Timothy Mason's Site

Creative Commons License
This work is licensed under a Creative Commons License.

This lecture, like the others in this series, was given to students of English at the Université of Versailles St. Quentin, for a course in the Didactics of English, which I taught from 1993 to 2002.

It offers a simplified introductory account. The embedded links, most of which point to material off this site, are for readers who are looking for greater depth and complexity.

Didactics - 5 : Critique of Krashen I

The Acquisition/Learning Hypothesis

A : Recap

We have seen that

  • according to Krashen, learning an L2 is very much like learning an L1. No conscious effort needs to be made to focus on the language as such, but if people pay attention to the sense of what they are hearing or reading, the ability to speak and to write will come more or less of their own accord .

  • We looked at an interesting experiment - Canadian primary school - French through immersion - producing children who had a native-level comprehension both in listening and reading, who had an excellent accent and who were favourable to the community whose language they had studied.
    But - it did not seem to give them the firm grasp of grammatical structures that were necessary for the production of well-formed utterances and of texts at native-speaker level.

  • This seems to stem from the fact that :

    • - the children were not expected to speak in full sentences
    • - the teachers and other pupils could always interpret what they meant by reference to the context.
    • - they did not have the formal French lessons that would have drawn their attention to the need for correct syntactic form.

So - this suggests that Krashen is wrong in at least two of his hypotheses.

  • a) The Input hypothesis - productive behaviour on the part of the learner is not necessary to language acquisition. If we mean by acquisition the ability to produce correct utterances, both orally and in writing, we suspect that he is wrong.

  • b) Also seems to be wrong when he suggests that learned language - the rules of grammar - are only of much use when writing - people do seem to need the rules in order to speak in well-formed sentences. This may lead us to believe that the first hypothesis itself - that of the distinction between learning and acquisition - is an oversimplification.

Nevertheless, if we look at Krashen's hypotheses, they may provide us with a framework within which a number of crucial questions can be asked.

B : The Acquisition/Learning hypothesis

We have seen that this has been criticised as an oversimplification. We can ask the question :

  • 'How is language stored in the brain, and in particular, how is a second language stored in the brain?'

Krashen does not really tell us :

  • a) what the process of acquisition is. (How does the learner build up a usable model of second language behaviour from input alone?)

  • b) why learned information is not accessible in the same way as acquired information

It is feasible that the distinction between learned and acquired knowledge is in fact more properly a distinction between two stages of the learning process. We are ready to agree with Krashen that the two are distinct kinds of knowledge because we have usually forgotten how we ourselves shifted knowledge from the conscious to the unconscious domain.

A rival hypothesis about learning comes from the work of Anderson, and other cognitive scientists. According to this perspective, learning is a process of assimilation whereby new information is processed by the brain in such a way as to be incorporated in already existing knowledge.

We may conceive of the brain as having two different types of memory :

  • 1. Short term memory - or working memory  - a finite storage space We use it to carry information while we make use of it. The information may come from external input - a telephone number, that we memorise long enough to make the phone call - or from -
  • 2. long-term memory - items are stocked in associative networks - thus the word 'home' will carry with it a whole series of associations - depending on our experience, our culture and so on.

For learning to occur, the information that is in short term memory must be incorporated in the network system. This may occur by simply adding it to a set of already existing associations :

  • - the child learns that Eskimos live in igloos, and adds that to the conception that he has of home.

- or it may need a major reassignment of links between different concepts

  • - when we learn that nearly 50% of murders take place within the home or are committed by members of the family, our concept of home changes radically, is linked to the possibility of violence and conflict, rather than simply the old log fire and mum cooking something nice in the kitchen.

Anderson distinguishes also between two different kinds of knowledge :

1. Declarative knowledge
This is conscious, and is stored as a series of statements or images. It is knowledge about the world - who is the President of France, for example, or the fact that 'move' is a verb, whereas 'movement' is a noun.
2. Procedural knowledge
This is unconscious, and consists of routines or procedures that allow us to bring declarative knowledge into use. Thus, our procedural knowledge of our mother tongue permits us to construct grammatically correct sentences without consciously thinking about it.

For Anderson, learning takes place in three stages :

  1. the Cognitive Stage - learner receives instruction, or watches an expert, or studies the question on his own - conscious effort, and - knowledge is formulated in a series of rules - conscious factual statements which Anderson refers to as 'declarative knowledge'.

    For L2 learning, this would include the rules of grammar and lists of vocabulary, as well as chunks of useful language that can help us out in specific situations.

    At this stage, the learner can describe the rules, but does not use them skilfully - performance is hesitant, and there are many errors.

    For Anderson, the model of L2 learning is the classroom, and the Cognitive Stage corresponds to being instructed by a teacher - the expert. You may also think of learning to drive, with an instructor. You learn how to use the clutch, how to learn the gear-lever, how to use the brakes and so on. The knowledge is there, but the learner may not be able to use it. Thus, as Rod Ellis suggests, the learner may know that the word 'drowned' consists of 'drown' + 'ed', but not be able to construct the word in conversation.

  2. The Associative Stage - two things occur :
    • a) errors in the declarative statements are detected and eliminated
    • b) connections between the different elements of the skill are strengthened

    This implies that, from being declarative, the knowledge becomes procedural. Thus the learner's knowledge abut 'drown' and 'ed' will be linked to his knowledge of 'save' and 'ed' and will be subsumed under a single production rule for producing the past tense - and indeed, the rule may be overgeneralized and used to produce 'goed', for example.

    The initial declarative representation is not always lost - we still remember the rules - but we no longer need to apply them consciously. This may be why Krashen can point to the existence of two separate systems - the old system, taken on board at the first stage of learning, remains, but is not directly used in language production.

    Performance begins to resemble expert performance, but may be slower. Instead of thinking of the clutch, the gear-lever and the brakes separately, we now mould them together into a non-verbalised procedure.

  3. The Autonomous Stage :

    - skill becomes virtually automatic and errors disappear. It can now be executed without attention - driving a car and having a conversation at the same time. With a complex skill, this stage takes a long time to reach.

    At this point, the conscious declarative knowledge may be lost - if you ask a competent driver to tell you how he drives, he may be unable to tell you - which is why we have specially qualified driving instructors, and also why the fact that you are a speaker of French does not qualify you as a teacher of French as a foreign language!

    The acquisition of the skill - like any other skill - depends on learning and practice - in this, Anderson differs radically from Krashen. For Anderson - the rules which we learn when acquiring a second language = rules of grammar as taught in class.

However, other observers have pointed out that this is hardly likely to be the case.

  • 1. First of all, many people learn foreign languages without going to classes and without begin presented with the formal rules.
  • 2. Secondly, the rules that are presented in class are by no means all the rules that are needed to speak a foreign language well.
  • 3. Finally, not all foreign language classes actually present the rules of grammar, and yet members manage to acquire the language.

So what does happen? It seems that individuals generate their own rules. Thus what is needed, it may be claimed, is not a class in which rules are dictated and learnt by heart, but what Gagné refers to as 'cued performance' and opportunities for practice.

This involves :

  • - modelling by an expert - the expert shows how it is done - the language teacher shows how to produce correct and/or comprehensible statements.
  • - active attempts by the learner to produce the activity himself - with the instructor cueing him at moments when he forgets the rules - think of the father running alongside the child who is riding a bicycle for the first time, and shouting instructions

The learner tries out the activity before possessing it completely. She is thus able to learn from her successes and her mistakes - learning by trial and error.

This is formalised as the construction and testing of hypotheses. Based on input, or on her understanding of the rule pronounced by a teacher, or whatever, the student makes a prediction about the effect of using a rule that she has generated. She can then test this through :

  1. Reception - compares the hypothesis to further input.
  2. Production - uses the hypothesis to generate language, and then assesses the result.
  3. Metalingually - consults a native speaker, or a grammar book.
  4. Interactionally - makes an intentional error to elicit a repair from a native speaker.

It will be seen that this view is in opposition to Krashen's input hypothesis. It is not sufficient for the learner to simply accumulate input - he or she must actively engage with the activity.

There are also problems with this account - in particular, the claim that is made in this tradition of cognitive science is that the learner proceeds through trying out hypotheses and receiving feedback. However, as we have seen - this is certainly not the case in FLA, where children do not tend to copy their parents' speech, and where parents do not inform their children when they have made a grammatical error.

Furthermore, the model has not as yet demonstrated how rules of language can be stored in a way that makes them usable by normal procedures - Chomskians argue that this failure is due to the fact that linguistic knowledge is not structured in the same way as other forms of knowledge.

Finally, these models rely on repetition and strengthening of the knowledge networks - but the child manages to learn the rules of her language with very little repetition. Furthermore, it does not explain why children and adult learners of an L2 make predictable, sequenced errors, which they have never heard, nor how it is that children avoid errors, when they have heard no evidence to suggest that the structures not employed are erroneous.

C : Conclusion

We have seen two models of SLA. One of these, which its author claims is based upon Chomsky's account of FLA, and central to which is the idea that language learning is a special skill, posits that SLA is similar to, if not exactly the same as FLA, and that the learner does not have to direct any conscious effort towards the language itself. Language lessons should, in fact, be about something else - something that the student wants to study, in which he is interested. This approach underlies the growing movement in educational establishments to have courses in one or other subject area given in a foreign language. As we have seen, in Canada, this means young anglophones learning French through doing maths. In Scandinavian universities, all science teaching is done in English.

The other model also assumes that SLA is similar to FLA - but does not agree that language learning is in any way different from other kinds of learning. Moreover, it suggests that work on the formal aspects of language is necessary, and that the learner needs to be given the rules of the language. Students need to be encouraged to build up their own rules, and to pack them into networks, where they will become available for work by procedural routines. This may leave the impression that speaking a language is effortless - but the conscious effort is mainly made during the first phase of learning.

In Anderson's model, and in particular in the work of those who have followed it up, the learner is expected to do the work for him or herself, and thus it is that the learner needs to be encouraged to develop specific learning strategies. We shall return to this question later on in the course.

Next week, we shall look more closely at the second of Krashen's hypotheses - that is the idea that the grammar of a language is learnt in a natural order.

(If you wish to comment or ask a question, please write to tmason@timothyjpmason.com)

Home Didactics Top of Page



Author's CV | Site Map | Contact Author |