Mike Long: Reply to Carroll’s comments on the Interaction Hypothesis

I’m very grateful to Mike Long for taking the time to wrie a quick response to Carroll’s comments on his Interaction Hypothesis, which I quoted in the latest post on Carroll’s Autonomous Induction Theory. His email to me is repoduced below, with his permission.

Important issues are raised – the roles of noticing versus “detection”, the reach of negative feedback, and, most importantly, perhaps, the statement “I side with Nick Ellis and the UB (not UG) hordes” – which I’ll try to address, once I’ve stopped sobbing, in Part 4.

Hi, Geoff,

Thank you for the valuable work you do with your blog, and just as important, the fact that it is also usually funny (in a good way).

It has been nice to see Susanne Carroll’s work getting an airing of late. I sent a several-page comment on your Part 3 via the Comment form yesterday, butit apparently disappeared into the ether. As I have a day job, and it’s term paper reading week, what follows is a quick and dirty rehash.

Much of the critique the pair of you leveled against the Interaction Hypothesis (IH) focused on one dimenson only: negotiation for meaning and the negative feedback (NF) it produces. But the IH and negotiation for meaning are a whole lot broader than that.

Genuinely communicative L2 interaction provides opportunities for learners focused on meaning to pick up a new language incidentally, as an unintended by-product of doing something else — communicating through the new language — and often also implicitly, i.e., without awareness. Interacting in the L2 while focused on what is said, learners sometimes perceive new forms or form-meaning-function associations consciously — a process referred to as noticing (Schmidt, 1990, 2010). On other occasions, their focus on communication and attention to the task at hand is such that they will perceive new items in the input unconsciously — a process known as detection (Tomlin & Villa, 1994). Detection is especially important, for as Whong, Gil, & Marsden (2014) point out, implicit learning and (barring some sort of consciousness-raising event)the end-product, implicit knowledge, is what is required for real-time listening and speaking. 

While communicating through the L2, a learner’s attention is likely to be drawn to problematic items by added salience resulting from typical characteristics of meaning-focused exchanges. For instance, NS or more proficient NNS interlocutors will consciously or unconsciously highlight important items, e.g., by pausing briefly before and/or after them, adding stress, repeating them, providing synonyms and informal definitions, moving them to more salient initial or final positions in an utterance through left-dislocation or decomposition, and through comprehension checks, confirmation checks and clarification requests, all triggering a momentary switch of the learner’s focal attention from meaning to linguistic form. In addition, NF — mostly implicit, mostly recasts — can have the same effect, while simultaneously providing whatever positive evidence is needed, whether a missing item or a model of more target-like usage. The same incidental learning process operates when learners read a book or a newspaper, listen to a radio interview or watch a movie. However, whereas those activities involve attempting to understand and learn from static spoken or written input intended for NSs and over which they have no control, face-to-face L2 interaction is dynamic, offering opportunities to negotiate for meaning. The negotiation work increases the likelihood that salience will be added, attention drawn to items uniquely problematic for them, and communicative trouble repaired.

You are both skeptical about learners’ ability to compare representations stored in long-term memory with the positive evidence contained in recasts, in order to “notice the gap”. I would be, too. But the comparison involves short-term, or working, memory (WM), not long-term memory. And why would that be inconceivable? Evidence that it is not only possible, but happens all the time (in L1 and L2) is abundant. For instance, what someone says or writes frequently primes, or triggers, use of the same lexis and syntax in that person’s own, or a listener’s,immediately following speech or writing (see, e.g., Doughty, 2001; McDonough 2006; McDonough & Mackey, 2008).

Then, think of the immediate recall sub-tasks common in language aptitude measures, such as LLAMA D and the n-back task in Hi-Lab. Essentially, learners hear/see a short string of sounds or letters and either have to say which ones in a new sequence they heard or read, or repeat them a few seconds later (n-back is a bit more complex). The same basic idea is employed in countless word-recognition post-tests in incidental learning studies. Everyone can do those tasks — some better than others, which is why they are used as a measure of language aptitude – and I reckon they tap roughly the same ability as that used in learning from recasts. The fact that the learner’s original utterance and a recast it triggers are both meaningful, unlike strings of random letters, sounds or words, can be predicted to make it even easier to hold and compare in short-term memory.

Recasts have seven additional qualities (Long, 1996) that make cognitive comparison and learning even more feasible. They convey needed information about the target language (i) in context, (ii) when listeners and speakers share a joint attentional focus, (iii)whenthe learner is vested in the exchange, (iv) and so is probably motivated and (v) attending. (vi) The fact that learners already have prior comprehension of at least part of the message the recast contains, because the reformulation they hear is of what they just tried to say, frees up attentional resources and facilitates form-function mapping. Indirect support for this idea may lie in the findings of a study of the value of content familiarity. Finally, and crucially, (vii) the contingency of recasts on deviant output means that incorrect and correct utterances are juxtaposed, allowing learners briefly to hold and compare the two versions in working memory (WM). Not convinced? Then considerthe fact that statistical meta-analyses (e.g., Goo, 2019; Li, 2010; Loewen& Sato, 2018) have shown that recasts result in measurable learning, with some evidence that they do so better than modelsand non-contingent speech, on salient targets, at least (Long, Inagaki, & Ortega, 1998).

And, again, it’s not just NF. Negotiation for meaning involves higher than usual frequencies of semantically contingent speech, including repetitions, reformulationsand expansions, sometimes functioning simultaneously as recasts, but more generally (something usually dear to UG-ers’ hearts) as (mostly)comprehensible, so processable, positive evidence usable for learning. Problematic forms are recycled, increasing their salience and the likelihood that they will be perceived by the learner. The positive evidence, moreover, is elaborated, not simplified (except for lower mean length of utterance), so retains the items to which learners need to be exposed if acquisition is to occur.

I side with Nick Ellis and the UB (not UG) hordes. Since learning a new language is far too large and too complex a task to be handled explicitly, and althoh it requires more input and time,implicit learning remains the default learning mechanism for adults:

“Even though many of us go to great lengths to engage in explicit language learning, the bulk of language acquisition is implicit learning from usage. Most knowledge is tacit knowledge; most learning is implicit; the vast majority of our cognitive processing is unconscious” (Ellis &Wulff, 2015, p. 8; and see Ellis &Wulff, 2019).

Sure, there are limitations. Williams (2009) notes, for example, that the scope of implicit learning may not extend to phenomena, such as anaphora, that involve non-adjacent items (Hawkins was arrested, and so were several members of his gang), which may no longer be learnable that way. And I have often pointed to evidence that the capacity for implicit learning, especially instance learning, weakens (not disappears) around age 12 (e.g., Long, 215, pp. 2017). But those are for another time, which, you will by now be relieved to hear, I don’t have now.

According to the Tallying Hypothesis (Ellis, 2002), ensuring that learners’ attention is drawn to learning targets that way, especially to abstract, perceptually non-salient items, can modify entrenched automatic L1 processing routines, thereby altering the way subsequent L2 input is processed implicitly. An initial representation is established in long-term memory and functions as a selective cue priming the learner to attend to and perceive additional instances when processingimplicitly. Ellis identifies what he calls “the general principleof explicit learning in SLA: changing the cues that learners focus on intheir language processing changes what their implicit learning processestune” (Ellis 2005, p. 327). Research is currently under way to determine whether it is possible to achieve the same results by unobtrusive,less interventionist means: enhanced incidental learning (Long, 2015, pp. 30-62, 2017, 2020).


Doughty, C. (2001a). Cognitive underpinnings of focus on form. In Robinson, P. (ed.), Cognition and Second Language Instruction (pp. 206-257). Cambridge: Cambridge University Press.

Ellis, N. C. (2005). At the interface: Dynamic interactions of explicit and implicit language knowledgeStudies in Second Language Acquisition 27, 2, 305-352.

Ellis, N. C. (2006) Selective attention and transfer phenomena in L2 acquisition: contingency, cue competition, salience, interference, overshadowing, blocking, and perceptual learning. Applied Linguistics 27, 2,164-194.

Ellis, N. C. &Wulff, S. (2015). Usage-based approaches to second language acquisition. In VanPatten, B., & Williams, J. (Eds., Theories in second language acquisition (pp. 75-93). New York: Routledge.

Ellis, N. C. & Wulff, S. (2019). Cognitive approaches to L2 acquisition. In Schwieter, J. W., &Benati​, A. (Eds.), The Cambridge Handbook of Language Learning (pp. 41-61). Cambridge: Cambridge University Press

Goo, J. M (2019. Interaction in L2 learning. In Schwieter, J. W., & Benati, A. (Eds.), The Cambridge handbook of language learning (pp. 233-257). Cambridge: Cambridge University Press.

Li, S. (2010). The effectiveness of corrective feedback in SLA: A meta-analysis. Language Learning 60, 2, 309-365.

Loewen, S. & Sato, M. (2018). Interaction and instructed second language acquisition. Language Teaching 51, 3, 285-329.

Long, M. H. (2015). Second language acquisition and Task-Based Language Teaching. Oxford: Wiley-Blackwell.

Long, M. H. (1996). The role of the linguistic environment in second language acquisition. In Ritchie, W. C., &Bahtia, T. K. (eds.), Handbook of second language acquisition (pp. 413-68). New York: Academic Press.

Long, M. H. (2015). Second language acquisition and Task-Based Language Teaching. Oxford: Wiley-Blackwell.

Long, M. H. (2017). Instructed second language acquisition (ISLA): Geopolitics, methodological issues, and some major research questions. Instructed Second Language Acquisition 1, 1, 7-44.

Long, M. H. (2020). Optimal input for language learning: genuine, simplified, elaborated, or modified elaborated? Language Teaching 53, 2, 169-182.

Long, M. H., Inagaki, S., & Ortega, L. (1998). The role of implicit negative feedback in SLA: Models and recasts in Japanese and Spanish. Modern Language Journal 82, 3, 357-371.

McDonough, K. (2006). Interaction and syntactic priming: English L2 speakers’ production of dative constructions. Studies in Second Language Acquisition, 28, 179-207.

McDonough, K., & Mackey, A. (2008). Syntactic priming and ESL question development. Studies in Second Language Acquisition 30, 1, 31-47.

Saxton, M. (1997). The contrast theory of negative input. Journal of Child Language 24, 139-161.

Whong, M., Gil, H.-G. and Marsden, E. (2014) Beyond paradigm: the ‘what’ and the‘how’ of classroom research. Second Language Research 30, 4, 551-568.

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 )

Google photo

You are commenting using your Google 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 )

Connecting to %s