Applied Linguistics 18/3, 1997, 279-313.


Theory and Practice in Applied Linguistics:

Disconnection, Conflict, or Dialectic?1




We are remaining at the moment the prisoner of our own categorisations.

— Chris Brumfit (1980: 160)


The enduring problems of co-ordinating theory with practice in applied linguistics and language teaching are surveyed in view of the symptomatic disconnections of theory from practice in theoretical linguistics, with the suggestion that how far a theory is applicable to practice is a good measure of how far the theory is valid as a theory. The basic frameworks of ‘doing language science’ are explored in terms of their applicability, including Krashen’s ‘theory’, and an alternative programme is proposed.


1. two ways of ‘doing language science

The relation between theory and practice can be a difficult issue in nearly all domains of human activity. In most domains, human practices were well established long before theories began to be provided and have also played a much more effective role in the history of societies. So we might expect that practice would play a decisive role also in deciding what sorts of theories ought to be produced. In fact, however, theory has typically taken over the leading role and at times has been disconnected from practice altogether. This tactic allows a society or its institutions, especially education, to maintain an official theory of humanity, equality, and efficiency, while also maintaining practices that are symptomatically inhumane, unequal, and inefficient and which consistently favour elite groups (cf. Apple 1985; Giroux 1992; Beaugrande 1996).

The relation between theory and practice also appears uneasy within linguistics, the ‘modern science of language’. In real life, we see a rich mosaic of practices relating to language, ranging from the general operations of language learning and ordinary conversation over to highly specialised strategies of communication such as translating poetry. But in ‘mainstream’ linguistics, we chiefly see a sparse array of theories, some of them self-consciously disconnected from all of these practices.2

A large-scale close analysis of the discourse of linguistic theorists (Beaugrande 1991) indicates that theoretical linguistics is currently in a stagnation of crisis proportions, though linguists are understandably reluctant to admit it. With due poetic justice, the domains where theoretical linguistics has encountered recalcitrant problems are typically also the domains where the theories are least suitable for applications. This correspondence suggests a principle which, according to H.G. Widdowson (personal communication) has hardly been raised in the central work on applied linguistics: how far a theory is applicable to practice is a good measure of how far the theory is valid as a theory. Still, I believe we can mount a fairly strong case for the principle.

We can start by drawing a basic distinction between the two main ways of going about constructing theories of language. Fieldwork linguists, as the term says, go out to ‘work’ in the ‘field’ of cultural and social activities and carefully record and describe what native speakers are actually observed to say. Their theories are both data-driven insofar as observation and induction exert prominent control, and practice-driven insofar as fieldworkers must join in the social practices of the community of speakers in order to gather and interpret the data. If your ‘theory’ of the language is wrong, you’ll soon find out by getting misunderstood, ridiculed, or ignored when you attempt to speak it.

In contrast, homework linguists (to coin a matching new phrase) ‘work’ at ‘home’ or in the office with data that may have come from a variety of sources, ranging from previously completed fieldwork, to corpuses collected from public discourse, to specific discourse types such as samples from language textbooks, and finally to data invented by the linguist during the homework process itself. As we move across this range in this order, control increasingly shifts from observation and induction over to introspection and intuition, whose task is to formulate and describe what native speakers (including the linguists themselves) are presumed to know about their language. The theories are mainly theory-driven insofar as homeworkers construct them according to predetermined methods and standards of design, such as formality, rigour, compactness, and so on. In exchange, data and practice may enter only obliquely and indirectly through artificial, isolated sentences invented by the linguists expressly to make a theoretical point — a situation utterly disparate from ordinary language practices where utterances are intended to convey a message, enhance co-operation, and so forth. Yet these linguists face about and assert that these artificial data are precisely the most revealing for the structure of the entire language.

My distinction cuts the pie a somewhat different way than does the usual distinction between the descriptive linguistics that dominated the ‘mainstream’ of research until the 1960s versus the generative linguistics that moved into the mainstream thereafter. Typically, we do find fieldwork methods at the base of descriptive linguistics, and homework methods at the base of generative linguistics, but not always. Still, my distinction may be the more relevant one for applied linguists and language practitioners, because fieldwork and homework correspond to two different approaches to language learning as well.  

2. fieldwork linguistics versus language learning

 To explore this correspondence, we might begin by asking how far fieldwork research may offer significant analogies to ordinary language learning. Before any theoretical statements can be made about the language, the fieldworker first has to learn the language reasonably well. The task is most challenging when using the ‘monolingual method’ developed chiefly by Kenneth Lee Pike (1944, 1967), where the fieldworker has no prior description of the ‘target language’ to be described and no helpful bilingual interpreters. In return, the fieldworker has the major advantage of doing the linguistic version of ‘situated learning’ (in the sense of Lave and Wenger 1990) within the community as fluency is gradually attained.

Still, the situation of the fieldworker differs from that of the ordinary language learner in at least seven ways: 

(1) The fieldworker is not accompanied into the field by an explicitly trained teacher, but tries to learn from a gradually accumulating circle of contacts and acquaintances, who represent the community’s language from multiple perspectives. Most language learners depend heavily at each stage on one teacher, whose version of the target language may be in some ways untypical.

(2) Fieldworkers are required to assemble their own systematic data corpus, whereas language learners are given textbooks compiled by other people and with means that are often rather unsystematic.

(3) The fieldworker is a mature and educated adult with an intense and round-the-clock commitment to the task as compared to those language learners who are children and adolescents and may find themselves in second-language classrooms for a variety of accidental or administrative reasons such as ‘language requirements’. Voluntary adult language learners (e.g. in the massive programmes in Scandinavia) fall in between the two situations.

(4) The fieldworker has received highly specialised training skills  for writing down utterances in a reliable phonetic alphabet or for distilling out regular patterns from the data that might correspond to categories like ‘nouns’ and ‘verbs’. Most language learners are untrained and unskilled, and the teacher may be uncertain about how they perceive or organise the data from classrooms or textbooks.

(5) The fieldworker is operating under conditions of total cultural immersion. Language is always being observed in relation to human activities that provide both important clues as to what certain utterances mean and a pervasive sensitivity to the general cultural setting of the target language, whether or not such factors are not addressed in the fieldworker’s official ‘linguistic theory’. Learners of foreign languages may have scant contact with the culture.

(6) The fieldworker is conversant with explicit theoretical concepts about language, directed partly toward the native language and partly toward experience with other languages during a period of pre-fieldwork training. Language learners have at most an implicit and disorganised theory of language derived only from their native language.

(7) The fieldworker’s results are intended for presentation to a scientific community of other experts, some of whom may re-examine the data and compare it to the theoretical statements made about the language. Language learners may have no clear idea where they will later display their results.

These seven factors ensure the success of the fieldworker as an ideal language learner, whereas the prospect of failure causes widespread anxiety for ordinary classroom learners.

Still, the parallel between the two learning situations does bear upon the question of how a theory of language might be made to order for practitioners (see section 6). First, such a theory would have to specify the relationships between native and second-language learning. If this is done in sufficiently concrete terms, the coordination of the native language and second-language instruction could be equally attentive to the requirements on both sides. We need to reassess the main differences between available theories of native-language learning — or, to use the conventional term, ‘language acquisition’ (cf. sections 4 and 5) — and of second-language learning. A child learning the native language is in a situation far more similar to the fieldworker’s in regard to cultural immersion; yet the social range of the child’s experiences is necessarily narrower, being more focused on a single family or even on a single parent. Both factors can assist the child: the immersion by supplying rich clues about what things mean, and the range by allowing attention to focus on what is interesting and relevant on a daily basis.

Conversely, both factors are heavily reshaped during second-language learning. In the standard situation, cultural immersion is not available, and the social range is limited to the classroom, where materials are to be learned by compulsion irrespective of whether the learner considers them interesting and relevant.

In absence of immersion, the second-language teacher in the classroom is conscripted into the role of a representative of the community of speakers of the second-language. This role is more problematic and daunting than many teachers would probably care to acknowledge. A fieldworker can circulate among farmers, shopkeepers, householders, and so on, in order to consult a cross-section of social types, each of whose perspective complements the others. In contrast, the second-language teacher may have only a single perspective which is strongly determined by an academic setting and profession, and which emphasises the reading of literary written texts and the grammatical analysis of banal sentences. What would be needed here would be a theoretical framework for expanding the education and training of second-language teachers to provide a spectrum of cultural outlooks on the language and its social and regional contexts and varieties.

Moreover, the teacher has the equally daunting task of trying to offset the learner’s lack of specialised training of the kind fieldworkers have. What is needed here would be a theoretical framework for attenuating this lack by training some basic techniques for describing and discussing language as language. Usually, teachers have to rely on a traditional ‘meta-language’ that is either too technical for learners (e.g. ‘gerund’ and ‘gerundive’) or too philosophical (e.g. ‘a complete sentence is a complete thought’). We need terms that exploit ordinary capacities to use the language (e.g. ‘a complete sentence is one that can be made into a sensible yes-or-no question’) (cf. Beaugrande 1985).

It is also essential that the meta-language be designed for parallel use in both native-language and second-language learning within the same school system. Most schools work at cross-purposes in this regard, so that learners encounter confusions and inconsistencies in terms when moving from one classroom to another. Or, terms are glibly transferred, as from Latin over to English, without appreciating the problems that can arise, e.g., that English categories are much less systematically distinguished by their forms than are Latin ones.

The theory should further provide a framework for a concerted and continual interaction among the personnel engaged in the teaching and learning of both native and all foreign languages. The various needs and issues could thereby be co-ordinated and made into an explicit context for developing flexible and sharable methods.

The desirability seems plain of making a vital part of the second-language teachers’ training be a period of actual cultural immersion for the learning of at least one foreign language, if at all possible the native language of the expected learner population. From a practical standpoint, this requirement would not be unduly difficult to meet for second-language teachers who go abroad and take up residence in a country where the learners’ language is spoken. It would suffice for the host country to provide a lodging and a living stipend for an appropriate period prior to beginning actual teaching. Immersion programs in Latin America or in the Middle East would also be ideal for teachers who will later be working in the United States with the many pupils whose native languages are Latin American Spanish and Middle Eastern Arabic.

This tactic would compensate for the commonplace situation in which the language of the host country does not belong to the handful of languages that are traditionally taught in the schools of such countries as the United States and the United Kingdom. It would not merely put the second-language teacher in a situation fairly similar to the fieldworker’s, but would also involve a similar commitment to the work and a similar opportunity to attain cultural sensitivity. To argue that this is not practicable or affordable for so-called ‘languages of lesser diffusion’ is a stingy rhetorical dodge that allows the bureaucrats in host countries to economise foolishly on the training of teachers of second languages that are vital to personal and social development; and, worse yet, it allows the bureaucrats in the ‘source country’ to maintain a colonial mentality of cultural and linguistic superiority in regard to the cultures and languages in the host countries (cf. Phillipson 1992; Pennycook 1994).

Again from a practical standpoint, it is plainly desirable if a period of actual cultural immersion can be made a vital part of the second-language learners’ training, for at least three reasons. First, the teacher would receive priceless support in the otherwise daunting task of representing the community of speakers. Second, the learners would be far more motivated and committed in their home schools if they were anticipating such an excursion. Third, their experience of a foreign culture would be the best antidote in later life against the rising tides of disrespect for cultural and linguistic human rights and against the language intolerance that threaten both global security and local coexistence (Phillipson, Skuttnab-Kangas, and Ranut [eds.] 1994).

Yet from a theoretical standpoint, the concept of ‘cultural immersion’ is not yet well accounted for. It is certainly not guaranteed by physical presence in the host country; visitors may simply form their own separate groups and keep their distance from the locals and the culture, chiefly for motives of language insecurity or personal anxiety. Or, they may mingle but with a group of locals who want to speak their language instead, a problem already noted by Henry Sweet (1899).3

Furthermore, the commonplace notion of culture as a unified, readily definable entity can be quite misleading. In a ‘modern’ society, ‘sub-cultures’ are proliferated as means for seeking status or compensating insecurities (cf. Apple 1985: Chs. 3 and 4). The culture of adolescents is likely to differ substantially from that of their parents and their teacher (Brake 1980). And further subcultures persist within the multi-culturalism that has reached even such officially mono-cultural societies as the US and the UK, and can be expected to rise considerably in the future (Giroux 1992).

So a practice-driven theory needs to explain how cultural contacts occur and how they might be guided to meet the specific needs of second language learning. Extremely valuable here would be case studies of the whole process of fieldwork not just in linguistics, but also in anthropology and sociology (see for example Hymes [ed.] 1964; Rabinow 1977; Baugh 1983). Case studies should also be assembled of teachers or learners who have been placed in cultural immersion programs.

A practice-driven theory also needs to explain how both the training of teachers and the role of learners in conventional classrooms might effectively offset the lack of opportunity for cultural immersion. Here, we would create a framework for assessing fieldwork on the activities of prospective or practising language teachers and language learners (e.g. Cohen 1976). These studies would offer a vital complement to studies of the informal diaries kept by classroom learners, such as carried out by Kathleen Bailey (see survey in Bailey and Ochsner 1983), which have shed new light on such vital factors as learner anxiety (Bailey 1983).

Without having undergone cultural immersion during the successful acquisition of foreign language, teachers need guidance in appreciating the typical problems of learners arising naturally from the contact between the first and the second language. A ‘practice-driven’ theory drawing on case studies could support an effective programme of deliberately building up cultural sensitivity.

 This resource would be all the more vital when the learners come from a whole spread of different native languages. This situation is a natural reflex of the mobility and multi-culturalism in a ‘modern’ society. The first phase might be to sort the learners out by cultural areas, such as Latin America, Northern Africa, or Southeast Asia. A language teacher with a knowledge of only one of the languages of the area could still attain a cultural sensitivity that would go a long way toward appreciating the situation of the learners and putting them more at ease.

A practice-driven theory should also describe methods whereby ordinary learners can attain at least some indirect access to cultural immersion via the classroom. Practical materials have long been available, such as films and videos; but to my knowledge, we still lack a theoretical framework indicating how such materials and technical resources might best be exploited, and what further materials should be developed and distributed. Video-taped television programmes sent out from the US and the UK to Africa or the Middle East might well be both culturally inappropriate to the host country and culturally distortive or disadvantageous for their image of the source country, particularly in respect to the wanton and graphic violence that passes for ‘entertainment’ in the ‘first world’ as a release valve for with its galloping social and economic tensions.

A second-language teacher whose training does not provide parallels to fieldwork methods and who has not successfully acquired any foreign language is in a position like that of a homework linguist, obliged to rely heavily on intuitions and introspections. Of course, a monolingual and monocultural second-language teacher can attain successful results, just as homework linguists can produce significant descriptions of their own languages. But due to the theoretical principles outlined above, the success rate would seldom be sufficiently reliable or widespread.

3. can fieldwork linguistics do the job?

We can now total up the central requirements for a practice-driven theory explicated so far. It should provide realistic data-driven and practice-driven descriptions of the emergence of the language system at various stages of fluency in both native and second languages and in both natural and pedagogical settings, plus an account of the more effective and expedient means of moving from a less advanced stage to a more advanced one. A body of such theorising should be closely co-ordinated with a library of fieldwork and case studies on the actual activities of the participants, made widely accessible and expandable via Internet.

What are the prospects of fieldwork linguistics supporting an ‘applied linguistics’ able to provide such a theory? In section 2, I have suggested some ways in which fieldwork linguistics could be a highly valuable source. Yet fieldwork linguists are hardly prone to go offering theory for its own sake in the academic marketplace, least of all if doing so means ignoring the different conditions between their own situation and that of teachers and learners of language. Their sense of professionalism rests on the sound conviction that a theory must be a guide for data-driven practice, and must therefore be data-driven itself (I shall return to this in section 4). Even the most monumental and ambitious theoretical framework in the fieldwork tradition, Ken Pike’s (1967) ‘unified theory of the structure of human behaviour’, clings tenaciously to the data. It chiefly explores the methods of phonology and morphology and only occasionally and cautiously reapplies them to the description of other modes of ‘behaviour’ (overview in Beaugrande 1991).

So the applied linguistics deriving from fieldwork linguistics has tended to favour sparse and sceptical theorising with a mechanical and behavioural emphasis. The ‘audio-lingual’ method followed these tendencies to such an extent as to seriously misrepresent the fieldwork tradition in at least three ways. First, its taboo on giving learners any descriptions and explanations of the target language put them in the bizarre position of fieldworkers with no training or guidance. Second, it equated language learning with imprinting behavioural patterns and sequences, thereby starving out the rich and explicit cultural orientation of fieldwork. And third, the method failed to appreciate the nature of the classroom itself as a culturally deprived behavioural ambience wherein rich clues about what to say and what other people mean may not be available to compensate for the sparseness of materials and activities like ‘pattern drills’. In all three ways, the ‘audio-lingual’ method was a practice actually at variance with the body of theory from which it was official derived.

Still, fieldwork methods from phonology and phonetics have contributed enormously to the teaching and learning of pronunciation — surely the most successful application of linguistics so far (see now Dalton and Seidlhofer 1994). Morphology and syntax made a smaller impact, partly just because fieldwork had concentrated on lesser-known languages that are seldom taught as second languages in schools, and partly because many teachers kept on relying upon the numerous traditional grammars devoted to the languages that are frequently taught.

What still remains to be done for the application of fieldwork linguistics is to supply a detailed theoretical and empirical account of the ways in which fieldwork both resembles and differs from ordinary language learning and from organising language materials for the classroom. As suggested in section 2, the issues are complex and have not received the consideration their potential importance would merit, mainly due to the practical difficulty imposed by the conditions for exploring them, e.g., the labour of doing large-scale case studies of cultural immersion among different groups of teachers or learners. Future work may profitably be invested here.

4. can homework linguistics do the job?

Now, what are the prospects for the application of homework linguistics? Its ‘mainstream’ branch has professed a lively interest in ‘language acquisition’, but has pointedly used this term in a rather different sense from ‘language learning’. A detailed analysis of their own theoretical discourse has led me to conclude that this interest and this difference were chiefly rhetorical  moves in their campaign gain control over theoretical linguistics while discrediting fieldwork. Chomsky’s followers evidently believe to have accomplished this high-minded goal. Newmeyer’s (1980: 249f) Linguistic Theory in America  presented Chomsky’s as ‘the world’s principal linguistic theory’, for which ‘no viable alternative exists’; ‘the vast majority’ of ‘linguists’ ‘who take theory seriously4 acknowledge (explicitly or implicitly) their adoption of Chomsky’s view of language’. However, this self-confidence of the Chomskyan school has grown increasingly brittle, nurtured mainly by not reading any linguistic research outside their own circle and by not acknowledging the burgeoning body of criticism and opposition.

In the late 1950s, Chomsky presented himself as a rescuer at the time when fieldwork linguistics was in trouble because its stringently data-driven methods were not transferring smoothly from phonology and morphology over to syntax (cf. Beaugrande in 1996). ‘Phonemes’ and ‘morphemes’ are theoretical units corresponding straightforwardly to practical units (sounds, word-parts, words) that are ‘in’ the data and can be reliably isolated by segmenting recorded utterances into the smallest pieces. ‘Syntactic rules’ are not ‘in’ the data and must be postulated or reconstructed. Chomsky’s ‘transformational grammar’ proposed a new and seemingly rigorous and compact way to do so: design a system of rules that describe sentence structures in terms of other sentence structures. Such a system would be an economic way for fulfilling the ambition of linguists to provide a complete description of a language. The net effects of this research, however, have in my view amply shown that no such system exists, due to one simple fact, already noted among the Prague School inaugurated by Vilém Mathesius (overviews in Beaugrande 1992, in press): the formation of sentences is not determined exclusively by linguistic rules, but also by the cognitive and social constraints of contexts (Lakoff 1987; Harris 1990; and see now Halliday 1994a; Beaugrande 1996). But a whole generation of ‘syntactic theories’, including Chomsky’s, continues to search in vain for such a system, and hides the stagnation and inadequacy of their projects behind an evasive rhetorical double-tracking we shall see later on, and behind a congealing thicket of impenetrable terminologies and formalisms (cf. Escribano 1993). These moves cannily cash in on the general valuation of theory over practice in modern society and especially in universities, as I mentioned at the outset.

Chomsky’s polemical talent for ‘negative campaigning’ was most conspicuous in his much-deserved deconstruction (Chomsky 1959) of the behaviourist theory of language spearheaded by B.F. Skinner (1957), which gave no sense of how language is organised as language and not simply as ‘verbal behaviour’. But more crucial to his project of gaining control over theoretical linguistics was his negative campaigning against fieldwork linguists, whose methods and cultural immersion ensured that their results would not be subject to the same strictures as Skinner’s vacuous extrapolations from animal conditioning experiments (cf. discussion in Beaugrande 1984).

A key argument in Chomsky’s campaign was that fieldwork would never attain ‘the deeper and more important notions of linguistic theory’, due to its ‘limitation-in-principle to classification and organisation of data’ (Chomsky 1965: 18f):

there is no reason to expect that reliable operational criteria for the deeper and more important notions of linguistic theory [...] will ever be forthcoming [because] knowledge of the language, like most facts of interest and importance, is neither presented for direct observation nor extractable from data by inductive procedures of any known sort.

This campaign was capped by the straightfaced declaration that the ‘inadequacy’ of the ‘grammars for natural languages’ based on fieldwork had ‘been established beyond any reasonable doubt’ (1965: 67). Moreover, Chomsky (1957: 52f) judged it ‘unreasonable to demand of linguistic theory’ that it stipulate a ‘discovery procedure’ ‘for actually constructing the grammar, given a corpus of utterances’ — just what his own theory obviously couldn’t do. ‘How one might have arrived at the grammar’ would not be ‘relevant to the programme of research’; ‘one may arrive at a grammar by intuition, guess-work, all sorts of partial methodological hints, reliance on past experience, etc.’ (1957: 57) — the most convenient ‘homework’ approach imaginable. He promised that ‘once we have disclaimed any intention of finding a practical discovery procedure’, ‘certain problems that have been the subject of intense methodological controversy simply do not arise’ (1957: 56). In the event, however, methodological controversies got more violent because linguists no longer needed to pass through a phase of heavily data-driven and practice-driven discovery before building a theory; so Chomsky was soon embroiled in disputes over ‘notions of linguistic theory’ with other linguists, some of whom had been his own pupils and had picked up their polemical talents from watching him.

Naturally, the fieldwork linguists were stunned by such accusations and caught wholly unprepared. They had never questioned for a moment that linguistic theory must supply discovery procedures, and had devoted their lives to developing and improving them. In their eyes, the ‘adequacy’ of a ‘grammar’ or any other description depended directly on ‘how one arrived at it’ — not just from ‘intuition, guess-work’, and ‘hints’, but from painstaking analysis of a ‘corpus’ of recorded authentic data. Any fieldworker would find it patently absurd to declare that we can ‘establish the inadequacy’ of not just one ‘grammar’ but of a whole class of ‘grammars’ at one stroke, and that this class is precisely the ‘grammars’ extracted through discovery procedures!

The context of this campaign is vital for understanding Chomsky’s professed interest in ‘language acquisition’. If you roundly reject fieldwork as base for theory, and if fieldwork resembles ordinary language learning in at least some ways, then you naturally want to invent a new theory of language learning that looks as little as possible like fieldwork and as much as possible like homework — and that is precisely what Chomskyan linguistics did. Cleared of its technical verbiage, their ‘theory’ says that the acquisition of language is not primarily data-driven or practice-driven. Instead, the child is cast as a miniature homework linguist and starts off with a ‘universal theory of language’ and specifies the theory for the native language by hypothesising from some abstract aspects of what Chomsky calls ‘primary data’ — what the child hears people actually say. To clinch this idea, Chomsky (1965: 201) pointedly devalued ‘primary data’ for being, ‘from the point of view of the theory he [the child] constructs, deficient in various respects’; ‘much of the actual speech consists of fragments and deviant expressions’. Such data would therefore be inadequate for a child learning the language just a much as for a fieldworker constructing a theory of language. Only if the child had a prior theory of language could acquisition proceed; and since no theory could be extracted from data before acquisition starts, the theory would have to be innate. Each rhetorical step in this line of reasoning compelled a further step, so that the initial resolve to undercut fieldwork linguists launched a circular chain of further claims, each of which was to be made plausible by the others and not by methodical ‘discovery procedures’. The gap between what real native speakers say or what real children learning the language do versus what the theory said or implied about them was continually blurred by rhetorical double-tracking to combine each theoretical construct with a conveniently commonsense (mis)interpretation. The very term ‘generate’, technically meaning ‘assign a structural description to’ a sentence that has already been produced (usually invented by the linguist), invites us to think of the human acts or processes of ‘producing’ the sentence; Chomsky warned us not to, but he himself pictured the ‘grammar’ ‘producing language’ or ‘strings’ or ‘sentences’, referred to ‘the process of generating sentences’, and equated ‘generating’ with ‘creating’ (1957: 48, 11f, 13, 18, 30f, 38, 45f, 103, 35; 1965: 135f).

Of special importance for the discussion here is the double-tracking by means of ‘systematic ambiguities’ that seem innocuous but in fact directly incorporated the reality and validity of ‘the linguist’s account’ into the discourse of theoretical linguistics:

Using the term ‘grammar’ with a systematic ambiguity to refer, first, to the native speaker’s internally represented ‘theory of his language’ and, second, to the linguist’s account of this, we can say that the child has developed and internally represented  a generative grammar in the sense described. […] we are again using the term ‘theory’ — in this case ‘theory of language’ rather than ‘theory of a particular language’ — with a systematic ambiguity to refer both to the child’s innate predisposition to learn a language of a certain type and to the linguist’s account of this. (Chomsky 1965: 25)

Thus, just to use the terms ‘grammar’ and ‘theory of language’ in his ‘sense’ would seem to automatically equate his ‘account’ with what it proposed to account for, and to assert that ‘the native speaker’ does indeed hold an ‘internally represented theory of his language’, and that ‘the child’ does indeed have an ‘innate predisposition to learn a language of a certain type’ — assertions craftily downplayed by unobtrusive possessive constructions (‘native speaker’s theory’, ‘child’s predisposition’), instead of more obtrusive subject-verb constructions (‘native speaker holds a theory’, ‘child has a predisposition’). Such discourse evades the scientific responsibility to provide empirical evidence to prove both that the ‘linguist’s account’ is valid, and that ‘internally represented theory of language’ and the ‘innate predisposition’ are there in the first place to require an ‘account’.

Of comparable importance for our discussion is the double-tracking about ‘underlying competence as a system of generative processes’ (1965: 4): 

Any interesting generative grammar will be dealing, for the most part, with mental processes that are far beyond the level of actual or even potential consciousness. Furthermore, it is quite apparent that a speaker’s reports and viewpoints about his behaviour and competence may be in error. Thus a generative grammar attempts to specify what a speaker actually knows, not what he may report about his knowledge. (1965: 8)

The speaker is not ‘aware’ or ‘conscious’ and is thus prone to make ‘statements’ and ‘reports’ of doubtful ‘interest’ and ‘accuracy’ or downright ‘errors’. In contrast, linguists ascend the status of exceptional human beings who are able to become ‘aware’ and ‘conscious’ of ‘mental processes’ after all, and who can ignore at will what speakers ‘report’ (cf. Beaugrande and Seidlhofer, in review). Technically, this rhetorical evasion was overkill, since we have already been informed that ‘linguistic theory is concerned primarily with an ideal speaker-hearer in a completely homogeneous speech-community, who knows its language perfectly’ (1965: 3) — and just try finding such a person to make ‘reports’!

It is in this rhetorical context that we can appreciate why Chomsky reanimated and reinforced the view, well-established in linguistics at least since Saussure (1916), that linguistics should describe language by itself. Fieldwork methods had ensured that the language was always observed in cognitive and social contexts, whence the notion of language as part of the ‘unified’ system of ‘human behaviour’ envisioned by Pike and cited in section 3. Because Chomsky’s own theory of language gave no clue of how language relates to such contexts, he was impelled to claim, in absolute terms, that language is a separate domain or faculty of human beings. The inevitable next step was to propose a theory of language acquisition also running only on language — in fact, on only those aspects of language addressed by his own ‘theory’. He thus had to invent his famous ‘language acquisition device’ (hereafter LAD), as a ‘hypothetical’ but ‘useful and suggestive framework’ for ‘posing and considering’ ‘certain problems of linguistic theory’: ‘the theorist has the problem of determining the intrinsic properties of a device capable of mediating’ between ‘the primary data’ as ‘input’ and the ‘grammar’ as ‘output’ (Chomsky 1965: 47). For rhetorical  expedience, its job was defined precisely as doing what Chomskyan homework linguistics did: extracting formal rules out of sets of sentences while abstracting away from the ‘deficiency’ of ‘primary data’ (quoted above). He gave no clear notion himself of how it did the job; after all, his campaign against fieldwork linguistics had led him to dismiss the question of how a linguist ‘might have arrived at the grammar’, as we saw. By the same token, he gave no clear notion of how a child could learn a language as a set of formal rules, witness his audaciously evasive ‘instantaneous model’ wherein ‘successful language acquisition’ happens in a single ‘moment’ (1965: 36). Otherwise, he would have had to specify how the LAD proceeds by building up rule after rule. Doing so would have unsettled his vision of language being a formal system of rules that operate on other rules; such a system can only operate in its complete state, and is not designed to ‘add’ rules but only to ‘generate sentences’. The alternative would be to use contexts to compensate for missing rules, just as fieldworkers do; and Chomsky’s campaign precluded this prospect. He dourly conceded that ‘it would not be at all surprising’ if ‘normal language learning requires use of language in real-life situations’, but he doubted, a bit absurdly, that ‘information regarding situational context’ ‘plays any role in how language is acquired, once the mechanism is put to work and the task of language learning is undertaken by the child’ (1965: 33).5  For good measure, the ‘acquiring’ of ‘grammar’ despite ‘the degenerate quality and narrowly limited extent of the available data’ was asserted to owe the ‘striking uniformity’ Chomsky believed it has to being ‘independent of intelligence, motivation, and emotional state’ (1965: 58).

I have recalled these well-known and often-cited lines of argument in order to highlight the rhetorical double-tracking whereby Chomsky came to propagate a ‘theory of language acquisition’ which was expressly not a practice-driven ‘theory of language learning’. His campaign depended critically on the bald assertion that data-driven ‘practical discovery procedures’ are forever cut off from the ‘deeper and more important notions of linguistic theory’, which he simultaneously claimed the right to identify and define; and he cynically, and rightly, expected ardent support from upcoming homework linguists had no taste for the arduous labours of fieldwork and who were delighted to find ‘intuition’ and ‘guess-work’ so nicely legitimised. This assertion and this claim authorised his own school to invent ‘notions of linguistic theory’ in advance of data, along with a forbiddingly complicated apparatus for relating the theory to data in technical and evasive ways that are deliberately hard to pin down and challenge.

Further authorisation was naturally sought by means of a theory of ‘language acquisition’ being a similar mode of theorising done by the child in advance of data and with a complicated apparatus furnished by genetics — the ‘LAD’. Ironically, whole branches of research ‘psycholinguistics’ and ‘second language acquisition’ have sprouted up, taking it as given that the LAD exists and only exploring miscellaneous claims about how it works or in which order.

 As time passed, all these confident, untested, and largely circular claims were entrenched by constant repetition and citation and were passed on to new waves of linguists and language specialists who were not properly aware of the original rhetorical context during an aggressive academic campaign. The claims provided the background and terminology in which further discussions were routinely carried on, like a dominant but invisible ideology that passes for reality and the natural order (cf. Beaugrande 1996). The participants in the discussions might then be quite unaware of how many untested theoretical claims they were taking as given when the real issue should be to subject the claims to empirical justification or practical evidence. Here, theoretical linguistics has already begun to made us ‘the prisoner of our own categorisations’, as Brumfit (1980: 160) commented on applied linguistics.

Today, the original rhetorical context for Chomsky’s ‘theory’ is rarely appreciated for what it was: a campaign for building a theory while discrediting a whole class of competing theories. Chomsky was strangely eager to throw out the baby with the bathwater and the bathtub as well, because he had a new bathtub to sell, which later became a whole series with a luxurious range of fixtures and attachments; and he had a new baby to put in too, one equipped with the LAD at no extra cost.

 It seemed like a good-news, bad-news situation. The bad news was that the ‘important notions of linguistic theory’ would have to be reformulated essentially from the ground up, irrespective of the impact on the careers and publishing opportunities of linguists using descriptive fieldwork approaches, who were firmly given to understand that their professional work in the ‘coverage of a large mass of data’ was not ‘an achievement of any particular theoretical interest or importance’ (Chomsky 1965: 26). Moreover, language learning would have to be reconceptualised as a heavily theoretical operation called ‘language acquisition’ and related in abstract, complicated ways to ordinary encounters with language data. The good news was that you could jump right into ‘linguistic theory’ without the intense labours of being trained and tested in fieldwork and of applying scrupulous analytic methods to large corpuses of authentic data; you could just invent a few sentences in your own native language and set to work inventing formal rules to ‘assign them structural descriptions’. Moreover, you could jump right into ‘language acquisition theory’ without the labour of teaching anybody a language or doing large-scale longitudinal studies of language learners. And best of all, your theory would be nearly immune to empirical or practical refutation because it would be encircled by protective theoretical constructions, abstractions, and idealisations founded upon double-tracking dichotomies (‘competence’ vs. ‘performance’, ‘generate’ vs. ‘produce’, ‘learning’ vs. ‘acquisition’, ‘deep structure’ vs. ‘surface structure’ etc., etc.), which all imply that, in Widdowson’s (1980: 166) words, ‘language, like God moves in a mysterious way and outside the range of the common man’s awareness’. Unfavourable evidence in ‘primary data’ could easily be deflected as mere ‘performance’ or ‘surface structure’, for which the theory could not be held accountable. We might well ask here whether Chomsky’s proposals even qualify as a theory, given these largely untestable factors.

A number of prominent linguists, including Wallace Chafe, Michael Halliday, John Sinclair, František Daneš, and Ruth Wodak, hold (as I do myself) that this ‘good news’ has in the long run turned out to be very bad news for the field as a whole. The homework linguistics of Chomsky’s school has fostered an unfortunate rhetorical ambience of top-heavy, divisive theorising and gratuitous confrontations, turning every theoretical issue of question into an occasion for vigorous sparring matches. It has become unreasonably hard to take a stand on any important issue without being promptly attacked or co-opted (or both) for implying a position on endlessly disputatious topics like ‘universal grammar’ or ‘deep structure’.

In such an ambience, a theory is prone to be designed chiefly as a weapon, not to describe or explain language as we find it in the world of real people, but to make language out to be exactly the opposite of whatever your opponent says, or (what is more to the point) whatever you allege your opponent is saying. Fieldwork linguistics never claimed to be a ‘theory of language acquisition’, but rather a theoretical guide for discovery and description; and as such it remains unexcelled. It had no great affinities for ‘notions of linguistic theory’ as an end in themselves, precisely because it had to be established in the sparse and sceptical environment of ‘unified science’ centring on the ‘hard sciences’, where it was academically prudent as well as productive to stick very close to observed data. Hardly had linguistics become fully established and begun to offer well-paying jobs when a crew of self-proclaimed ‘revolutionary’ homeworkers appeared with the plan to ignore the significant success of fieldwork linguistics in describing hundreds of previously undescribed languages, many of them threatened with extinction, and to denounce fieldwork for being incapable of attaining a set of ‘deeper and more important notions of linguistic theory’ whose necessity had not been realised before and has, I submit, never been convincingly shown since then, but merely imposed by self-confident hand-waving rhetoric of the type we have seen here.

Sadly, this campaign on behalf of chimerical and divisive ‘deeper notions’ has distracted many linguists, both theoretical and applied, away from the really urgent issues, such as the massive communication problems in our increasingly global, multicultural, and multilingual societies (cf. Halliday 1994b; Pennycook 1994; Beaugrande 1996). We should devote our efforts to a generation of practice-driven theories that are ‘deep’ and ‘important’ not just because they serve the rhetoric and academic interests of one cadre of theoretical linguists over the others, but because they account for the rich diversity of languages and of their connections with society and cultures, including the processes of learning to speak and learning to speak better.  Some of the learning is biologically prepared and genetically transmitted as innate capacities; some of it is derived from experience in real-life situations; and some of it is explicitly and consciously learned from parents, siblings, friends, teachers, and so on. So humans come by their ‘language competence’ partly by subconscious, automatic processes, partly by conscious learning, and partly by making canny guesses about what words or people mean in the contexts where you hear or read them.

We can now return to the principle proposed at the outset: how far a theory is applicable to practice is a fairly good measure of how far the theory is valid. By this standard, far from being an explanation or account of language and language acquisition, Chomskyan theory is a static enclosure of circular technical constructs created chiefly to subserve academic politics. If we ask, ‘can this homework linguistics do the job for language teaching?’, the answer is likely to be: according to this linguistics, there is no job, because acquisition is the competent hands (or cogs) of the ‘language acquisition device’ that always works and in the same way for everybody, serenely undaunted by the woeful ‘deficiencies’, ‘fragments, and deviant expressions’ in ‘actual speech’. Perhaps our job would be to make language learners conscious of these deficiencies’ in hopes of getting them to match up better their ‘performance’ with their ‘competence’, and their ‘surface structure’ with their ‘ deep structure’ — which would be, under more technical terms, essentially what authoritarian traditional grammar proposed to do. But how we would go about it is totally unclear insofar as we can’t get a good look at ‘competence’, and ‘ deep structure’, because they have always been abstract idealisations belonging to the ‘ideal speaker-hearer in a completely homogeneous speech-community, who knows its language perfectly’ — and whom we’ll never get to see or hear, not even in the most elite departments of linguistics and philosophy. 

5. a case study: the discourse of stephen krashen

 Interestingly, the ‘theory’ of Stephen Krashen proposed to be an application of this homework linguistics after all. He claims his ‘theory’ ‘support Chomsky’s position, and extends it to second-language acquisition’ (1985: 3).6 Finding such an enterprise paradoxical and anticipating some interesting rhetoric, I have carried out a discourse analysis of his the 1985 volume The Input Hypothesis: Issues and Implications. He certainly does follow Chomsky by mounting a campaign to gain control by means of complicated theoretical arguments, albeit this time over applied linguistics. His discourse is a bit less self-confident than that of Chomsky’s school (e.g. Newmeyer’s cited in section 4) about whether he has achieved this. To be sure, he equates ‘current second-language acquisition theory’ with his own work (e.g. 1985: 100). But the extreme difficulties of ‘extending Chomsky’s position’ this way have rendered him a bit more cautious than Chomsky’s school and have impelled him to read linguistic research outside his own circle in his continuing campaign to show that it either supports his theory or at least doesn’t contradict it — which he likes to suggest is much the same thing.

Like Chomsky, Krashen has launched his campaign by telling people — practitioners and teachers this time rather than fieldworkers — that what they have been doing all along is at odds with ‘linguistic theory’, i.e. with Chomsky’s. Doing so capitalises on Chomsky’s own prestige on the canny (and justified) assumption that many practitioners have no clear idea of what Chomsky’s theory really says (after all, they were never supposed to). The ones who do, like myself, won’t be impressed of course, but the number of those who don’t, or who don’t feel competent to challenge Chomsky, might suffice to muster a substantial following, as Krashen apparently has.

Just as Chomsky presented himself as a rescuer when fieldwork linguistics was in trouble with syntax, Krashen has exploited the insecurities and uneasiness of language practitioners and teachers who had long been discontented with the otherwise sparse and sceptical applied linguistics and were thus eager for change. They were on the rebound, as it were, from various theories and methods whose results had proven disappointing. Krashen easily profited from on Chomsky’s well-orchestrated rebuke of Skinner’s behaviourism, and on the growing perception that its widely accepted ‘audio-lingual’ method had not fulfilled its promises. The newer theories that had been put forth in the interim had not managed to achieve an equally wide acceptance or to provide a new consensual framework: the ‘cognitive code-learning theory’ because the notion of ‘language as code’ is not very helpful or well-defined; and the ‘communicative approach’ because so much of both theoretical and applied linguistics had not yet dealt with communication in realistic and productive ways (cf. Beaugrande, in review).

Krashen fervently adopted Chomsky’s (1965: 47) ‘language acquisition device’, which, as I noted, had originally been proposed  as a ‘hypothetical framework’ during Chomsky’s plan to invent a theory of language learning that looks like homework linguistics. The frank resemblance between what the child was claimed to do and what the homework linguist actually does do was handled not as a self-serving cagey move to replace the child with an idealisation and replace the hard labour of studying real children with glib speculation based on what linguists do, but as a serious argument for the validity of Chomsky’s theory. Besides, hearing that children will automatically and inevitably learn their language, thanks to the stalwart LAD, has immense appeal for pedagogues and teachers, even when, for reasons I hope to have made clear, they are not told how the device does the job.

This strategic ‘hypothetical device’ has gradually moved out of its original rhetorical context of theoretical argument and switched from a ‘hypothetical framework’ to a separate ‘mental organ’. This odd term would seem to designate a piece of the human anatomy — Krashen (1982: 96) at one point portrayed the LAD as ‘a part of the brain’ — but one that cannot even be detected, let alone observed. It’s not a physical organ, yet is claimed to resemble one by only ‘functioning automatically’ and ‘subconsciously’ (Krashen 1985: 4, 100, citing Chomsky 1975). So learners could never even become aware of this ‘organ’, let alone control it, any more than ‘certain cells in the embryo choose’ ‘to become an arm’  (Chomsky 1975: 71). Equally expedient was the implication of ‘uniformity in the language faculty’, such that ‘the language acquisition device operates in fundamentally the same way in everyone’ (3). This proviso authorises you to disregard variations in personality, social status and so on, about which Chomsky’s ‘theory’ had nothing to say anyway.

The LAD became an anchor-point for Krashen’s own static enclosure of technical constructs, these too, like Chomsky’s, evasively disconnected from real data. These constructs too prop each other up in circular ways, e.g.: language must be acquired by the LAD; the LAD must exist because language gets acquired. Just in one book, it’s easy to spot Krashen’s blunt rhetorical strategy of promoting his theses — he calls them ‘hypotheses’, though they aren’t tested but merely imposed by his own rhetoric, again like Chomsky — simply by repeating them like mantras, especially when he wants to reject an alternative account. As for Chomsky’s proposals, we might well ask here whether  Krashen’s  even qualify as a theory, given the obstacles to testing it; surely brute repetition cannot replace empirical validation.

In sum, Krashen has transposed, with minor modifications, Chomsky’s theory of first language acquisition over into a theory of second language acquisition composed largely of hypothetical processes and undetectable operations and programmatically disconnected from the ordinary observable realities of deliberate language learning. The most breathtaking property of this brave new world (that has strange people in it) is its resolute disconnection from the ordinary world of teachers and learners. To ‘support Chomsky’s position’ and his own, Krashen is prepared to declare acquisition completely independent of learning. So Krashen presents a theory which, at one and the same time, purports to explain the result of practices, i.e., the fact that humans do various activities and at some point have learned a language, while expressly denying that the usual practices can have led to that result. Such a theory, for which I know of no parallel in any science, requires a divisive, confrontational rhetoric of affirmation (my way is how acquisition does and must proceed) plus exclusion (their way is how acquisition does not and cannot proceed).

Applied to language pedagogy, the ‘theory’ views acquisition as an ‘automatic’ and ‘subconscious’ outcome serenely indifferent to what teachers and learners are ‘consciously’ doing. Like Chomsky, Krashen equates the acquisition of language with the acquisition of grammar rules, and competence with the person’s set of acquired rules. He also invokes a ‘natural order’ prescribed by Mother Nature herself and presumably programmed into the ‘mental organ’ of the LAD, where we’ll never get a good look at it. ‘We progress along the natural order’ by ‘understanding input that contains structures at our next “stage” —  structures that are a bit beyond our current level of competence’ (2). Since the child can’t know what rules are yet to be acquired, this ‘order’ must somehow be managed by the LAD; yet if so, it would apparently have to the same order for all languages, because the LAD can acquire any one of them.7 That implication seem highly suspicious, the more so when Krashen defines the ‘natural order’ in terms of ‘structures’ and avows that ‘natural orders have been found’ for ‘several languages’ that were ‘investigated’ (2, 20). Actually, these findings bear only on the ‘acquisition’ of certain morphemes (cf. Gregg 1984; White 1987), and not on the rest of the grammar, and even less on lexicon or vocabulary.

At all events, a damaging contradiction impends: competence is the human ability to understand language, yet humans somehow proceed by understanding things beyond their competence — unless ‘competence’ is define more evasively not as the ability to understand  but as an entity merely ‘underlying’ that ability.8 But, as with Chomsky, we may detect some rhetorical double-tracking here to combine a theoretical construct with a conveniently commonsense (mis)interpretation, namely the ordinary sense in which a real person who has learned a second language would be judged ‘competent’ to actively speak it and not just, say, to passively get the gist of other people’s speech.

A problem arises: insofar as the theoretical construct of ‘competence’, being a system of formal rules, is deterministic, it could not allow for (or ‘underlie’) the understanding of structures for which no rules have been ‘acquired’, except by analogy to those it has.9 Krashen’s solution is to modify Chomsky’s denial, cited in section 4, that ‘use of language in real-life situations’ ‘plays any role in how language is acquired’: ‘we are able to understand language containing unacquired grammar with the help of context, which includes extra-linguistic information, our knowledge of the world, and previously acquired linguistic competence’ (2).

Significantly, Krashen invokes ‘evidence’ that the ‘natural order’ is ‘independent of the order in which rules are taught in language classes’ (1). So ‘the language teacher need not attempt deliberately to teach the next structure along the natural order — it will be provided in just the right quantities and automatically reviewed if the student receives a sufficient amount of comprehensible input’ (2). I don’t see how one structure could be ‘provided in quantities’; perhaps Krashen means it will be exemplified by quantities of input. But if so, the quantities could hardly be called ‘just right’, seeing that he nowhere gives us measures of quantity and generally just implies the more the better.

At all events, Krashen roundly avows, in magisterial tones reminiscent of Chomsky’s rhetoric about the ‘deeper and more important notions of linguistic theory’, that the role of ‘comprehensible input’ is ‘the fundamental principle in second-language acquisition’ and therefore ‘the most important part of the theory of second-language acquisition’ (vii, 3). His ‘Input Hypothesis’ states that humans acquire language in only one’ ‘amazingly simple way’ — ‘by understanding messages’, or ‘by receiving comprehensible input’ (vii, 2). He sells this idea with a grandly confident promise: ‘to the extent’ his ‘Input Hypothesis’ ‘is applied, to that extent will our language programmes be more productive and efficient for our students and easier and more pleasant for teachers’ (viii). Despite ‘objections’ ‘made on practical grounds’, ‘the theory promises much more successful language acquisition in the classroom, both in fluency and accuracy, and is far easier to apply than any of the alternatives’ (54).

Stated in these ‘amazingly simple’ terms, what Krashen offers is not a ‘hypothesis’ at all but an incontestable truism (cf. Gregg 1984). Of course people must have input in order to learn; otherwise they would be cut off from the outside world. And of course the input must be comprehensible; what is incomprehensible would not be ‘input’ at all except in the technical or mechanical sense, in communications engineering, that a system can receive an input of mere ‘noise’.

How then can Krashen claim that ‘comprehensible input’  is ‘the one essential ingredient’ that ‘has escaped us all these years’ (ii)? The answer would seem to be that ‘comprehensible input’ has a more technical and evasive meaning: not so much the messages you do understand in practice, as the structures you can understand in theory by virtue of your competence, i.e., your set of ‘acquired rules’. This account brings back the same problem of a deterministic ‘competence’ that would rule out learning in principle if you could only comprehend what is already comprehensible. So, as we saw, Krashen primly makes allowance for a narrow margin where ‘extra-linguistic information’ from ‘context’ and ‘knowledge of the world’ enables people to ‘understand language containing unacquired grammar’. But he unwisely retains the stringent restriction that this margin can absorb only a ‘rule’ that is ‘i + 1’, i.e., ‘next in line within ‘the natural order’ (39). Unwise because he might be asked to say just what the natural order is and not just that ‘natural orders have been found’ and that the ‘natural order is independent of teaching order’.

Yet Krashen is so intent on establishing his ‘theory’ over all others that he feels impelled fend off any prospect of ‘beating the natural order’ whereby ‘any rule can be acquired at any time’ (41). He does so by boldly denying that ‘learning can become acquisition’, lest there be ‘two paths to acquisition, one way via comprehensible input, and another via conscious rules’. Here, Krashen’s self-centred exclusion of rival theories animates his rhetoric to create an unbridgeable split between conscious learning versus unconscious acquisition. In a stern warning, any ‘approach assuming the correctness of the “learning becomes acquisition” view’ is accused of ‘creating an impossible situation for teacher and student; not only do we have to teach and learn all the rules of grammar, we must also teach and learn the subtle and numerous rules that relate language functions and grammatical rules’ (55). ‘If sentence-level grammar is too difficult and complex to teach and learn, which it is, adding sociolinguistic rules to the students’ burden can only make the situation worse. In reality, many of these rules are acquired, both in the language classroom and in the real world. Our responsibility in language teaching is only to put the student in such a position that he can continue to acquire such rules outside the language class’ (55f).

The rhetorical work achieved by these discourse moves is further animated by the prospect that Chomsky’s own theory projects a view of ‘sentence-level grammar’ that really is far too ‘difficult and complex to teach and learn’ — a problem that probably helped compel him to propose the ‘language acquisition device‘ in the first place, as I suggested in section 4. Also, Chomsky’s theory is notoriously shy on ‘rules that relate to language functions’, and even more on ‘sociolinguistic rules’. So the major gaps in Chomsky’s theory become for Krashen the issues that should not or cannot be taught, and teachers are warned to fear and shun them or suffer ‘impossible’ consequences. The warning both plays on teachers’ present insecurities and purports to explain why the teaching of rules hasn’t done so well in the past. I have already stated what I submit is the real reason: much of language use is simply not governed by linguistic rules at all, but by strategies for fitting utterances to contexts  and to cognitive and social constraints.

Besides, ‘providing more contextualised practice of grammatical rules’ and ‘teaching sociolinguistic rules’ are hallmarks of the ‘communicative approach’, Krashen’s main strongest rival upon which he visits the unintentionally comic dismissal that it doesn’t ‘fit the theory’ (55), whereas his theory was of course designed such that it wouldn’t fit the ‘communicative approach’. In real practice, though, good teachers frequently do provide functional and sociolinguistic information, though they may well not call it by those terms, for example, whenever they indicate that a certain expression is ‘colloquial’, ‘slang’, ‘respectful’, ‘old-fashioned’, and so on. Since neither Chomsky’s nor his own theory takes account of all that, Krashen advises teachers to leave it all to ‘the real world’ and to teach only what is ‘necessary to avoid truly insulting and impolite behaviour’ (56), Evidently, the lesser perils of seeming nosy, quaint, foolish, snobbish, and so on, would be matters for students to learn the hard way out on their own.

Yet another self-centred rhetorical exclusion is performed against the ‘output hypothesis’ he ascribes to ‘communicative theories of acquisition’ (35, 55), his leading rival. There, ‘competence develops via output practice in communicative situations. Specifically, the performer acquires rules by “trying them out” in communicative situations: when he experiences communicative success, his hypothesis about the rule is confirmed; if this happens often enough, the rule is acquired’. The ‘output hypothesis’ is made into an obvious straw man: a flat denial that input alone can ever do the job and that ‘input-type teaching methods’ can have any ‘success’ (36). This straw man ‘predicts that language acquisition is impossible via listening alone, that radio and television are always useless, that acquisition via subject-matter teaching will occur only when the student talks, and that reading is never helpful’ (35f, i.a. [= italics added]).

Krashen does hedge, but only a little: a ‘denial of the Output Hypothesis is not a denial that language acquisition involves hypothesis-testing. This hypothesis-testing, however, according to the Input Hypothesis, takes place on a subconscious level. In addition, it does not require production, nor does it involve communicative success’ (36). Moreover, he does not deny that ‘the performer’s own output can serve as comprehensible input to his own language acquisition device. Even if rules are consciously learned by the performer, if he uses them correctly, he conceivably understands the message he conveys [I can’t ‘conceive’ how the ‘performer’ could be said not to ‘understand the message’!] and thus provides himself with comprehensible input containing a structure he had not yet acquired.’ But here too, ‘the performer’s own output will “count” as input for language acquisition only if the structures involved happen to be at the acquirer’s current i + l’ (cf. § 56). These predictable hedges enable Krashen to steer clear of a genuine compromise by making modest expropriations from output strictly on his own terms: by emphasising the ‘subconscious’ activities over ‘production’ and ‘communicative success’, and by circuitously interpreting output as input. Compare also his stark opposition: ‘the crucial element that peer interaction provides’ ‘is not output practice but low-filter comprehensible input’ (66); we shall examine his ‘filters’ in a moment.

Self-centred rhetorical exclusion wins out once again, albeit less openly, when theory and practice get placed in opposition. ‘Much of the difficulty we experience in language education comes from our efforts’ to ‘develop speaking skills via speaking practice’, and to ‘develop grammatical accuracy via grammar drill and error correction’; this tendency is encouraged when we use tests that require and focus on output and grammatical accuracy’ (92). Why attention to ‘output’ must produce a ‘focus on grammatical accuracy’ is not explained; learners can surely focus just as well on its ‘communicative success’, another factor Krashen has uncompromisingly rejected, as we saw (cf. Widdowson 1990: 21). In an afterthought unobtrusively placed in a fine-print footnote at the end of the book, he concedes he ‘does not mean that output practice should be avoided; some practice in producing language may help the student gain confidence, and in the case of writing may help the student develop an efficient composing process’ (93). But his ‘theory predicts that a programme emphasising only output will not be effective, even if the students’ goals are only writing or speaking, without comprehension’ (91). Of course, only a tiny handful of radically behaviourist programmes ever proposed to disconnect output from comprehension in order to emphasise pure ‘conditioning’; once more, Krashen unfairly overstates rival positions to make them obviously untenable.

Taken together, all these rhetorical exclusions make ordinary language learning seem both horribly difficult and rather irrelevant to what Krashen calls ‘language acquisition’. Apparently, we should just regale our ‘acquirers, no longer calling them “learners”’ and now viewing them as ‘humanoid receptacles in a maximum state of receptivity’ (Widdowson 1990: 21), with torrents of ‘comprehensible input’ and let the LAD do its job. We might even conclude that ‘materials, lesson plans, etc. are not necessary’; and Krashen calmly tells us that ‘theoretically, this is so’ (55) — a stupefying move if you don’t know that he’s building up to a sales pitch for his own favoured method. No one seems to notice here, and Krashen would be crazy to say so, that, by the same logic, his ‘theory’ also ‘predicts’ that we could fire teachers and replace them with automatic input-providing devices like radios and televisions, or with naive native speakers brought in off the streets of foreign cities, who would work for low wages and would require no expensive training in pedagogical methods he has declared to be all wrong anyway.

How could Krashen’s ‘theory’ account for the fact that many would-be acquirers don’t succeed so ‘automatically’, even when input is comprehensible? He cannot say that the LAD doesn’t work so well for some people; he won’t even allow that it works in different ways, as we saw. After all, his brash good news that it uniformly and automatically must work is one of the theory’s big selling points, because it predicts that we could totally eliminate failure. So he has to introduce some additional mechanisms that intervene between the LAD plus competence versus the outside world where real learners do real things. His proposals have varied in the past; in our discourse sample, he has three mechanisms, one on the input side and two on the output side. The imbalance of one against two is an interesting feature, as we shall see.

On the input side, Krashen postulates a filter that can prevent some comprehensible input from getting in to the LAD; incomprehensible input needs no filter because it couldn’t get in anyway. (This design does not allow the possibility that input might be noticed but not comprehended until a later time and then acquired, as I have observed in my own learning of languages, especially using written materials.) He builds in, on the ground floor of his theory, ‘the Affective Filter Hypothesis’ stating that ‘comprehensible input is necessary for acquisition, but it is not sufficient; the acquirer needs to be “open” to the input’ (3, i.a.) The ‘affective filter’ is a mental block that prevents acquirers from fully utilising comprehensible input for language acquisition. ‘When it is “up", the acquirer may understand what he hears and reads, but the input will not reach the LAD. This occurs when the acquirer is unmotivated, lacking in self-confidence, or anxious, when he is “on the defensive”, when he considers the language class to be a place where his weaknesses will be revealed. The filter is down when the acquirer is not concerned with the possibility of failure in language acquisition and when he considers himself to be a potential member of the group’.

I have postponed mentioning this main ‘hypothesis’ until a point where we can better see Krashen rhetorical motives for needing it. He cannot simply deny, in the face of practice, that language learners can and do attain widely varying degrees of success and failure; yet Chomsky’s idealised theory making uniform ‘competence’ be the ‘automatic output’ of the ‘LAD’ predicts they should all do much the same, and this ‘uniformity’ is taken over by Krashen too, as I noted. Nor can Krashen allow for such prime factors as varying degrees of conscious attention or varying amounts of deliberate practice in giving output, because his self-centred rhetorical exclusions deny the relevance of all these factors in principle, as we have seen.

So he constructs hypotheses that contribute blocking factors that are not defined in terms of language and especially not in terms of Chomsky’s theory — they are plainly ‘outside the language organ’. They are performance factors, although they are not actually among the ones cited in Chomsky’s (1965: 3) Aspects model, namely: ‘memory limitations, distractions, shifts of attention and interest, and errors (random or characteristic)’. For Chomsky (1965: 4), such factors affect output in ‘natural speech’, producing ‘false starts, deviations from rules, changes of plan in mid-course, and so on’. But he did not, indeed dared not, imply that such factors materially affect input while ‘the child learning the language’ is ‘determining from the data of performance the underlying system of rules’ (1965: 4). Instead (as quoted in section 4), he stressed the robustness of the process whereby ‘children acquire first or second languages quite successfully’, even when ‘actual speech consists of fragments and deviant expressions’. Moreover (again as quoted above), Chomsky salvaged the ‘striking uniformity of the resulting grammar’ by declaring them ‘independent of intelligence, motivation, and emotional state’ (1965: 58) — the latter two being factors Krashen now invokes as major determiners of acquisition. In effect, Chomsky’s theory had saddled Krashen’s with a LAD that is just too robust to be seriously marketed among experienced practitioners. Since the robustness is a big selling point that cannot be sacrificed, Krashen can only adduce a set of exclusionary performance-based constraints and ‘filters’ to exonerate the LAD for not doing in practice what the theory says it should.

Empirical research on cognition has firmly established that both learning and performance are indeed significantly supported by motivation, self-confidence, and a sense of belonging to a group, and impeded by anxiety and defensiveness, feelings of weakness, and expectations of failure (cf. surveys and references in Beaugrande 1980, 1984, 1996). But the research takes all these factors to apply to human capacities at large, and does not give them the special job of filtering the ‘input’ to an independent ‘device’ or ‘language organ’ that would otherwise ‘automatically’ do ‘perfect’ work. The consensus is rather that processing resources are always limited; models which, like Chomsky’s, airily set aside factors like ‘memory limitations, distractions, shifts of attention’, are unrealistic and unproductive, and — a main point in my argument here — untestable.

Krashen either doesn’t see the danger or hopes we won’t, of admitting on principle that performance can crucially determine the development and quality of competence, seeing how critically his ‘theory’ hinges on the thesis that the usual performance tactics in language classrooms definitely cannot do so — only comprehensible input can, and it is determined by competence. If performance can, then ‘communicative success’ and ‘output practice in communicative situations’ can help ‘competence develop’ after all — if not directly, then indirectly by building self-confidence, lowering expectations of failure, and so forth. Such activities would exert not a filter effect, but a booster effect on both competence and performance.

But Krashen can’t allow an input booster that could justify classroom methods he has disdained. So, opposite to just one filter on the input side, he installs two mechanisms on the output side: a filter and a booster. With a mild show of ‘reluctance’ (44), he adds an ‘output filter hypothesis’ (64) but doesn't capitalise it or enshrine it among the official ‘five hypotheses’ of the theory (1-4). As he did when contemplating two ‘language acquisition devices’, one for native and one for non-native language, he coyly protests that ‘the prevailing philosophy of science encourages us to use a minimum number of theoretical constructs — the “simplest” theory to account for existing data is considered to be the one closest to reality’ (44), where we again have to wonder which ‘data’ and ‘reality’ can be meant. He invokes ‘largely non-experimental evidence’ he finds ‘nevertheless compelling’ (45).

 ‘The output filter’ is a handy ‘device that attempts to explain why second-language users do not always perform their competence’; it also help out by ‘adding another explanation for variation in performance’ (45, 64) despite the ‘striking uniformity’ inherited from Chomsky’s LAD. Again, the LAD is exonerated: ‘there has been real acquisition, but affective forces’ ‘prevent us from showing this competence’ — presumably ‘just those factors responsible for the input or affective filter’ (46) too. ‘The output filter’ ‘prevents acquired rules from being used’ (45).

The output booster does get a place of honour in the main theory under the title of ‘the Monitor Hypothesis’ ‘stating how acquisition and learning are used in production’ and how ‘the output of the acquired system’ is ‘changed’ ‘before we speak or write’  (1f). The Monitor is free to apply all the ‘learning’ and ‘conscious knowledge’  Krashen’s rhetoric of exclusion has strictly sealed off from ‘acquired competence’ and ‘subconscious knowledge’. This ‘Hypothesis’ offers yet another chance to recite his twin mantras: ‘claiming’ that ‘acquiring via comprehensible input’ and ‘learning via conscious rule teaching’ are two systems’ ‘used in very different ways’; and that ‘learning cannot become acquisition’ (22, 24). Krashen even contemplates enshrining his exclusions in an anatomical split: ‘Monitor use involves the left cerebral hemisphere’ and gives an ‘advantage’ for ‘listening’ with the ‘right ear’, whereas ‘dichotic’ ‘listening’ with both ears ‘taps only acquisition’, though he admits he is on ‘far shakier ground’ here than with his ‘non-interface’ position (64f). Indeed.

By now, his rhetorical motives should be transparent. Denying conscious control altogether would be far more audacious for giving output (which can be closely observed) than for getting input (which cannot), and would lose credibility among teachers. So Krashen admits it but wants it kept tightly contained. He at once imposes ‘two conditions’, ‘both difficult to meet’:10 ‘the performer must be consciously concerned about correctness; and he or she must know the rule’ (2); later in the book, he imposes three conditions: ‘focus on form, rule knowledge, and time’ (22). He also stipulates that ‘the gain in grammatical accuracy achieved by utilising the conscious Monitor is modest’ (21). He decries ‘Monitoring while performing’ as a ‘risk’ against which only ‘very advanced and linguistically sophisticated second-language performers’ can ‘succeed’ (22). Or, to criticise ‘drill and conscious attention to form’ once more, he warns that ‘using the conscious Monitor will only cover up the error temporarily’  because (cue the mantra) ‘learning does not become acquisition’ (48).

To further conjure ‘the dangers of Monitor-overuse’, he links his output booster with his ‘output filter’ by saying that both can ‘impede fluency’ (38, 64). In place of ‘Monitor use’, he recommends’ ‘lowering the output filter’ by ‘focusing off form and on meaning’ (64), though we might wonder if ‘focusing’ does not entail some conscious attention. Yet the split-up exclusionary design of his theory impels him to claim that the ‘operation of the output filter does not affect the Monitor’; and the claim causes problems for at least two reasons. First, an experienced teacher knows that ‘affective forces’ can easily hamper second language learners from watching their grammar when they speak (give ‘output’). Second, Krashen's counsel for ‘acquisition with optimal efficiency’ is to ‘temporarily forget’ you are using another language (101); and his leading piece of evidence for his ‘output filter’ was a case where ‘students’ were impelled to ‘forget’ because of ‘strong feelings  about the topic’ (45). So the same factors that ‘lower the filter’ would also affect the Monitor. Indeed, ‘lowering the filter’ and ‘raising the Monitor’ might be corresponding descriptions or explanations of the same operation if Krashen weren’t so determined to split them apart.

Trying to teach to an LAD so hemmed in with input and output gadgetry might seem a parlous venture. But it’s another good-news, bad-news situation quite like the one I diagnosed for Chomsky. Krashen too is eager to throw out the baby with the bath water and the bathtub, because he too has a new bathtub to sell, namely, a teaching method. ‘If the Input Hypothesis and Fundamental Principle are correct’, runs his forecast, ‘we may be facing a “period of adjustment”’ (57f). ‘If the theory is correct, it will find its way into general education, and language students will no longer expect learning to be the central component of their language course’ (58). The ‘Natural Approach’ he has developed along with Tracy Terrell and others bears a label which was already found in the writings of Henry Sweet (1899 [reprinted 1964]: 74f), who roundly rejected for ‘putting the adult in the position of an infant’ and ‘not allowing him to make use of his own special advantages’ of ‘generalisation and abstraction’, and ‘greater powers of concentration and methodical perseverance’. Now, the same label hints the method is meticulously designed to follow the ‘natural order’ in which Krashen claims rules must be ‘acquired’, as we have seen. But our discourse sample gives little evidence that he has specified (or knows how to) in any detail what that order might be, and gives many arguments against teaching rules at all. Without those specifications, the label is hardly more than a catchy brand-name like breakfast cereals have, and slyly implies that teachers who don’t use it might be performing unnatural acts in the classroom!

The design of the method is actually quite spontaneous and informal, and by no means so different from the ‘communicative approach’ as his exclusionary rhetoric implies. ‘It uses a semantic, or notional, syllabus, simply a series of topics that students will find interesting and that the teacher can discuss in a comprehensible way, supplemented by games, tasks, and other activities that provide comprehensible input’ (55). Not at all surprisingly, it ‘de-emphasises production’ (34) (i.e. ‘output’). We are assured it has ‘been compared to traditional approaches and demonstrated to be significantly and clearly better’ (13). After a ‘year’ in this approach, ‘an adult foreign-language student’ ‘will be able to converse comfortably with a native speaker (who adjusts his speaking a bit to the level of the student) on a variety of everyday topics; this is a great success when compared with the results of the usual second-language class’ (71). ‘Involvement in a topic of real interest has a chance of resulting in the students’ focusing on the message — a prerequisite’ ‘for real language acquisition’ (74). Moreover, ‘interesting materials’ ‘should be far easier to create’ than ‘bone-dry exercises’ with a ‘grammatical focus’; there is no need to ensure that particular grammatical rules or vocabulary are practised, and initial field testing need determine only whether the materials are interesting and comprehensible for the intended student audience’ (56).

Unlike most of Krashen’s claims, these are easily testable: either the approach works as advertised or it doesn’t. But its success rate is in no way a test of the sole validity of Krashen’s theory. It offers no proof of fostering an ‘acquisition’ that does not pass through ‘learning’; nor of bypassing ‘conscious monitoring’; nor of getting around the various ‘filters’ to and from the LAD. At most, students soon get the message that they’d better not do things in class that look like conscious learning or monitoring — and that message already came packaged with the audio-lingual method too!11 The success is far more likely due to creating a relaxed, non-threatening environment wherein students are not just passively regaled with ‘comprehensible input’, but are doing interesting activities instead of doing grammar drills, like discussing their own topics and playing ‘games’, where there is a deal of output practice, albeit of a more spontaneous kind.

I submit that interest is the real key, and that it is not a concept accounted for by Krashen’s theory, and still less by Chomsky’s. Interest is not a linguistic category, and Chomsky’s Aspects model (1965: 3) expressly excluded it as ‘irrelevant’ for ‘linguistic theory’, as I have shown. Moreover, since the LAD is exclusively a ‘language organ’ and (Chomsky says) does its work without having to rely on ‘real-life situations’, it would be incapable in principle of ranking input by interest. It can only rank by ‘comprehensibility’ assessed in terms of current ‘competence’ — a set of language rules. And a great deal of comprehensible input is not at all interesting, as almost any commercial language textbook can demonstrate.

Nor does ‘interest’ appear in Krashen's opening procession of ‘five hypotheses’ constituting ‘an overall theory of second-language acquisition’ (1-4). Instead, it blithely pops up halfway through the book: ‘according to the Input Hypothesis, we need simply present students with messages that are interesting and comprehensible’ (55, i.a.). An equally sly alteration happens to his flamboyantly heralded ‘fundamental principle in second-language acquisition’; at first it says that ‘people acquire second languages only if they obtain comprehensible input and if their affective filters are low enough to allow the input in’ (3). Later on, it stipulates that ‘a language teacher is first of all someone who can present messages of interest, help make them comprehensible, and put students at ease; in short, a communicator’ (57f) — even though we recall Krashen castigating the rival ‘communicative method’ for promulgating a ‘futile approach’ on the wrong assumption that ‘competence develops via output practice in communicative situations’ and by achieving ‘communicative success’ (35, 55). These are certainly not equivalent formulations of his ‘Hypothesis’ or ‘principle’; and the discrepancies among them are essential because, if taken literally and practised radically, Krashen's theory — and not the communicative method he accuses of doing so really does ‘create an impossible situation for teacher and student’. As teachers, we would have to continually assess comprehensibility without letting learners consciously monitor whether they are comprehending. We would have to assess the competence level of an inaccessible LAD hedged round with filters that can visibly misrepresent its ‘real acquisition’. We would have to actively discourage learners from learning explicit rules by rebuking them that they are just wasting their time if not indeed hurting themselves. And we would heavily emphasise input at the expense of output. Only a vehemently Unnatural Approach would even try to do all these things, and would certainly not have ‘great success when compared with the results of the usual second-language class’.

The most flagrant restrictions and pressures would result for teaching or learning vocabulary. Chomsky's Aspects model made no mention of ‘vocabulary’ (except as a technical term of the theory),12 and certainly not in his account of ‘acquisition’, which is all about ‘grammar’ and ‘syntax’. And his dim view of the ‘lexicon’ can be plainly seen when he said that ‘the lexical entries constitute the full set of irregularities of the language’ (1965:142). The same argument implies that the acquisition of vocabulary cannot in principle be entrusted to the LAD. If not even formal linguistics attributes the acquisition of vocabulary to ‘unconscious rules’, then the only alternative, obvious to anyone who has successfully learned a second language, is that vocabulary items must be learned in the conscious awareness that they are vocabulary items and not just some spontaneous mumbles or some mistaken pronunciations of other items. So the ‘comprehensibility’ of input does not automatically lead to the ‘acquisition’ of vocabulary items but presupposes the learning of the items, just as no LAD can supply an innate dictionary of words, even if it can supply some principles for organising input that consist of words.

Krashen's theory leads to the queasy conclusion that acquired rules are constantly interacting with learned items, in open defiance of his mantra, cited repeatedly here, that the two domains have ‘no interface’. And in fact, Krashen does associate the use of ‘second-language vocabulary’ with the use of the ‘Monitor’ as ‘a way of outperforming one's competence’ and ‘sounding far more advanced than the users really are’ (13, 26); and he even describes ‘teaching vocabulary’ as ‘relying on deception’ (28). But unless vocabulary has been learned, there would simply be no comprehensible input for the LAD to do its unconscious handiwork. You may forget the experience of having learned the items as you become more fluent, though some cognitive theories suggest that the contexts of learning persist as robust memory traces, e.g. the ‘encoding specificity principle’ of Tulving and Thomson (1973). But somehow, consciously learned items must get into unconscious competence, most probably by a gradual lowering of the consciousness threshold. The plain conclusion is not so much that ‘learning becomes acquisition’ but that learning is the front end of acquisition — just as practice is the front end of theory, at least in real life if not in linguistics. When theory tries to throw out the front end of practice or to take over its role — as have the theories of both Chomsky and Krashen in their self-centred exclusionary campaigns — they can only lead to practices that are either inconsistent with the theory or else downright impossible.

So to the degree that the Natural Approach does succeed, it is not an application, let alone a test, of Krashen’s split-up theory and provides no ‘support for Chomsky’s position’. Instead, the whole ‘theory’ has been a rhetorically driven by promotional campaign contrived to do in applied linguistics just what Chomsky’s theory was contrived to do in theoretical linguistics, as shown in section 4: to justify for the theorist a leading role in theorising and to make everyone else feel pressured to cite him and adopt his terminology. Chomsky has evidently been much more successful. After all, his theory was expressly crafted to be untestable and irrefutable; it legitimised large savings of labour for linguists; and it brought no obligation to produce any practical results such as a teaching methods. Krashen’s job is much harder, since he is resolved to exploit Chomsky’s theory while producing a method. Doing both at once is so hard that he has split his job into two jobs and hopes practitioners won’t notice that his theorising doesn’t at all enforce his teaching method nor make him the chief architect of the ‘period of adjustment’ we would all agree we are ‘facing’. The zeal of Krashen the theoretician to totally exclude all other accounts and seal them off from his own puts an divisive strain on Krashen the practitioner.

Why should other practitioners flock to a ‘theory’ (or to public lectures about it) implying that most of what they have been doing is irrelevant to ‘language acquisition’ — in fact, as I see it (though Krashen of course doesn’t say so), implicitly providing a rationale for a mass firing of well-trained teachers and hiring naive ones or just buying television sets and VCRs. The chief attraction no doubt lies in the theory appearing both to explain why conventional second language teaching methods have been so disappointing, and to open up a brave new world of ‘language programmes being more productive and efficient for our students and easier and more pleasant for teachers’. (Notice that the ‘efficiency’ goes to the students, while the ‘ease’ and ‘pleasantness’ go to the teachers!) The ensuing euphoria has no doubt helped to paper over the extent, shown above, to which the theory does not underwrite the practice in any but the most general and evasive terms.

Not surprisingly, Krashen’s theory signally fails to meet the requirements advocated in the foregoing sections of this paper. Instead of specifying the relationships between native and second-language learning, Krashen simply equates the two as twin jobs for the LAD. His theory would strongly discourage the use of a meta-language in second-language learning and the training of basic techniques for describing and discussing language as language, on the grounds that it’s all too ‘conscious’ and fosters ‘monitoring’. He gives no guidance to teachers for appreciating the typical problems of learners arising naturally from the contact between the first and the second language, but remarks glibly, and a bit absurdly, that ‘pedagogy does not need to help the acquirer fight off the effects of the first language’ (Krashen and Terrell 1983: 41); he presumably expects the LAD and ‘universal grammar’ to overcome all such problems. He does not reflect upon how a teacher might better be able to act as representative of the community of speakers of the second-language; ‘comprehensible input’ could just as well come from anybody, a principle which, as I said, secretly implies we can largely dispense with teachers and with teacher training, too. And worst of all, his theory is silent about cultural outlooks on the language, social and regional contexts and varieties, and on ways for guiding how cultural contacts can meet the specific needs of second language learning. Like Chomsky’s ‘theory’ with its ‘completely homogeneous speech-community’, Krashen’s ‘theory’ projects language learning to be an unconscious process in a cultural vacuum. But surely learning about other cultures and interacting them are prime motivations for learning a second language?

6. back to the drawing board

I hope to have shown that the application of theories taken over from theoretical linguistics — what Widdowson (1980) calls ‘linguistics applied’ as compared with ‘applied linguistics’ — has, aside from phonology and phonetics for pronunciation, been problematic enterprise to be approached with cautious attention to terms and concepts and to their rhetorical contexts and implications. The theories of fieldwork linguistics are tailored to the situation of the specialised fieldworker, who is a language learner with special qualifications; such theories might be extended to developing methods that give second-language teachers and learners at least a modest proportion of the advantages fieldworkers enjoy (section 3). The theories of ‘homework’ linguistics of the Chomskyan type are tailored to the situation of fiercely competitive theoreticians who do not need to have learned any more languages than their native language; such theories frankly work at cross purposes with the development of either fieldwork methods or teaching methods (section 4). Fieldworkers would rightly scoff at the notion that their work follows some ‘natural order’ genetically programmed into their LAD; instead, they proceed according to the quality and quantity of the data as it happened to be collected, depending on which persons or groups in the community are disposed to be recorded. Pending much better evidence, teachers should also scoff at the notion that their learners follow some ‘natural order’ but that we needn’t follow it in our teaching.

So we are back to the question of how an applicable theory might be made to order for practitioners (section 1), and how we may cease to be, in Brumfit’s (1980: 160) words, ‘prisoners of our own categorisations’. Instead of describing a language in its full and finished state, as theoretical linguistics almost always has, such a theory needs to describe the degrees of approximation to the language that are appropriate and manageable for communication from the very earliest stages onward.

Here, the implications for practice seem to be somewhat radical and disorienting from the standpoint of conventional pedagogy. Such a theory could specify a succession of interlanguages that share features both of the native language and of the second language. These would differ from the ‘interlanguages’ as the term’ has usually been used (e.g. Selinker 1972) in having been expressly and strategically designed on a theoretical foundation rather than being spontaneously produced by individual learners or learner groups in real-life situations. The characteristics of the native language would be heavily represented in the early stages and gradually phased out over a suitable period of time in favour of second language features — not in a hypothetical ‘natural order’ but in a demonstrably workable order.

However unorthodox this approach may sound, it might offer some significant advantages:

(1) It resembles the way languages are normally acquired in authentic multilingual settings both by children or adults, much more than do conventional classroom methods. The strategies people naturally select when left on their own resources to do their best job are surely obvious candidates for renewed attention from a theoretical standpoint.

(2) All the learners could freely participate in a communication, rather than restricting their contributions to the small handful of occasions where they feel reasonably secure of the second language vocabulary, grammar, and pronunciation.

(3) The typical stress and anxiety expressed for instance in learner diaries would be considerably lessened because distinctions between ‘correct’ versus ‘incorrect’ usage would be made only to the degree that is appropriate at a particular stage. The criteria of native speaker ‘correctness‘ would only become significant in the more advanced stages.

We cannot predict on theoretical grounds alone how long the progression would take before we could expect learners to perform within a reasonable approximation of the performance of native speakers of the second language. But I would predict that at the end of that period of time, fluency and self-confidence would be on the average considerably higher than what we are obtaining now, and would be much more democratically distributed among the general population of learners and not just among the modest portion that are obtaining reasonable fluency with current methods.

To sum up: if the prospects of application were determined before the theories were built, we could expect a generation of theories that share the following features:

(1) The theories would be not just descriptive for current, ordinary practices but also instrumental in the development of more powerful practices.

(2) The theories would give higher priority to criteria of effective application than to idealised standards such as theoretical formality, rigour, or compactness.

(3) They would not be theories of language separated off by itself but theories of language as it is organised partly by its own criteria and partly by its use in organising people’s models of world and society.

(4) The role of the theoretician and the practitioner would be explicitly accounted for within the theories, stipulating under what conditions a theory or practice is related to a given language or to language as general conception, and how  the theoretician and the practitioner can claim to represent a language community or its knowledge of the language.

(5) The cycles of theory and practice would dialectically feed back to and refine each other as the theory becomes steadily more able to specify their respective roles, and as the practice brings steadily more evidence to bear on the requirements for an applicable theory. Once theory and practice each has the job of writing and rewriting a script for the other, their interaction would finally take on a systematic continuity often lacking in the past.

(6) The theories would be ‘driven’ by large corpuses of authentic data, such as the International Corpus of Learner’s English (ICLE) being assembled for eight language groups and banked at the University of Louvain. Such data would help us to build a series of models of the language at various stages and in various settings to take into account the age of the learners, their social background, their native language, their culture, and the environment in which acquisition or learning occurs.

(7) The theories would also be ‘driven’ by corpuses of case studies of people learning languages in various situations such as fieldwork linguists, language teachers in training, learners in language classrooms, and work trainees learning new special-purpose languages on the job.

(8) The theories would seek to specify the correlations among various modalities of spoken and written language, which have traditionally been crudely divided into ‘four skills’. New theories will doubtless show that ‘speaking’, ‘writing’, and so on are not unified blocks of activity but subtly nuanced complexes of activities keenly sensitive to the respective practical uses they serve.

(9) The theories would draw together observational studies of conventional classrooms and experimental classrooms, attempting to reason back from the more successful practices toward theories that can account for them.

(10) The theories would indicate the position of second-language learning within the overall curriculum, especially in respect to the native-language curriculum and to those subject areas in which the second language will later be used, e.g. when students in South-East Asia learn English in as a step toward careers in the computer industry.

At the start of this paper, I remarked that practice has usually preceded and determined theory, but theory has typically taken over the leading role and at times has been disconnected from practice. In this respect education and linguistics reflect the priorities of society. I can’t help wondering if the low importance some ‘mainstream’ linguistic theories assign to practice, and the problems of doing applied linguistics at cross-purposes with theoretical linguistics has not ominously encouraged the stingy commitment of resources by governments and institutions to language pedagogy, on the assumptions that the elites who need the language will learn it anyhow by studying abroad or hiring special tutors, and that the development of genuinely new theories and methods is not likely to result from increased investments.

At all events, theories which sharply raise the status of practice are one indispensable prerequisite to improving public and institutional attitudes. A generation of theories that can put language back into authentic cognitive and social contexts and can demonstrate the merits of multilingualism for the future well-being of our increasingly global society will also create an ambience in which theory and practice can finally be united as equals and dialectical partners in the enterprise of guiding the teaching and learning of languages.



1 This is a revised version of a paper prepared for the 1994 AAAL and TESOL Conventions in Long Beach but informally summarised there rather than read out. I am deeply indebted to Prof. Barbara Seidlhofer and to Prof. H.G. Widdowson for discussions of this topic and for carefully reading and assiduously commenting upon earlier drafts of the paper, and to three anonymous referees of this journal their helpful advice.

2 The term ‘mainstream linguistics’ is used here merely heuristically for conventional notions which were either stated in central works and frequently cited from these or else taken for granted; and which dominated the agendas in academies, universities, professional conferences, and so on (see my survey in Beaugrande 1991 for specific sources and quotes. Pennycook (1994: 25) observes: ‘there has clearly been a rejection of connections between language and its contexts in much of mainstream linguistics’.

3 For example: ‘it is often almost impossible for an Englishman to learn educated colloquial German in the county because all the Germans want to practice their English upon him’ (Sweet 1899 [reprinted 1964]: 76).

4 Insofar as the book directly equates ‘theory’ with Chomsky’s, the claim is a mere tautology, since there would be nothing else left to ‘take seriously’.

5 He presumably meant ‘language acquisition’ rather than ‘language learning’, but his phrase ‘undertake the task’ inappropriately suggests that the child makes a deliberate effort.

6 For compact citation, I shall omit the ‘1985’ in references to this one volume; and I also omit a reference when it is identical to the one just before it.

7 Widdowson (1990: 18) draws the same conclusion and points out that the ‘evidence’ actually bears on ‘consistency in accuracy’ and not on ‘internalisation’. Conversely, Gregg (1984: 97) argues that the natural order cannot be ‘explained’ via the LAD, but neither he nor Krashen explains how it could possibly operate without the LAD.

8 I am indebted to one of my reviewers for pointing out this possibility.

9 This prospect was in fact entertained by the Chomskyans for time as a promising way to get ‘semi-sentences’ or ‘partially well formed sentences’ into the theory, but Krashen doesn’t pursue it in the discourse I analysed.

10 Elsewhere, Krashen (1982: 18) says that ‘most people, even university students’, need ‘a real discrete-point grammar-type test to meet all three conditions for Monitor use’; if so, it could not be so grave a ‘risk’ as he alleges.

11 Widdowson (1990: 25) notes some interesting ‘similarities’ between Krashen’s ‘theory’ and behaviourism, including the ‘capacities for making mischief’. I would add that some key terms, such as ‘input’, ‘output’, and ‘affect’, were also made current by behaviourism, though in more precise and operational senses than Krashen’s.

12 As in: ‘the grammatical formatives and the category symbols’ ‘are selected from a fixed universal vocabulary’ (1965: 65f).



Apple, M. 1985. Education and Power. Boston: ARK.

Bailey, K. 1983. ‘Competitiveness and anxiety in second-language learning: Looking at and through the diary studies’, in H.W. Seliger and M.H. Long (eds.), Classroom Oriented Research in Second Language Acquisition. Rowley, MA: Newbury House, 67-102.

Bailey, K. and R. Ochsner. 1983. ‘A methodological review of diary studies’, in K. Bailey, M.H. Long, and S. Peck (eds.), Second Language Acquisition Studies. Rowley, MA: Newbury House, 188-198.

Baugh, J. 1983. Black Street Speech: Its History, Structure, and Survival. Austin: University of Texas Press.

Beaugrande, R. de. 1980. Text, Discourse, and Process. Norwood, N.J.: Ablex.

Beaugrande, R. de. 1984. Text Production. Norwood, N.J.: Ablex.

Beaugrande, R. de. 1985. Writing Step by Step. New York: Harcourt Brace Jovanovich.

Beaugrande, R. de. 1991. Linguistic Theory: The Discourse of Fundamental Works. London: Longman.

Beaugrande, R. de. 1992. ‘The heritage of functional sentence perspective from the standpoint of text linguistics’. Linguistica Pragiensa 34/1-2: 2-26 and 55-86.

Beaugrande, R. de. 1996. New Foundations for a Science of Text and Discourse. Norwood, N.J.: Ablex.

Beaugrande, R. de. In press. ‘On  history and historicity in modern linguistics: Formalism versus functionalism revisited’. Functions of Language.

Beaugrande, R. de. In review. ‘Theory versus practice reconsidered: Grammar versus communication in the teaching and learning of languages’. Submitted to Applied Linguistics.

Beaugrande, R. de and B. Seidlhofer. In review. ‘“Nativeness” and “naturalness” in the teaching and learning of foreign languages’. Submitted to International Journal of Applied Linguistics.

Brake, M. 1980. The Sociology of Youth Culture and Youth Subculture. London: Routledge and Kegan Paul.

Brumfit, C.J. 1980. ‘Being interdisciplinary’. Applied Linguistics 1: 158-64.

Chomsky, N. 1957. Syntactic Structures. The Hague: Mouton.

Chomsky, N. 1959. ‘Review of Verbal Behavior, by B.F. Skinner’. Language 35: 28-58.

Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge: M.I.T. Press.

Chomsky, N. 1975. Reflections on Language. New York: Pantheon Books.

Chomsky, N. 1980. Rules and Representations. New York: Columbia University Press.

Cohen, A. 1976. Educational Research in Classrooms and Schools. London: Harper and Row.

Dalton, C. and B. Seidlhofer. 1994. Pronunciation. Oxford: Oxford University Press.

Escribano, J.L. 1993. ‘On syntactic metatheory’. Atlantis 15/1: 229-67.

Giroux, H. 1992. Border Crossings: Cultural Workers and the Politics of Education. London: Routledge.

Gregg, K. 1984. ‘Krashen’s monitor and Occam’s razor’. Applied Linguistics 5: 79-l00.

Halliday, M.A.K. 1994a. An Introduction to Functional Linguistics, Second Edition. London: Arnold.

Halliday, M.A.K. 1994b. Language in a Changing World. Sydney: Australian Association of Applied Linguistics.

Harris, R. 1990. ‘On redefining linguistics’, in H. Davis and T. Taylor (eds.), Redefining Linguistics. London: Routledge, 18-52.

Hymes, D. (ed.). 1964. Language and Culture in Society. New York: Harper and Row.

Krashen, S. 1982. Principles and Practice in Second Language Acquisition. New York: Pergamon Press.

Krashen, S. 1985. The Input Hypothesis: Issues and Implications. London: Longmans.

Krashen, S.  and  T. Terrell. 1983. The Natural Approach. San Francisco: Alemany.

Lakoff, G. 1987. Women, Fire, and Dangerous Things. Chicago: University of Chicago Press.

Lave, J. and E. Wenger. 1990. Situated Learning: Legitimate Peripheral Participation. Cambridge: Cambridge University Press.

Newmeyer, F. 1980. Linguistic Theory in America. New York: Academic.

Pennycook, A. 1994. The Cultural Politics of English as an International Language. London: Longman.

Phillipson, R. 1992. Linguistic Imperialism. Oxford: Oxford University Press.

Phillipson, R., T. Skuttnab-Kangas, and M. Ranut (eds.). 1994. Linguistic Human Rights. Berlin: Mouton de Gruyter.

Pike, K.L. 1944. ‘Analysis of a Mixteco text’. International Journal of American Linguistics 10: 113-38.

Pike, K.L. 1967. Language in Relation to a Unified Theory of  the Structure of Human Behavior. The Hague: Mouton.

Rabinow, P. 1977. Reflections on Fieldwork in Morocco. Berkeley: University of California Press.

Saussure, F. de. 1916. Cours de linguistique générale. Lausanne: Payot.

Selinker, L. 1972. ‘Interlanguage’. IRAL 10: 219-231

Skinner, B.F. 1957. Verbal Behavior. New York: Appleton-Century-Crofts.

Sweet, H. 1899 [reprinted 1964]. The Practical Study of Language. Oxford: Oxford University Press.

Tulving, E. and D. Thomson. 1973. ‘Encoding specificity and retrieval processes in episodic memory’. Psychological Review 80: 352-73.

White, L. 1987. ‘Against comprehensible input: The input hypothesis and the development of second-language competence’. Applied Linguistics  8: 95-110.

Widdowson, H.G. 1980. ‘Models and fictions’. Applied Linguistics  1: 165-170.

Widdowson, H.G. 1990. Aspects of Language Teaching. Oxford: Oxford University Press.