E is for Emergence

23 07 2017

path.JPG“Out of the slimy mud of words … there spring[s] the perfect order of speech” (T.S. Eliot).

Eliot’s use of the verb ‘spring’ suggests that language emerges instantly and fully-formed, like a rabbit out of a hat. Historical linguists, sociolinguists and researchers into language acquisition (both first and second) suggest that the processes of language evolution and development are slower – and messier. To capture this messy, evolving quality, many scholars enlist the term emergence.

In what sense (or senses), then, does language emerge? There are at least three dimensions along which language, and specifically grammar, can be said to be emergent: over historical time; in the course of an individual’s lifetime; and in the moment-to-moment interactions in the language classroom.

Languages emerge over time. Pidgins, for example, emerge out of the contact between people with mutually unintelligible mother tongues. Creoles emerge when these pidgins are acquired as a first language by children in pidgin-speaking communities. English itself is the product of creolizing processes, as speakers of different local dialects came into contact with each other and with successive waves of invaders.  There are some that argue that ELF – English as a lingua franca – is yet another instance of an emergent variety.

Because, of course, English continues to evolve. The emergence of the future marker ‘going to’ is a case in point: in Shakespeare’s day, if you were to ‘going to meet someone’ you were literally moving in the direction of the projected meeting place. Over the course of a century or so, ‘going to’ became a metaphorical way of expressing a future intention. By the twentieth century it had further metamorphosed into the contracted form ‘gonna’. Such changes do not happen overnight nor are they ordained by some higher authority or by some genetic disposition. Arguably, everything we call grammar has emerged through similar processes, whereby lexical words become ‘grammaticalized’ to perform certain needed functions, and then, through repeated use, become established in a speech community. According to this view, ‘grammar is seen as … the set of sedimented conventions that have been routinized out of the more frequently occurring ways of saying things’ (Hopper 1998: 163).

Language emerges, too, in the course of an individual’s lifetime, primarily their infancy, as argued by proponents of usage-based theories of language acquisition – those theories that propose that linguistic competence is the product of an individual’s innumerable experiences of language in use.  As Nick Ellis (1998, p. 657) puts it:

Emergentists believe that simple learning mechanisms, operating in and across the human systems for perception, motor-action and cognition as they are exposed to language data as part of a communicatively-rich human social environment by an organism eager to exploit the functionality of language, suffice to drive the emergence of complex language representations.

path 01.JPGThese ‘rule abstraction’ processes have been modelled using connectionist networks, i.e. computerized simulations of the way neural pathways are sensitive to frequency information and are strengthened accordingly, to the point that they display rule-like learning behaviours – even when they have no prior grammatical knowledge (Ellis et al. 2016).

In other words, the system continuously upgrades itself using general  (rather than language-specific) learning faculties, a view that challenges ‘innatist’ theories of language acquisition, as argued by – among others – Steven Pinker in The language instinct (1994).

From a complex systems perspective, the emergent nature of language learning is consistent with the view that, as John Holland (1998, p. 3) puts it: ‘a small number of rules or laws can generate systems of surprising complexity,’ a capacity that is ‘compounded when the elements of the system include some capacity, however elementary, for adaptation or learning’ (p. 5). While humans have this capacity, they are also constrained in terms of how information (in the form of language) can be processed in real time, and these constraints explain why languages share common features (so-called language universals) which, as Christiansen and Chater (2016) argue, are simply tendencies, ‘rather than the rigid categories of [Universal Grammar]’ (p.87).

Finally, language emerges in second language learning situations, especially when learners are engaged in communicative interaction. The learner talks; others respond. It is the scaffolding and recasting, along with the subsequent review, of these learner-initiated episodes that drives acquisition, argue proponents of task-based instruction, with which Dogme ELT is, of course, aligned. ‘In other words, the emphasis shifts from the traditional interventionist, proactive, modelling behaviour of synthetic approaches to a more reactive mode for teachers – students lead, the teacher follows’ (Long, 2015, p. 70). Or, as Michael Breen (1985) so memorably put it: ‘The language I learn in the classroom is a communal product derived through a jointly constructed process.’

A recent book that attempts to unify the different dimensions of emergence – the historical, the biographical and the moment-by-moment – enlists a felicitous metaphor:path 02

 ‘The quasi-regular structure of language arises in rather the same way that a partially regular pattern of tracks comes to be laid down through a forest, through the overlaid traces of endless animals finding the path of local least resistance; and where each language processing episode tends to facilitate future, similar, processing episodes, just as an animal’s choice of a path facilitates the use of that path for animals that follow’ (Christiansen & Chater, 2016, p. 132.)

Is teaching, then, simply a matter of guiding the learners to find the tracks laid down by their predecessors?


Breen, M. (1985). The social context for language learning – a neglected situation? Studies in Second Language Acquisition, 7.

Christiansen, M.H. & Chater, N. (2016) Creating language: integrating evolution, acquisition and processing. Cambridge, Mass: MIT Press.

Ellis, N. (1998) Emergentism, connectionism and language learning. Language Learning, 48/4.

Ellis, N., Römer, U. & O’Donell, M.B. (2016) Usage-based approaches to language acquisition and processing: Cognitive and corpus investigations of construction grammar. Oxford: Wiley.

Holland, J. H. (1998) Emergence: From chaos to order. Oxford: Oxford University Press.

Hopper, P.J. (1998) ‘Emergent language’ in M. Tomasello, (ed.) The New Psychology of Language: Cognitive and Functional Approaches to Language Structure. Mahwah, NJ.: Lawrence Erlbaum.

Long, M. (2014) Second language acquisition and task-based language teaching. Oxford: Wiley-Blackwell.



24 responses

23 07 2017

thanks for opening up the discussion.

usage-based theories, along with all other current theories of SLA seem to agree that we can control the rate of acquisition, but not the route. the research and implications for the classroom present a significant challenge to the predominant coursebook-driven methodology. so why does the grammar syllabus persist? its like the invasive kudzu (japanese arrowroot) in the US south that keeps climbing and spreading, suffocating life around it.

23 07 2017
Scott Thornbury

Donald Freeman used a similar metaphor. The grammar syllabus is like bamboo in the garden – you keep rooting it out and it keeps springing back.
Mike Long likens grammar teaching to feeding seals with sardines. ‘Here, catch that one!’ says the teacher as he/she tosses the students a new grammar item. (Although seals are probably better at catching sardines than learners are at ingesting grammar mcnuggets!)

23 07 2017

“The learner talks; others respond. It is the scaffolding and recasting, along with the subsequent review, of these learner-initiated episodes that drives acquisition, argue proponents of task-based instruction, with which Dogme ELT is, of course, aligned”.

1. If this is your view of language learning, why, in your books with Peter Watkins on the CELTA course do you not mention it, but instead help teachers to use “the traditional interventionist, proactive, modelling behaviour of synthetic approaches”?

2. Many proponents of task-based instruction think that Dogme places too much reliance on “the learner talking and others responding”. They insist that needs analysis, the identification of target tasks and the subsequent design of pedagogical tasks should inform syllabus design. Furthermore, Breen and Long, plus Wilkins, Candlin, Willis, Skehan, Robinson, Schmidt, VanPatten, Hulstijn and a very long list of others who criticise synthetic syllabuses, assume an explanation of language learning that has nothing to do with emergentism.

3. There are 2 strands of emergentist approaches to language acquisition. First, that focusing on the importance of the input (or usage); Nick Ellis takes this approach. Second, that focusing on the role of the processor–working memory interface in language acquisition; William O’Grady is the main proponent. Both approaches have been evaluated by Kevin Gregg. You have consistently failed to give a coherent account of Nick Ellis’ approach, or William O’Grady’s, and you have likewise failed to answer the points made by Kevin Gregg.

23 07 2017
Scott Thornbury

Thanks, Geoff.

1. The CELTA Course was somewhat constrained by the requirements of that course as laid down by Cambridge ESOL at the time. (‘So why write it then?’ I hear you yell). I’d love to (co-)write an alternative pre-service course that wasn’t so closely aligned with an existing provider. The TBLT Course, perhaps. Care to join me?

2, Breen & Long et al. do not subscribe to emergentism, it is true – their best work predates its, erm, emergence. But it seems to me that there is no contradiction between Long’s ‘the students lead; the teacher foillows’ philosophy and an emergentist theory of SLA.

3. For a coherent account of Ellis’s theory, look no further than the man himself: http://www.uttv.ee/naita?id=25911&keel=eng

For O’Grady’s ‘version’ of emergentism, see the chapter by O’Grady, Lee and Kwak in Ritchie & Bhatia (2009) The New Handbook of SLA (Emerald Press) in which they do not totally disassociate themselves from the frequency-oriented, usage-based approach of Ellis and co, but simply foreground the role of processing factors in explaining the route and rate of language emergence. As they put it: ‘a simple processor, committed to reducing the burden on working memory, lies at the heart of the human language faculty. Although such a processor makes no use of grammatical principles, its operation plays a key role in explaining the properties of many core syntactic phenomena—binding, control, agreement, island constraints, scope, and so forth.’

This is very much the line taken by Christiansen and Chater (2016), i.e. that ‘in language acquisition, as in other areas of perceptual motor-learning, people learn by processing , and that past processing leaves traces that can facilitate future processing’ (p. 115). I urge you (and Kevin) to read their book, which offers a more coherent refutation of the latter’s arguments than ever I could.

23 07 2017

1. That’s the best offer I’ve had today! 🙂 I’d be delighted to join you in writing such a course; any profits go to the Hands Up project or the school in Palestine where you help, maybe. But is there no way you could sneak just a bit of criticism of current teaching orthodoxy past the Cambridge ESOL watch dogs?

2. I don’t think it’s fair to say that Long’s best work predates emergentism; you cite his fine 2015 book. Anyway, I agree that there’s no contradiction; my point was that you don’t have to adopt an emergentist view of language learning in order to argue for the kind of change you advocate.

3. Thanks for the great link. I haven’t seen this before and it looks like a really good talk. However, I’ve read a lot (tho nothing like all!) of N. Ellis’ work, and it seems to me that your accounts of emergentism don’t do justice to the evolution of his thinking, to his current concerns with saliency, and to the criticisms of his work by Gregg and others.

4. I’ve read the O’Grady (2009) chapter, and I think your quote is fair, but there’s no denying the very big differences between Ellis, Tomasello and O’Grady, And if I may quote O’Grady (2010, downloadable here: http://ling.hawaii.edu/wp-content/uploads/OGrady_Emergentism.pdf )

“There is currently no comprehensive emergentist theory of language or its
acquisition,….. For the most part, this work is based on the simple thesis that the core properties of language are best understood by reference to the properties of quite general cognitive mechanisms and their interaction with
each other and with experience. The viability of this idea can and must be
measured against its success in confronting the classic empirical challenges of
linguistic analysis—figuring out how language works and how it is acquired”.

I think you’ll agree that they still have a long way to go, and that we shouldn’t present emergentism as a complete unified theory.

5. I haven’t read Christiansen and Chater (2016), but I will, and I’ll pass on your recommendation to Kevin Gregg.

23 07 2017

Would love to see Messrs Thornbury and Jordan collaborate to create an alternative ITT course. It would, if nothing else, lend some credence to the idea that all not flaws with the CELTA lie with how it is taught (an idea which I challenged quite recently, and received a less than lukewarm response).

A danger may be, given the reputations you both have, that the course could be implemented by many as a “this is the way to teach” course, thus suffering the same one-size-fits-all attitude that prevails throughout CELTA.

25 07 2017
Scott Thornbury

That is always a danger, admittedly, Robert. I take heart from a recent article on applying TBLT, which reports some teacher training initiatives – one at pre-service level – which were able to shift teachers’ beliefs about the viability of TBLT and which ‘led in practice to increased emphasis on meaningful, authentic interactive tasks that students found motivating and that develop[ed] students’ fluency in the target language’.

East, M. 2017.’Research into practice: the task-based approach to instructed second language acquisition.’ Language Teaching , 50/3.

26 07 2017

this article by East looks really interesting- thanks for sharing. if only there were more teacher educators working inside the Universities who seem passionate about ‘providing a perspective on where [they] see theory, research and practice and interacting and not interacting successfully.’

24 07 2017

I wonder if it would not be overly cumbersome for the brain to handle such an insurmountable number of neural connections if connectionists rule out rule formation to opt for pattern formation, an argument supported by usage based theories.

25 07 2017
Scott Thornbury

The brain seems to be pretty good at handling a massive number of parallel distributed interconnections: without its pattern abstracting capacity we would be hard pressed to name a tune or to recognize our neighbour on the stairs.

24 07 2017

Michael Swan’s thoughts on Rod Ellis’s strong form TBLT (equally applicable to Mike Long’s) are worth airing:

‘It seems likely, in fact, that TBLT is principally valuable for two kinds of student: those who have already been taught more language than they can use, and those who have substantial out-of-class exposure. For lower-level secondary-school students who have perhaps three hours a week of language study in countries where the target language is not spoken (that is to say, probably the majority of the world’s language learners), the value of the approach is substantially less clear.’

Interestingly, in his TBLT book, Long speaks approvingly of an activity he learned at Swan’s Oxford language school, “overt plagiarism”, sounds fun, think I’ll give it a go.

25 07 2017
Scott Thornbury

Swan’s critique deserves to be taken seriously, and it should be a cause of concern that the most vocal proponents of TBLT (Ellis, Long, Bygate etc) don’t seem to have really engaged with it. Correct me if I’m wrong.

26 07 2017

Ellis’s response seems to be (i) Swan miscategorised Ellis’s particular approach to tblt, which apparently is less prescriptive, or more flexible, than some other TBLT approaches (ii) EFL expanding-circle government school contexts (ie those referred to in Swan’s second sentence above) are particularly likely to benefit from tblt.

Personally, I agree with both positions, the key differentiator in my opinion being the teacher rather than the method.

What I do share very much with Swan is his exasperation with academic researchers making pronouncements from on high about things and situations they have little direct experience of, but now’s not the time to go into that!

Long’s strong form TBLT would be prohibitively expensive, culturally unacceptable and pedagogically unimplementable in my context, and I’m sure in many others. However, we have been embedding tblt into the curriculum for some time and have no plans to stop.

25 07 2017
Sulabha Sidhaye

Dear Scott and other teachers,

When we talk of emergent language , how important is the trainer’s role of “correcting”/ modifying the responses of students while conducting a course ? Especially when certain language variations get repeated in the classroom over the years ? For eg Indians tend to use ‘prepone”instead of “advance’, no matter how often the standard pattern “advance” is suggested; or the case of using enter “ïn’ a room in spite of exposing them to using enter without ïn”.

-Sulabha Sidhaye.

26 07 2017

Hi Sulabha, the OED gives various examples of “prepone” in use, this one caught my eye:

‘‘Hinglish may be catching, but it could be a while before a British man says to his wife in the morning: ‘Darling, can you prepone (bring forward) my meeting with the bank manager or ask my secretary to do the needful?’’ Could be the start of a fun class consciousness-raising activity.

To me as an erstwhile RP Londoner, “prepone” seems like a charming and ‘legitimate’ variation rather than an ‘error’.

Re cases like “enter in”, if you can find a way to ‘fix’ these you will be in line for the Nobel award! It doesn’t impede communication, so I would advise ignoring it, or delaying the fix until “enter in” is the only outstanding ‘problem’, not that I usually take my own advice.

The emerging orthodoxy seems to be that ‘native’ speech patterns are not the end goal, and I see no reason to disagree with that.

27 07 2017
Sulabha Sidhaye

Thanks Patrik ! Your reference to “variations” reassures me that I need not spend precious classroom time on them . That’ll allow me greater focus on important errors ( like correcting * “did not went”) caused by mother-tongue influence !

-Sulabha .

26 07 2017
Justin Willoughby

I take emergent language as grammar, phonology and lexis that comes up during the lesson which was outside of the lesson plan. I record it in a google spreadsheet, which the students have access to. Usually at the end of the lesson I ask the students to review the list and discuss language items that they are not sure about. I also have a few games and activities to review language items weekly or monthly. As we return to speak within similar contexts the students have opportunities to use the language again, however, it’s not always the case and more often than not, students tend to forget and need to be reminded in some way.

In my own experience as a B2 Spanish speaker, I have learned purely from interactions with friends and family and Spanish society. I have had few Spanish classes and to be honest it’s pretty slow going. I guess I am walking blindly down tracks through the forest and stumbling across interesting landmarks and then quickly forgetting them and then stumbling over them again and so on. I definately could have used some type of guide to help me negoiate my way through the forest.

Just to clarify the metaphor, the tracks would be the language and the predecessors would be proficient English speakers, right? So then, to guide in the classroom would be like helping students to remember previously learned items as well as helping them work out the meaning/use, form,etc of new items and offering them many opportunities or affordances to use those items or at least notice them again and again with texts that reflect real usage.

As teachers, should we be able to accurately anticipate language items prior to any given class? How much should be planned for? Should we have a map of the forest and the route laid out or just leave a trail of breadcrumbs?

28 07 2017
Scott Thornbury

Thanks, Justin – in the ‘forest track’ metaphor, I think the point that Christiansen and Chater were making is that languages evolve this way – over time – but I also see ti as the way they may emerge in the course of one’s lifetime – as we revisit not only the language that we are exposed to, but our own language use – repeating, refining, renewing. What then is the teacher’s role? Well, not to pre-define the path, but to motivate the learner to want to follow their own path, and then be there,by their side, as it were, pointing out landmarks along the way – and helping them out of the holes they may fall into!

31 07 2017
Justin Willoughby

Thanks very much for your reply Scott. So then, it’s a good idea to work out what the students are interested in and focus on that. For example, if a student is mad about video games, it would be more motivating to discuss that with them rather than baseball. During the discussion/conversation you could also help the students to notice interesting language by boarding it or, if you recorded it, listening to it again together and when they get stuck you can provide them with the language they need.

28 07 2017

Re Emergentism and language acquisition. My own views on first language acquisition are pretty much in line with Halliday’s ideas about (i) abstracting generalisations from the lexicogrammar (ii) social interaction and “learning how to mean”, ie usage-based and entirely compatible with the work of Nick Ellis and co.

Based on longitudinal study, rather than the limited cross-sectional experiments that often constrain academic research, Halliday’s work on childhood language offers a refreshing (to me) alternative to the reductionist dogma of some other approaches.

Various people, including Nick Ellis and Annamaria Pinter, have called for more longitudinal study in this area, but maybe the politics of academia militates against, certainly the wham-bam 3-6 month MA theses do so.

Re L2 pedagogy, it’s obviously not possible to recreate the conditions of 1st language acquisition in the 2nd language classroom – time-wise or interaction-wise (and most of our students are not babies). I do try to remember this when I develop class activities, but truth to say I often fall back on what I might call “guided-conversation” with small adult classes!

Re adult L2 acquisition and adult L2 pedagogy, surely the fastest and most reliable way for anyone to form their own opinion about these 2 key areas is to attempt to study/learn/acquire a second language and observe and reflect on the processes involved (pedagogic and cognitive).

28 07 2017
Scott Thornbury

Thanks, Patrick. I strongly agree with the last point you made (about studying a SL). My blog about attempting to ‘de-fossilize’ my Spanish (the link is on the right) describes the steps I took and what i learned (about language acquisition if not about Spanish!).

28 07 2017
Scott Thornbury

Just to pick up on your point about Halliday, Patrick. I was interested to see what he might have said about emergence, and while there are no index references (in the collection of papers on the language of early childhood) he does seem to have been remarkably prescient in describing the way that a probabilistic grammar is derived from frequency of occurences in the input, which forms a core principle in the usage-based view developed by N. Ellis and others several years later:

Learning a semiotic system means learning its options together with their relative probabilities, and so building up a quantitative profile of the whole. This concept is familiar in linguistics with regard to word frequencies… Grammatical probabilities are no less part of the system of the language; and they are more powerful than lexical probabilities because of their greater generality. Children construe both kinds from the very rich evidence they have around them. By five years of age, a child is likely to have heard between half a million and one million clauses, so that, as an inherent aspect of learning the principal grammatical systems of the language, he has learned the relative probabilities of each of their terms. An important corollary of this is that children are able to sequence the learning of the grammar, beginning with those options that stand out as being the most frequent.

Halliday (1993) ‘Towards a language-based theory of learning’ in Webster, J. (Ed.) 2003. The Language of Early Childhood: Volume 4 in the Collected Works of M.A.K.Halliday. London: Continuum, p. 342 – 343.

2 08 2017

Just wanted to follow up on some thoughts on Halliday and emergentism in language acquisition.

The idea of frequency patterns in the lexicogrammar is core to Halliday’s linguistics:

“lexicogrammar is probabilistic in nature, and probabilities vary across varieties of English” p73 Introduction To Functional Grammar, Michael Halliday and Christian Mathiessen 2014 4th edition.

And also to his ideas on language acquisition:

“…our discourse as a whole will pattern according to the probability profile of the grammar. This feature of the discourse is one of the principal factors enabling children to learn their mother tongue” P 292 The Essential Halliday , ed. Jonathan Webster 2009.

Christian Mathiessen shares your view Scott that Halliday was ahead of his time, and had no-one in neuro-linguistics “to resonate with” in the 1970s (I’ve forgotten the provenance), but later, for example in On Language in Relation to the Evolution of Human Consciousness, 1995 (reprinted in On Language and Linguistics 2003) Halliday emphasises the alignment between his work and the work of Nobel prize winning biologist Gerald Edelman.

I’m not aware of any specific references by Halliday to Nick Ellis’s work, or Emergentism with a capital E, but Christian Mathiessen, Halliday’s long-term collaborator, has a chapter in Language as a Complex Adaptive System, Nick Ellis and Diane Larsen-Freeman 2009.

All in all, Halliday’s work seems to be aligned with so many of the important developments of the last few decades – corpus linguistics, CLT, discourse analysis, critical discourse analysis, the CEFR, emergentism…, and has inspired a lot of it. And I’ve recently discovered he was the mastermind behind the UK school English language curriculum that framed my adolescence!

Somehow in tracking down these references I came across a reference to your Conversation: From Description to Pedagogy with Diana Slade. Sounds interesting, I thought I knew of all your books but that’s new to me.

2 08 2017
Scott Thornbury

Thanks, Patrick – very helpful, and confirms my view that Halliday is surely the greatest living philosopher of language. One important development that he algns with, that you didn’t mention, is Vygotskian social-cultural theory, and Gordon Wells wrote a whole book about this very alignment – called Dialogic Inquiry: Towards a sociocultural practice and theory of education (Cambridge University Press, 1999) – in which he points out that – among other similarities – both theories share the view that ‘In learning their mother tongue through situationally based conversation, children also appropriate the knowledge and practices of the culture’ (p. 35).

Halliday has also been enlisted into the ecolinguistics ‘camp’, e.g. by van Lier (2004), since systemic functional linguistics is essentially a ‘situated’ linguistics, both constructed by, and constructive of, social and contextual (hence environmental) factors. Van Lier connects Vygotsky, Halliday and emergentism very neatly:

Learning and development are not linear, slowly cumulative and incremental processes that basically just increase in size of volume. Rather, they include series of transformations and emergent configurations of activity and cognition, with phase transitions (the gateways), where the nature of the new processes cannot be reduced to the nature of the preceding ones.

(The ecology and semiotics of language learning: A sociocultural perspective. Boston: Kluwer, p. 104).

Mathiesson, in the paper you mention (in Ellis and Larsen-Freeman 2009), makes a similar point: ‘Learning a language means learning how to mean by learning a complex, adaptive, inherently variable system…’ And he adds: ‘However, although learning a complex system may sound forbidding, learnability has of course, always been a central feature of language. Language has evolved as a learnable system: its adaptiveness and inherent variability make it easier to learn because we do not have to learn it in one fell swoop; we learn it in a cumulative way, building up the complexity gradually from texts instantiating different registers’ (p. 214).

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s