules’ in language at all, only ‘representations’ – and these representations co-occur with the language that instantiates them. For L2 learners, this means that everything a learner needs to know about the ‘grammar’ is in the instance of language-in-use (taken to include rich context, such as speaker relations, gesture, cultural norms, etc).
One (perhaps controversial) implication for the classroom, I think, is that we need to reconsider active memorisation as a learning tool. One construction grammar inspired process might be to include practice of the kind:
1) learners memorise fully substantive, likely-that-they’ll-need constructions (in context, of course, via meaningful exchange, as needed/wanted by learners, etc). So, perhaps beginners learn something functional for the classroom, like ‘could I borrow a pen’. This is activated again and again (meaningfully – this is not old-school behaviourism…), so that it begins to achieve ‘unit status’ (i.e. is stored and retrieved as a single item, and is not ‘built’). This creates a low-level (contentful) ‘exemplar’ for the learner, from which they will begin to explore variation.
2) Learners become (or are made) aware of a generative pattern within these now memorised ‘units’, such that they see applications to new contexts. In the example here. then, we encourage learners to replace ‘pen’ with something else (‘phone’ or whatever) – but we keep the variation to a minimum. Thus internal structure begins to abstract, through the construction [could I borrow a N], but as immanent in full examples, not as a ‘semi-fixed expression’ with a ‘free slot’ as the Lexical Approach would teach it. This is a mundane example and may seem very slow, but I think practice of this kind with adults can probably quickly build a ‘construction competence’ that allows the process to become a self-generated strategy outside the classroom. A bottom-up, usage-based view (such as this), would then predict that frequency of input-and-use will lead to differing constructions with differing levels of content and abstraction – even in native speakers.
This approach would also suggest that perhaps we should not teach (e.g.) the whole of a verb paradigm in one class (as no doubt still happens in language classrooms everywhere). We build it up slowly, on a need-by-need basis, via fully contentful phrases that learners want and need. It’s an ‘intelligent phrasebook approach’, I think. As learners continue to meet rich input (through video / text / etc) they begin to see instances of constructions they know, but used with slight variations – and their sense of structure evolves. In this way we create an approach to learning that may mirror somewhat the way grammar may actually emerge (notwithstanding the likely co-occurence in the learner’s mind of metalinguistic knowledege of ‘rules and exceptions’, etc, that learners will inevitably gain from self-study and traditional teaching / learning). We also build early islands of fluency this way (in a similar way to that suggested by Lewis in the Lexical Approach), since our learners are retrieving from memory more than they are building from scratch (just like native speakers) – and this is of course faster for both comprehension and production.
There is empirical evidence that this works for children. Kids of around 5-6 were able to correctly ‘learn the rule’ of sentences containing both novel non-English word order (NP NP VP for a transitive sentence) and one of 5 nonsense verbs, while watching a video of puppets performing actions (Casenhiser & Goldberg 2005). Of real interest from a constructions (and memorisation for L2) point of view, though, is that they abstracted the new pattern for the purposes of basic creative use faster when presented with one new verb/video 4 times and the other 4 verbs once, than when each was presented twice. The repetition allowed the children to remember the first token as an ‘exemplar’, and then to abstract the type (the new grammar) through comparison with analogous structures containing slightly differing content. I have a hunch (if there are not already L2 studies) that the same would be true for adult second language learning. We need to engage learners’ natural analogical abilities, by first giving them something to draw on for comparison. Memorised constructions are, I believe, one way. We introduce too much type variation in how grammar teaching is usually done. I think we need more teaching of tokens.
Finally (since it’s almost Sunday again and I’m going on too long again), a CG view highlights, as Scott mentions in a comment above, that it is actually the meaning of the whole that guides the parts and not vice versa. In the utterance ‘she sneezed the napkin off the table’, sneezed is being used as a transitive verb – but look it up in a dictionary and you’ll be told it’s intransitive. It is the meaning associated with this caused-motion construction itself that licenses the transitive use. Since the S V O PP structure exists as the common patterning across thousands of instances of use, but not as separate from these, it retains meaning. ‘Pure grammar’ (in traditional terms) carries meaning. For the classroom, I think we might use this insight to sometimes start with peripheral and creative instances of language, in order to teach the construction – and its creative limits. ‘Collostructionally’, we might come across the caused-motion construction with verbs like ‘pushed’ more often (‘we had to push the car up the hill’), but teaching an example ‘on the edge’ can bring, I think, far deeper insight into the why of grammar in ways that may allow our learners to take faster ownership of constructions and make them their own.
Thanks, Steve, for that lucid account of how a ‘construction grammar’ pedagogy might work in practice. It does seem that an exemplar-based approach offers a way forward, but that we can’t trust the learner to “abstract the patterns” unaided. That is to say, whereas in first language acquisition exemplars ‘release’ their grammar implicitly, in second language acquisition they may not — but simply remain as memorised chunks in a largely lexicalised competence. In his excellent book, Cognitive Linguistics and Language Teaching (Palgrave Macmillan 2009) Randal Holme dismisses the Lexical Approach for this very reason: “Students doubtless benefited from an enhanced repertoire of lexical phrases, but over-encouraged to seek the safety of the fixed forms, they still lost control of the larger meaning and found themselves stranded on phrasal islands of incongruous correctness by the error-infested nature of the larger sentence” (p.6). This is what I have referred to elsewhere as “all chunks and no pineapple”!
Steve,
I first heard of the approach you’re advocating (from whole to parts, from synthesis to analysis, from ready made to generative) in the early 90s, when I read a gem of a book called Lexical Phrases, by Nattinger and DeCarrio. In fact, I think the Lexical Approach owes a lot – conceptually, at least – to the perhaps underrated and under hyped work of Nattinger and DeCarrio.
So what you’re saying makes a lot of sense to me on a lot of levels and, yes, I do think it’s time we revisited the role of memory in language learning in its own terms, without the inevitable (and sometimes ill-conceived) references to Pavlov and Skinner.
But let me tell you about my recent experience in Paris. I know very little French (A1+, tops) and was lucky enough to have traveled with a friend who’s practically bilingual. So, as it turns out, she’d give me (and sometimes drill!) lots of ready-made chunks (wonder if they’d qualify as constructions, though – guess I still have some reading to do…) to use in a number of everyday situations. Since those chunks enabled me to get the work done, as it were, the natural thing to do would’ve been to stop there – at phrase book mode – and get on with the rest of the trip, right?
Well, no, not quite – not in my case, at least. As soon as I’d used the ready-made phrases / sentences, I’d then proceed to break the chunks down into their constituent parts and create new sentences in my head – lots of them. I really wanted to explore the generative potential of all the new chunks and I found myself either murmuring the new utterances or saying them out loud over and over.
So I think the point I’m trying to make – very tentatively, since it’s based on nothing but a personal anecdote – is that the synthesis / analysis, whole / parts transition might happen more or less naturally and organically depending on the learner. I, for one, am NOT the kind of learner who allows himself to simply be immersed in the new input and let the new grammar blossom on the basis of the formulaic repertoire built over time. But each to his own, right?
First of all, I can’t resist saying that I found Steve’s posting outstanding – stimulating, , well worth reading several times. Thank you. The suggestion that it might make sense not to teach all the forms of a verb at the same time but at the point of communicative need (or just before it) is surely what many learners have always done learning a highly inflected language like Russian, for example. I suspect you can get on pretty well for a long time without the vocative.. I suppose this learning of all forms together is yet another continuation of good old/ bad old traditional Latin learning pedagogy:: Nauta, nauta, nautam —–, amo, amas, amat etc. As for Luiz being drilled in Paris, and, it seems getting something from it – along with a mention of memorizing and Skinner, certainly not for the first time I’d like to suggest that there is a helpful comparison – as long as it is is not taken too far – with learning to play a musical instrument and learning to speak in a foreign language – both being performances. And the two require systematic practice and a great deal of repetition.