Linguistics and the Critique of Domination 1
An extended critique of the East Anglia school of Critical Linguistics, represented in the two books Language and Control by Roger Fowler and others and Language and ideology by Gunther Kress and Roger Hodge
(Superscript numbersx refer to the endnotes, bracketed names refer to the bibliography)
In an essay on 'Illocution and Ideology' Keith Graham remarks that if 'ideological distortion plays an important part in perception of social relations ... it would be unthinkable that language itself, as the basic form of so much social interaction did not contribute to the process' (Graham 1981, p. 188), and he gives a footnote to my own previous work (Pateman 1975 and 1980a) as attempting to show how indeed language does contribute to such a process, where the emphasis is on forms of utterance and interactive sequence. Other researchers have also had the idea of looking for the reflection and reinforcement of ideologies, and for a contribution to the reproduction of domination, in the textures of discourse, including a group of four linguists and philosophers, all working originally at the University of East Anglia, two of whose works - Language and Control (Fowler, Hodge, Kress and Trew, 1979) andLanguage and Ideology (Kress and Hodge 1979) - I shall consider in detail here. The argument I develop is also implicitly critical of my own earlier work, and leads to some conclusions which come close to Graham's 'unthinkable': if 'language' is narrowly defined, it is difficult to show how it might contribute to 'ideological distortion'. This is the conclusion indirectly reached by Black and Coward in their review of Dale Spender's Man Made Language (Spender 1980), where they remark that, 'The theories of discourse or 'pragmatics' developed within modern linguistics have little to offer feminist analysis' (Black and Coward 1981, pp 78-79), since they seem unable to show up the distortions assumed to exist. This leads, I think, to a broader construal of 'language', but when that happens, it is all too easy to end up talking about something which is notlanguage at all but, rather - ideology!
However, before coming to the analysis of what I shall call for convenience 'the East Anglia theory' of domination through discourse in section 4, I must discuss two of the more general contexts of the works in question: first, the epistemological context (section 2) and second, the 'language and thought' context (section 3). I draw attention here to footnote 10 at the beginning of section 4 which explains why I think Tony Trew's contributions to Language and Control are largely immune to the criticisms I bring in section 4 against the approach I take to be common to his three co-authors. I abbreviate the title of Language and Control to LC, and of Language as Ideology to LI, and give the references to LC and LI rather than to Fowler et.al. 1979 and Kress and Hodge 1979.
1.2 Epistemological Foundations of Critical Linguistics
The final essay in LC is entitled 'Critical Linguistics', and it is as critical linguistics that the authors of LC and LI represent their practice in those two books. Critical linguistics offers 'resistance' to 'mystificatory tendencies in language', and because of the direction of determination from society to language, it is consequently 'a critique of the structures and goals of a society [contemporary Britain - TP] which has impregnated its language with social meanings many of which we regard as negative, dehumanising and restrictive in their effects' (LC, p. 196).
These sentiments and similar ones scattered through the two books, together with the occasional allusion or reference, serve to place the project of LC and LI very much in the tradition of Frankfurt School Critical Theory. That tradition is underpinned by an epistemology which seeks to establish the scientific character of criticism, and ground as non-paradoxical the concept of objective partisanship central to it. At least one answer, and not an implausible one, is consequently available against the objections of the positivistically-inclined who will read LC and LI as exercises in mere partisanship, not advancing our knowledge of language at all, but simply passing more or less arbitrary judgment upon selected aspects of it. In subsequent paragraphs I will sketch out that answer in the terms used by Jürgen Habermas in his 1965 Inaugural Lecture 'Knowledge and Human Interests' (Habermas 1972, pp 301-317), after briefly sketching the philosophical context of Habermas' work.
If it is possible to make any true or false statements about the world, then it is possible to make an infinite number of true and an infinite number of false statements. It is easy to show why this is so. Suppose I drop my pen on the floor, a single event in the world truly described by (A). But it is also truly described by (B). Further, since length is infinitely divisible, an infinite number of true statements like (A) and (B) can be generated. 2 The reader can easily use the example to yield a proof of the infinite number of false statements which can be made.
(A) At time t, T.P. dropped his pen on the floor.
(B) At time t, T.P. dropped his pen, which is less than twelve inches long, on the floor.
It follows that from the simple injunction to pursue truth no rational scientific practice is deducible. For someone could truly claim to be engaged in the pursuit of truth who spent their life producing statements like (A) and (B) - and they would be considerably more successful in their pursuit than most scientists!3
What more is required, then, to specify a rational scientific practice? Two different answers are provided by realism and positivism. The realist tells us to pursue truth in the form of identifying the finite number of causal mechanisms which produce the infinity of effects constituting the actual and experienced world, and stating the laws which govern the operation of those mechanisms. Unfortunately, this does not of itself cut down the search space of science any more than does the simple injunction to pursue truth. For since laws (which are theories) are always and necessarily underdetermined by the data (evidence) which supports them, there are an infinite number of candidate laws observationally or descriptively adequate to account for any finite set of data, and the problem of selecting explanatorily adequate laws remains. 4 The realist is now faced with the problem of how to constrain the search among candidate laws in such a way that the true one has some chance of being hit upon. At this point, at least one realist has opted for pre-established harmony between mind and the rest of nature, 'Unless man have a natural bent in accordance with nature's, he has no chance of understanding nature at all' (Peirce 1940, p. 156).
The Viconian argument that we can hope to understand only what we ourselves have produced serves the same purpose as Peirce's pre-established harmony in establishing the rationality of the practice of the cultural sciences. It is to similar ideas that Lévi-Strauss resorts in seeking to ground the plausibility of his logic of myths (Lévi-Strauss, 1964).
This is heady stuff for the positivistically inclined and they have their own more down-to-earth solution to the would-be scientist's dilemma about where to begin. Since the search space must be cut down in some way, let it be our interests (including our values) which regulate our pursuit of truth. Of course, once we have chosen a domain for study on the basis of our interests, it is rational from then on to make sure we respect all the Lockean injunctions against bias and partiality (Locke 1966), and any judgments we may pass on what we discover will still be extraneous to the matter in hand, and have their basis once again in our interests. It is impossible to avoid selectivity, but entirely possible to sustain objectivity.
For a positivist of this sort the enterprise of LC and LI will be construed as selected by the authors' subjective anti-capitalism, and its judgments on the data gathered as deriving from the same extraneous standpoint. It is the object of Habermas' reconstruction of the epistemological foundations of critical theory to show that this positivistic interpretation of interests as regulative and critique as subjective is fundamentally mistaken.
In a tradition of thought which dates back at least to Kant, Habermas challenges the assumption, common to positivism and some versions of realism (though not Roy Bhaskar's: Bhaskar 1979) that reality can be represented at the level of observation statements in a form which is essentially value-free and theory-free. In other words, Habermas stresses the theory-laden character of all evidence, though where Kant did so in terms of an a-historical theory of the constitutive activity of the human mind, Habermas presents an anthropological and quasi-historical account of the knowledge - constitutive character of human practical activity, including scientific activity. Kant drew as a consequence of his theory that reality as the 'thing in itself' remained necessarily unknowable to humans constituted as they are. Habermas, I think, wishes to resist drawing this conclusion, though his theory may force him to it - a question I must leave aside here.
For Habermas, human interests are not merely regulative, but constitutive of the pursuit of knowledge, through the practical activities in which they are realised. Specifically, they give rise to - or, more accurately, transcendentally constitute the conditions of possibility of - three kinds of scientific activity, evidence, proof and refutation. First, in the empirical-analytic sciences, theories, 'disclose reality subject to the constitutive interest in the possible securing and expansion, through information, of feedback-monitored action. This is the cognitive interest in technical control over objectified processes' (Habermas 1972, p.309). This description would also be applicable to work in artificial intelligence (AI), including computational linguistics, where it is indeed legitimate to accept all negative feedback (bugs etc.) as straightforwardly falsifying à la Poppero (see Lakatos 1970). In both AI and the empirical-analytic sciences the 'disclosure of reality' can come to be seen as essentially irrelevant; all that matters is that the theory or program works, as is argued in various kinds of pragmatism and operationalism. (For a sophisticated and attractive version of this idea, see Sloman 1978, ch.2. He argues that science is concerned centrally with the discovery of sufficiency conditions, not necessary conditions.)
Second, in what Habermas calls the 'historical-hermeneutic' sciences, inquiry, 'discloses reality subject to a constitutive interest in the preservation and expansion of the intersubjectivity of possible action - orienting mutual understanding' (Habermas 1972, p.310). Quine's famous portrayal of the activity of radical translation in Word and Object (Quine 1960, ch.2) is exactly an account of hermeneutic inquiry, presuppositions (or postulates) of the rationality of which is that 'the attainment of ... consensus' (Habermas 1972, p.310) is desired by participants and is possible for them - though Quine argues that it is in principle impossible for them to be certain that they have attained a genuine or true consensus. In Habermas' terms, the 'linguistics of languages' would be a central example of inquiry constituted by an interest in mutual understanding, for in it languages are described in ways which make it possible for non-native speakers to understand native speakers. If the belief in the possibility of understanding and desire for it are the faith and hope of this activity, charity enters as a (methodological) principle of interpretation the precise status of which is the subject of much current debate (see, for instance, Putnam 1978, Part 1, and references there).
Third, in the critically-oriented sciences, an emancipatory cognitive interest constitutes the effort to, 'determine when theoretical statements grasp invariant regularities of social action as such and when they express ideologically frozen relations of dependence that can in principle be transformed', and, more generally, to 'release the subject from dependence on hypostasized powers' (Habermas 1972, p.310). Freudian psychoanalysis is Habermas' paradigm case of such a critical science, at this stage in the development of his thought. It relies for its practical success on the analysand's own interest in securing freedom from domination through (though not only through) enlightenment. Habermas traces such an emancipatory cognitive interest to the fact that we are language-using creatures, arguing that in acquiring a language, 'autonomy and responsibility are posited for us', (ibid., p.314) however much they are denied in reality. Though in later work, Habermas has greatly developed this last idea, and modified his 1965 account of knowledge constitutive human interests (see McCarthy 1977 for discussion) I have, I think, said enough to locate the project of LC and LI.
For in LC and LI the authors challenge the hypostasization of spoken and written natural languages as neutral, transparent, reflective media, always and everywhere indifferently usable for purposes either of enlightenment or domination, and offer their work 'as a contribution to the unveiling of linguistic practices which are instruments in social inequality and the concealment of truth' (LC, p.2, my italic). The 'Critical linguistics' of LC and LI offers resistance to 'mystificatory tendencies' in language (see section 3 below); it 'is not resistance to language itself, nor to individual users of language, but to the social processes which make language work in communication as it does.' (LC, p.196)
These sentiments place LC and LI squarely in the tradition of critical theory. Here, the autonomy and responsibility universally posited for us as language-users - as speaking subjects who are the responsible authors of our utterances - are the basis from which a critique of actual languages (exclusively English in LC and LI) and their use is mounted. That this can be done in the language criticised does not constitute a paradox, since there is no logical obstacle to a language being it own (no doubt imperfect) metalanguage. But that fact must immediately eliminate as serious substantive theories of critical linguistics some of the stronger neo-Whorfian claims advanced in LC and LI. It is these I turn now to consider.
1.3 Functionalism, Whorfianism and Materialism
The orientation of LC and LI is functionalist in the sense of 'a theory of language which proposes that the structures of language have developed in response to the communicative needs that language is called upon to serve rather than a linguistic model which assumes that structures are natural, universal properties of the human mind and so unaffected by social function' (LC, p.3). This functionalism takes its inspiration from the work of Michael Halliday, and in the above quotation is counterposed to the nativism of Chomsky. However, Chomsky is not so foolish as to believe that all the grammatical structures of all natural languages can be wholly explained by natural, universal properties of the human mind, since obviously different languages have different as well as common grammatical properties, and not all their properties need have the same ontological status. Equally, to claim that social function is the only source of linguistic structure would involve denying that the human mind and the natural conditions of human existence impose any limits on the languages which are possible for humans, a position which I take to be obviously implausible. In consequence, the only tenable position on the relation between communicative function and linguistic structure is surely that stated by the Prague School functionalist linguist, Peter Sgall, who writes that, 'it may be assumed without any contradiction that linguistic structure is determined by innate properties in some aspects, others being conditioned functionally' (Sgall 1977, p.18). Of course, there is still plenty of room for disagreement on the relative weight to be accorded to each causal mechanism affecting linguistic structure, and Sgall would weight non-innate sources more heavily than Chomsky does. What is important in the present connection, however, is simply to avoid an oversocialized conception of human beings, the conditions of their existence, and their languages, and to note the difference which any assumptions about innateness make to any theory.
Whatever their scope, all 'functional' (in the sense ofpurposive) explanations of linguistic structures are going to require concepts like 'communicative needs' to do the explanatory work, that is to identify causes or causal mechanisms. But what are 'communicative needs'? The expression is neither general enough or appropriate in work devoted to showing that functional pressures on languages are not necessarily democratic or egalitarian in character. In place of 'communicative needs' we could speak, like Lewis 1969, of the interests people have in controlling the beliefs and actions of others, for the pursuit of which interests language serves as an instrument. These interests will be causally active in human interaction, but need not correspond in a one-to-one fashion with subjectively-perceived wants; they will arise from both idiosyncratic and structurally-conditioned experience, desires, beliefs, values, etc. It will be the structurally conditioned independent variables which presumably will be most important in any functional explanation of linguistic structures. For example, in a two-class society of rulers and ruled, the rulers may have a structurally-conditioned, standing interest in limiting flows of information from themselves to the ruled, and possibly a belief that this ought to be done as unobtrusively as possible. If the theoretical claims of LC and LI are true6 such people - whether conscious of it or not - would have a clear reason to develop or favour a grammatical passive with compulsory agent deletion (as in Latvian: Comrie 1977; cited in Brown and Levinson 1978, p.275).5
In this example, the explanatory functionalism has been married to a substantive materialism: human interests, desires, beliefs, etc. themselves arise in and are conditioned by determinate material circumstances, of which economic and political circumstances form a historically and geographically variable sub-set. (For more universal aspects of material circumstances, see Timpanaro 1975.) This marriage accords with tendencies LC and LI, in which specific debts to the materialisms of both Marx and Freud are acknowledged. But it cannot be reconciled with the Whorfian tendencies of their arguments.
For if Whorfianism is any distinctive theory at all, it must be a theory about the conditioning of people's experience, desires, beliefs, values etc. by the language they speak (Whorf 1956). The direction of causation is exactly the reverse of that proposed in any materialism. In addition, it is not a functional (purposive) explanation at all, and can only be made so at the price of absurdity - the attribution of purposes to a language, rather than to its speakers.
Of course, if Hopi speakers believe in the metaphysics of the Hopi language this helps ('contributes to the survival of') Hopi just as, à la Durkheim, criminals help ('contribute to the survival of') the police. This may encourage the idea that the contradiction I have sought to generate can be dissolved by saying that, of course, circumstances and language 'interact'. But this doesn't work. For either it means that Hopi beliefs and the Hopi language ( or criminals and police) form a closed system - a symbiotic relationship not acted upon by external causes - which is patently false in both cases; or else it allows that either Hopi beliefs and criminals or the Hopi language and police or both are acted upon by external causes, most obviously by material causes, such as natural events and economic circumstances. But to allow this second position is to reinstate some kind of materialism incompatible with any non-trivial Whorfianism: the experience of Nature can alter Hopi beliefs in a way which overrides Hopi, and economic circumstances produces criminals who are not creatures of prior police practices.
Since there are independent grounds for rejecting Whorfiansim in the version I am considering (see notably the work of Rosch 1977a,b), and independent grounds for accepting purposive and materialist explanations (see Pateman 1987, chapter 2), I propose that Whorfianism be abandoned, despite its attractiveness as portraying yet another aspect of the domination of individuals by the institutions they have created but are unable to transcend. However, everything of value that Whorfianism seeks to do can be done by a functionalism which avoids the fallacy of interpreting a functional explanation of something as a claim that it is eufunctional for all those affected by it (eufunctionalism).8 Functionalism in the sense of purposive explanation is perfectly compatible with a sectoral or conflict theory. In the case of languages, if their structures encode sectoral interests, and Whorfianism is false, we should expect to find evidence for the partisanship of linguistic structures in actual conflicts over language between those whose interests are served, and those whose interests are not served, and this is exactly what we do find, at the everyday conversational level of metalinguistic enquiry and critique and at the theoretical level represented most recently by work like LC, LI and Spender 1980. Volosinov's concept of 'multiaccentuality' (see Volosinov 1973) is a theorisation of such situations.
What cannot be salvaged are the excesses of Whorfianism: 'only what has a name can be shared' (LI, p.5; compare Spender 1980 on 'Experience without a name', pp.182-190), which ignores the vitally important possibilities of metaphor and paraphrase (though see LI, p.63); 'the world is grasped through language' (LI, p.9), false if it means that the world is only grasped through language, true but less interesting if it means that language is one of the ways in which the world is grasped; 'The grammar of a language is its theory of reality' (LI, p.7), but even if it is that, it is only one among the theories of reality held by humans, theories which obviously must have sources in experience outside of language if humans are to survive in their changing environment.9
1.4. Sentences, Utterances and Reality
Even if the origin of linguistic structures could be completely explained by reference to causes of the 'communicative needs' variety, the use of a given structure on a particular occasion would not necessarily arise from the same causes as did the structure itself. Conversely, the causal origins of a structure cannot be inferred from the causes of its use on a particular occasion, though the causes of its continued reproduction are the causes which occasion its use. Inferences from past to present and present to past are blocked in the case of natural languages because linguistic structures (forms) do not possess the causal power to determine their own use in utterances (functions). If psychologically-real linguistic competence is at least the power of speakers to generate (and hearers to understand) novel system-sentences, in virtue of possession of the grammar of a language, communicative competence is at least the power of speakers to generate (and hearers to understand) novel uses for system-sentences, so that utterance tokens of the same sentence type can be used in quite different ways. (See Lyons 1977, ch.1 for an exposition of type/token and related distinctions.)
The indispensable distinctions between sentence types and utterance tokens have been unknown or blurred in traditional rhetoric and literary stylistics. Of the former, Sperber writes of, 'La confusion presque générale dans la littérature rhétorique entre phrase et énoncé' (Sperber 1975b; pagination missing on my copy). Pratt 1977 provides a critique of the formalism which characterises the latter. But some years ago now the distinctions were effectively denied in a version of generative semantics associated with J.R. Ross (see Ross 1970), whose kind of theory is indispensable to the critical enterprise of LC and LI, exception made of the important contributions in LC by Tony Trew.10 It seems to me that Rossianism is simply a version of a kind of formalism or structuralism which places all meaning in a context-free language. But unfortunately for LC and LI it is an untenable doctrine, both for local reasons and because of the weakness it has in common with formalism/structuralism, as will become clear in what follows, if by a roundabout route.
Here is a quotation from a linguist, Geoffrey Sampson, poles apart theoretically and politically from the authors of LC and LI, but expressing the kind of sentiment which is the pre-theoretical starting point for their critical work:
The passive construction, in sentences such as Adoption of the proposal is felt to be inadvisable, is beloved by bureaucrats aiming to disclaim responsibility for their decisions'. (Sampson 1980, p.105)
If we accept the claim that bureaucrats do love the passive construction, and Sampson's account of what their purpose (or one of their purposes) in using it is, and ask why they love that particular construction, the obvious answer is that in English the passive construction allows 'optional deletion' of agents, or - more neutrally (note 11) - is fully grammatical without specification of agent. (For further discussion of 'passive motivation' see Stanley 1975). Thus (1) is grammatical as a declarative whereas (2) is only grammatical as an imperative, and (3) not at all grammatical:
(1)Adoption of the proposal is felt to be inadvisable.
(2)Feel the adoption of the proposal to be inadvisable.
(3)*Feels the adoption of the proposal to be inadvisable.
Now the theory developed in LC and LI holds that (1) is the (surface-structure) product of the application of (at least) a passive 'transformation' and deletion 'transformation' to an underlying semantically organised deep structure (or 'basic model' in their terminology) which has an 'actional' (in fact, 'transactive' - see LI, p.8) form and specified agent. This specified agent has been 'got rid of' in the derivation of (1), using an optional agent deletion 'transformation' permitted after passivization, and it seems to be this especially which leads to talk of 'The mystification inherent in the passive transformation' (LI, p.33; at LC, p.208 'deletion of participants' is presented as a consequence of the passive 'transformation', but in English it is clearly not a grammatical consequence.12)
Assume this account to be true (see footnote 11 for doubts about its truth). Nonetheless, the basic model from which (1) is derived will be very different depending on whether we treat (1) as a system-sentence or an utterance token of that sentence. If (1) is a system-sentence, its underlying semantically-interpreted form cannot be more specific as to agent than I have indicated in (4) (where I use English without underlining as a notation to represent one possible underlying form), though it could be less:
(4) There is at least one person (A) such that that person reports to at least one person (B) [where (B) may be identical to (A); a proper sub-set of (A); etc.] that at least one person (C) [where (C) may be identical to (A), (B); a proper sub-set of (A), (B); etc.] feels adoption of the proposal to be inadvisable.
To convert (4) to an English surface-structure, underline it. The result is, according to the letter of LC and LI, a relatively unmystified sentence, at least up to 'feels adoption of the proposal to be inadvisable'. More seriously, the point to be made is that as a surface-structure system-sentence (1) could only be derived from something no richer in content than (4), where (4) does no more than make explicit our semantic knowledge that feelings about proposals can be attributed without category mistake only to people - and even that is to take a broad interpretation of semantic knowledge.13
As a concrete utterance, and assuming the psychological reality of passivization and deletion in the generation of the sentence-token from basic models or underlying forms modelled on those of J.R. Ross, the derivation of (1) would look quite different. Suppose Sampson's account of the motivation for (1) on some occasion of its utterance was correct. Then according to LC and LI, (1) would derive from forms including something like those represented in (5), (6), (7) and (8), where 'I' is the bureaucrat-author of (1):
(5) Adoption of the proposal is felt by me to be inadvisable.
(6) I feel that adoption of the proposal is inadvisable.
(7) I tell you that I feel that adoption of the proposal is inadvisable.14
(8) I tell you that I feel that adoption of the proposal is inadvisable and I tell you that I have not adopted it and I dare you that you challenge my decision.15
To begin with, convert (5) and (6) to English sentences by underlining them, and then consider their use in place of (1) in the same context of utterance as we have imagined for that sentence. There is a major difference in the informativeness of (1) and (5), (1) telling hearers or readers less because of the specified agent deletion which is not intra-linguistically recoverable (the 'transformation' is not reversible). But this is also to say that (1) and (5) are not synonymous. They do not have the same truth-conditions. In contrast, though (6) is an active rather than a passive structure, as an utterance it would be no more and no less informative than (5). In transformational terms, the process by which (5) is derived from (6) is reversible. The truth conditions of (5) and (6) are the same, and in that sense (5) and (6) are synonymous. (5) and (6) differ stylistically and this would no doubt affect their total meaning as utterances in a specified context. In the case we are considering, whether (5) or (6) is used would actually appear to make very little difference, whereas the difference between (1) and (5) is considerable.
However, even if (1) derives from (5) by a psychologically real linguistic process in the speaker's head, which is not linguistically recoverable (reversible) by the hearer, it does not follow that an addressee cannot in context recover the information contained in (5) from the utterance of (1). Indeed, we make such 'recoveries' all the time, and Sampson does it in the quotation with which I began. This is because real-time interpretation of utterances need not be a mirror-image of the linguistic processes by which LC and LI claim they are produced, even if they are produced in the manner claimed. In interpretation, we draw upon our knowledge of the world, including specifically the speaker, past utterances of the speaker, context of utterance, and so on. This is by now a commonplace. It is the subject of several articles by Straight (e.g. Straight 1976. It is discussed explicitly in relation to truncated passives by Stanley and Wolfe (Robbins) 1977 and Stanley and Robbins 1977. It is difficult to see how it could be otherwise. It is acknowledged by Chomsky (1980, esp. ch.3. And it is inconsistent with the claim that, 'The interpretation of utterances entails reconstructing their derivation' (LI, p.34; this claim is said to be one made by 'generally accepted [transformational] theory'). The claim is only intelligible as the effect of a conflation of utterances with system-sentences (the latter have no real-world context), and is part of a systematic failure in LC and LI to appreciate the role of non-linguistic processes in the interpretation of utterances (see, for example, the discussion of a children's story in LI, pp.47-50, and compare with Sacks 1974 or Charniak 1976 or Schank and Abelson 1977). In short, it is partly because of what we know about bureaucrats, not English, that we can get from (1) to (5), (6), (7) or (8), and knowledge of English alone couldnever get us from (1) to (5), (6), (7), or (8). Understanding bureaucratese is like understanding indirect speech acts, and in both cases if we have the right extra-linguistic knowledge, we can achieve such understanding.
What is really wrong with (1) as an utterance, relative to (5) or (6), is its uninformativeness, but uninformativeness is not a linguistic fact or effect of purely linguistic facts at all. An utterance changes a context (Levinson 1980 and references there), and we usually have well-defined expectations about how any utterance should change the context, if it is to satisfy the validity claims its utterance raises. These expectations are the subject of all the literature deriving from Grice 1975. Undoubtedly, people can and do violate our expectations and thereby mislead us, 16 and some people (e.g. bureaucrats) may try to do so quite regularly. But is is only relative to a context of expectations that (1) can be rationally criticised as less informative than we require, by application, for example, of a Gricean maxim of quantity. If this were not the case, and more information was always better than less, the best utterance would be that which inscribed the history of the world in its surface structure. But just as utterances can mislead us by being too little informative, so they can mislead by being too much informative.17 The theoretical apparatus of LC and LI cannot handle such widely-recognised and discussed facts.
It is now possible to proceed to consideration of (7) and (8). (7) applies to the derivation of an utterance the substance of the performative analysis of sentences, developed by J.R. Ross and others, in which it is claimed that every declarative sentence of English must contain in the deep (semantic) structure from which it is derived, 'a higher subject 'I', ... an indirect object 'you', and ... some performative verb, possibly abstract [see note 14 - TP], as the main verb of the highest clause' (Searle 1975a, p.28; see also Gazdar 1979, pp 18-19 for precise formulation of 'the strongest possible performative hypothesis' in the form of eight sub-claims). If you'll swallow this claim you'll swallow anything, including the claim that (1) is derived from something which (8) represents, and the treatment of indirect speech acts in LC which I shall consider a few paragraphs further on. Hence it is worth pausing to consider the performative analysis of sentences and utterances.
The motivation for the performative analysis, as for generative semantics more generally, is to account for apparently grammatical facts by rules of grammar. If (13) and (14) are synonymous sentences, then the grammar which generates them ought to account for their synonymy. This can be done by arguing that (14) derives from (13) by meaning - preserving transformational deletions and other minor adjustments:
(13) I request you to go home
(14) Go home
Such an approach very rapidly runs into insuperable difficulties (catalogued in Gazdar 1979, ch.2; pp 23-25 for the question of meaning preservation), in the present case that (14) as a system-sentence could equally derive from (15), which is why abstract performative verbs, whatever they are, have to be brought in.
(15) I order you to go home.
Considered as utterances, (13) and (14) are intuitively related - we may imagine that anything which could be done by a surface token of (13) could be done by a token of (14), and vice versa.18 The performative analysis is also an attempt to explain this. But it fails (apart from Gazdar 1979, see Gazdar 1976; Lyons 1977, ch.16.5; Searle 1975a). And there is a theory which explains it successfully, the pragmatic theory according to which the apparent synonymy of (13) and (14) is due to extra-linguistic factors only present, of course, when (13) and (14) are used. (14) when used has an author and an addressee, and a force, all usually recognizable in the context of use. We represent the author, addressee and force in any description of the sentences used, but that is quite different from attempting to represent those descriptions in the grammar. LC and LI confuse these two levels. The result is the assimilation in to 'language' of the world in which it is used, a slide which vitiates the claim to be criticising ideology in language.
While the performative analysis of sentences and utterances may not look prima facie absurd, any claim that (1) might be derived (in a psychologically real process) from a basic model or underlying form represented by (8) is surely ludicrous. Is there any evidence that in LC and LI the authors are committed to anything like such a claim? The evidence I adduce is, first, that they impose no constraints on the 'transformations' by which they derive surface structures from basic models, so if a deletion 'transformation' can remove the speaker between (5) and (1), it could quite well remove all the material in the second and third conjuncts of (8). In other words, the 'transformational' arrow ' = > ' they use appears in a theory no more specific than an unrestricted rewriting system, and this is psychologically implausible. Second, and more important, there is evidence that the authors slide between using an undefined notation to represent underlying forms (basic models) in speakers' heads, and using it for rich descriptions of aspects of reality selected by the authors. This leads them to put into underlying forms much more than is possibly warranted, an error which is simply a magnified version of Ross's, and to which the criticism at the end of the previous paragraph is equally applicable.
I offer two examples to sustain these criticisms. Consider first the following quotation:
We start by looking at a noun which may or may not be in a dictionary of the English language, 19 that is, one whose status as a stable word is somewhat doubtful. Here it is in the sentence in which it occurs in the text [under analysis - TP]:
Picketing .... curtailed coal deliveries.
If we asked speakers of English what the meaning of picketing was, they would probably explain it by describing the kinds of things involved: strikers, the action, a factory, or, in this case, a coal-depot. The noun is a contraction of a significant kind. The single word necessarily implies a particular kind of actor and a particular object of action. We might represent the process in this way:
strikers picket a factory => picketing
Because we can interpret the nominal as being 'derived from' the full sentence, the deletion of actor and affected is not a complete elimination ...
An activity which was initiated and performed by the miners, in a specific place and time, now seems to have autonomous existence, and can appear as the actor in a new construction,
Picketing curtailed coal deliveries
The affected entity in that sentence is yet another nominalization: again we would not meet the noun coal deliveries in a dictionary of English. Its source seems to be a sentence of the form,
Someone [rail drivers] delivers coal
(LI, pp 20-21)
In the first part of this quotation a representation of a state of affairs in the world which it is supposed would be used to explain ostensively the meaning of the word 'picketing' becomes the underlying form of picketingwhich is '"derived from"' it by means of a non-truth preserving transformation. (But why the hedging scare quotes? Either it is derived or it isn't. And the theory says it is.) It is not truth preserving because picketingcan be truly predicated of the actions of people who are not strikers and who do not picket a factory, so the derivation is not reversible. In the second part, it is the reality that rail drivers deliver coal that is the source ofcoal deliveries, not 'a sentence of the form, Someone [rail drivers] delivers coal', which is rather a representation of that reality. To suppose otherwise opens the way to the claim that coal deliveries has as its source a sentence of the form of (16):
(15) Someone [rail drivers [Joe Smith, Fred Jones, George Herbert [driving diesel locomotives numbers 007, 008, 009]]] delivers coal.
- and so on ad infinitum.
There is nothing in LI to block (16), which I take to be areductio ad absurdum of the argument advanced in the passage quoted.
What would be required to block (16)? Or, in other words, what would be required to limit the amount of information contained in a basic model or underlying form in a psychologically plausible way, and which would avoid the suspicion that what goes into the basic model depends not on psychological reality but the purposes of the critical linguist? It seems to me that what would be required is some frame or schematatheory, of the kind developed in artificial intelligence (MInsky 1977; Schank and Abelson 1977), which would indicate how reality is perceived and conceived from a point of view and at a level of descriptive richness dependent upon the frames or schemata activated in the subject. But what frames or schemata are mobilised on any given occasion will depend on (i) the subject's stock of frames, and this will be variable; and (ii) on the subject's interests, which define what is relevant on some occasion. Though this may not seem at all incompatible with LC and LI, in fact it removes the notion of a privileged kind of basic linguistic model or underlying form on which LC and LI rely for their critical purposes. Though the subject's set of frames and selection from them on some occasions can be explained, criticism of the set and the selection now becomes external rather than internal. It is true that internal criticism of the processes by which the frame-defined cognition is converted to a linguistic utterance remains possible, in terms of the operation of self-censorship, for example. Along these lines, (16) could represent not a sentence, but a cognition of a particular subject (a spy, for instance), which would differ from the cognitions other subjects had of the same reality (like those of Kress and Hodge in the quotation), and which might or might not surface as an utterance.
In other words, ideological distortion can be 'base-generated' in the sense that the ideological distortion is present in the schemata or frame activated to conceive a situation, and which distortion is simply 'passed on' to the linguistic expression of the conception. In this way, the subject can be entirely unaware of the ideological character of what he or she is saying. The bureaucrat who uses passives uses them, on this view, because that is the way he or she sees things from the outset, not because the bureaucrat is a rewriting machine (cf. Keat 1981, p.179). In contrast, the Kress and Hodge model is one in which subjects 'know the truth' in their heart of hearts - or at least in the underlying forms of their language. Their demystification can then take the form of reversing the 'transformations' they have themselves applied to restructure and truncate their underlying knowledge before it reaches surface expression. The inspiration here is clearly Freudian, but what has happened is that the idea of the Freudian unconscious has been assimilated to the cognitivist view of unconscious linguistic representations, and this is far from obviously legitimate. In terms of the models outlined in footnote 10, the processes with which Freud is concerned have to do with the revision of representations which did once have or could have (had) a surface existence, rather than with the generation of such representations as envisaged in transformational grammar. Dream work, for example, is a function which maps representations into representations on the model R1=>R2 (where R = a representation). What then tells us about the operations of the human mind (with which the cognitivists are concerned) is the nature of the operations represented by '=>' not the content of R1. This becomes obvious in the case of Lévi-Strauss's theory of myth, where it is clear that there is no sense in searching for an ur-myth (M0) and where the interest is in the light on the human mind which the transformational relationships between myths (M1=>M2) display. In general, then, the error of LC and LI is to confuse both the 'transformation' (rewriting) of linguistic forms, texts and representations on the axis of historical time and the generation of linguistic forms in real-time psychological processes, with the generation of forms, texts and representations in a timeless formal theory (of which transformational grammars generating system sentences are undoubtedly instances ).
This error can be further illustrated by considering the treatment of indirect speech-acts in LC and LI.
Consider the following passage:
When the language of control is backed up by the reality of power, the more massive the deletion, especially of the fact of power and coercion, the more powerful, mystified and irrational is the control. Anyone who can give orders without even acknowledging this in the surface of his utterances has access to an insidiously powerful form of command. For instance, someone who can say 'the door is open' and be interpreted as saying 'close the door' has issued an imperative which has been totally deleted yet is fully effective. The person who obeys accepts the reality of a power that has not been claimed, which has been completely mystified into the form of an apparently neutral, factual observation. (LC, p.18)
This passage is typical of many (see, for example, LI, ch.5, pp 122-5, p.139; LC, ch.2). The common error is to suppose that if some act X (for example, my ordering you to shut the door) is performed by means of some words p (for example, the door is open), then underlying p and in the speaker's head there must be some specifically linguistic form which represents the act X in a specific way and from which p is derived, such that in the present example:
(17) I [legitimately] order you that you shut the door [on the basis of my power over you] => the door is open.
But this is simply a straightforward confusion of words and things20 which ends up formulating as an arbitrary base and arbitrary 'transformational' rules of language the very thing which needs to be explained - the fact that it is indeed sometimes possible to successfully and non-defectively order someone to close a door by saying to them the door is open, the kind of fact on which all attempts to specify language use as a function of linguistic form or form as a function of use are wrecked. (Compare Searle's critique of the Gordon and Lakoff 1975 theory of conversational postulates in Searle 1975a, pp 33-38).
Now there is a literature, much of which is to be found in Cole and Morgan 1975, and subsequent discussion: Cole 1978 and 1981; Gazdar 1979; Bach and Harnish 1979; Searle 1979c; and all of it having its origins in Grice 1957, 1961, 1968, 1969, 1975; Austin 1962; Searle 1969, etc. which attempts to understand the possibilities which LC and LI think can be explained by '=>', but which cannot.
There are occasional references to this literature in LC and LI, but its explanatory and critical power is nowhere utilised. Even if, per impossibile, the rewriting account given in LC and LI was correct, the conclusions the authors draw simply do not follow from it. For if I get you to close the door by saying the door is open, then there are two possible explanations of how this is brought about. (1) Either you have understood my words as constituting an order to close the door, which is what we are assuming they are, and take that to be sufficient reason to close the door (in which case there is no mystification of hearer by speaker). (2) Or you close the door for some other reason - for example, that you want the door closed - the utterance of door is open drawing your attention to the fact that it is open.
In this latter case, would it not be true to say that the speaker exercises power over the hearer without the hearer recognising it? Alternatively, might not clever speakers conspire to get their way by means of utterances which will be incorrectly interpreted by hearers but which are just as effective in producing the results desired by the speakers? The answer in both cases is basically yes, and there is a dizzying literature which discusses such cases usually as counter examples to the analysis of meaning given in Grice 1957 (see, notably, Strawson 1964; Schiffer 1973; Bach and Harnish 1979, pp 103-7. Compare also Pateman 1980a, Appendix 6). But though such phenomena are important to the student of mystification, their properties are not all alarming. If hearers simply misunderstand speakers, there is an important sense in which speakers have failed to achieve their communicative purpose, though the real world results may be no different (as in the example of someone's shutting the door because they want it shut, when they have been ordered to shut it). Again, if speakers are vague or ambiguous with a view to getting their way without having their intent recognised, or with a view to having it misconstrued, then it is quite generally the case that they thereby make available to hearers an interactional resourcewhich hearers can exploit. This is often overlooked (but not by Weiser 1974 and 1975. Cf. Giddens 1979, p.93). Thus, because an indirect speech act retains its literal or direct illocutionary force at the same time as it is used with a secondary illocutionary force,21 it is possible either for the hearer to misunderstand (as in (18)) or to exploit the indirection for his or her own purposes (thus, it should be easier to refuse the polite, indirect request in (19) than the direct, 'unmystified' request in (20), for reasons which Brown and Levinson 1978 explicate):
(18) S: The door is open
H: So it is! Let's close it [closes door]
(19) S: Would you mind if I smoked?
H: I'm afraid I would 22
or H: I'd prefer it if you didn't.
(20) S: May I smoke?
or H: I'd prefer it if you didn't.
The examples of language use analysed in LC and LI are primarily taken from written texts, including literary ones. The speech-act and conversation theory approach is perfectly applicable to written as well as spoken texts. Pratt 1977 has shown this for literary texts (see also Traugott and Pratt 1980), and Quentin Skinner has developed a speech act theoretic approach to the interpretation of texts in the history of political theory (see; e.g., Skinner 1972 and the discussion of Skinner's work in Graham 1980 and 1981). In the appendix to this chapter, I re-analyse a passage from Edmund Burke analysed in LI, but making use of a speech-act oriented approach close to that of Skinner and Graham.
It might be objected to my critique that the fact that as hearers interpreting utterances we need such an elaborate extra-linguistic apparatus as Searle, Grice, Bach and Harnish, etc. envisage only goes to prove that much language use is not transparent in meaning. If it was, the apparatus would be unnecessary, and we would all be better off. The answer to this is that there could not be a human language which did not at least contain deictic elements used by speakers ('I', 'you', etc.), and an indexical language can only be understood by hearers who have extra-linguistic competences at their disposal by means of which they can relate indexical elements in utterances to their intended referents. Nor is it possible to construct a language in which it would be impossible (ungrammatical) for speakers to say less than they know, or deny what they know, when they know that what they know is relevant to the (conversational) interests of addressees. 'Prevarication' is one of Hockett's defining design features of natural languages (see Lyons 1977, pp 83-4) and the fact that speakers often have cause to equivocate can be used to explain ( in the sense of a 'rational reconstruction') the origins and fuctions of regulative social norms of language use. Such norms appear to be universal, and are arguably necessary conditions of the possibility of the reproduction of any language use whatsoever, given humans and languages as they are. (See Pateman 1987, chapter 7). Hence, I conclude that the pragmatic, extra-linguistic apparatus we require to understand utterances cannot be taken as an index of anything essentially unsatisfactory about the language in which those utterances are produced.
1.5 Concluding Remarks
It would be entirely possible for speakers in a more egalitarian society than ours to be more explicit about the speech acts they perform, and to make less use of indirection. The study of indirection can indeed beheuristic of a culture and a society, which is one central aspect of the endeavour of LC and LI, which I have not dwelt on here. But the causes of indirection (notably politeness23) require an independent critique and cannot be criticised on the basis of the linguistic indirection which is the effect of these causes. I come away from LC and LI feeling that there are few or no clear cases of purely linguistic mystification except those like (21), where a speaker who uses a self-embedding structure rather than a right-branching structure, as in (22), poses a hearer considerable computational problems, which are usually unnecessary and may be intended to baffle:
(21) The elephant the giraffe the keeper bit kicked bellowed.
(22) The keeper bit the giraffe which kicked the elephant which bellowed.
But even then the cost of sending a telegram might well cause us to be grateful for the grammaticality of self-embedding structures!
Though in this paper I have been sceptical about claims made in LC, and LI, and critical of their theories, these books are a significant contribution to a kind and use of linguistics which, sadly, still has few practitioners. Where they fail is, essentially, in not having an adequate theory of the relationship between texts and contexts - in other words, a pragmatic theory by means of which we progress - to use a Freudian idiom - from symptomatic to scenic understanding.
1.6 Two Ways of Reading Edmund Burke
In chapter 7 of LI, 'Reality, Power, and Time', Kress and Hodge discuss English modal auxiliaries (must, may, can, etc.) which are generally thought of as having both epistemic and deontic senses and uses, though it is an issue what the connection between the two senses is. Whatever it is, it allows modals to be ambiguous in use. Of the third-person non-transactive form, He must come, as used to command, it is said that it, makes no claims that the speaker is the source of the command. The command is depersonalized and given an impersonal authority which is not specified and therefore is more difficult to challenge. The third-person non-transactive, then, has clear advantages for an ideologue, who wishes to mystify the basis of his authority and the fact that authority rather than reason is being appealed to (LI, pp 123-24)
To illustrate this claim, LI introduces a passage from Edmund Burke's Reflections on the Revolution in France (Published 1790. See Burke 1968), and comments at some length on it. Rather than just show the applicability of the critique in 1.4 above to the comments made about modals and the passage from Burke, I want primarily to illustrate what a better-grounded approach to the passage in question would have at least to involve - and there are other things which it would have to involve which I don't consider.
Here, first of all, is the quoted passage from Burke:
Good order is the foundation of all good things. To be enabled to acquire, the people, without being servile, must be tractable and obedient. The magistrate must have his reverence, the laws their authority. The body of the people must not find the principles of natural subordination by art rooted out of their minds. They must respect the property of which they cannot partake. They must labour to obtain what by labour can be obtained; and when they find, as they commonly do, the success disproportioned to the endeavour, they must be taught their consolation in the final proportions of eternal justice. (Burke 1968, p.372; quoted LI, p.124)
And here is what LI says about the passage from Burke:
The musts here do not all have the same force, but all are to some degree ambiguous, ranging between the first, must be tractable, and the last, must be taught their consolation. The sentence the people ... must be tractable and obedient is offered as a deduction or necessary consequence. If they want to acquire, then they must be tractable. But though there is a reason or motive given for this tractability an obligation is involved. The context for an 'inference' interpretation of the sentence would be something like: 'Look, the people are lying on the ground to let the magistrate's coach roll over them. They must be tractable and obedient. I deduce from this expression of devotion that they are tractable.
However, as the musts continue, the obligational/ coercive sense becomes increasingly prominent, until we reach they must be taught their consolation. This is a passive, which has an underlying X must teach them their consolation. Exactly who this X is, is unclear. It is also unclear whether Burke is commanding, advising, or, just possibly, deducing. He may simply be drawing X's attention to the need to teach the people their consolation if X is to avoid bloody revolution. The non-transactive speech model conceals the force of the utterance. Combined with the passive all signs are removed from the surface of the utterance that Burke has shifted the target of this command - if it is a command. Burke exploits the indeterminacies of the form to negotiate the contradictions in his role as ideologue, both instructing the people in their duties and reminding his masters of their responsibilities' (LI, pp 124-25).
This commentary, typical of many in LC and LI, is not exactly perspicuous: the text does not even speak of the people wanting to acquire, but rather of their being enabled to, it being assumed that they want to. More importantly, it repeats what I have already argued to be errors: the idea that force is 'concealed' if it is not explicitly represented in surface structure, and the attribution to properties of sentential forms what are more likely (possible) properties of the utterance ('indeterminacies'). But much more importantly still, the commentary suggests that what Burke is doing (and/or what is being said) in this passage, can be recovered by considering the linguistic forms and modifications of 'basic forms' which Burke employs. The assumption is that forms will yield up the meaning of a text if only 'we' read the sentences there on the page and evaluate them from the standpoint of a linguistic theory. If only hermeneutics were that simple! It isn't and what Kress and Hodge are missing is all the non-linguistic inferential work the competent reader engages in when reading a text - a vast work as attempts in artificial intelligence to achieve machine understanding of texts has shown.
What is the problem Kress and Hodge are trying to answer? It appears they are trying to establish the illocutionary force of each textual utterance in the passage, taking sentence boundaries as utterance boundaries - a disputable procedure, but one I shan't challenge. They suggest that at the outset the utterance force is that of a deduction,24 but that it then becomes 'unclear' whether Burke is 'commanding', 'advising' or 'deducing', though clear that he is 'instructing' one audience and 'reminding' another. If this is the problem, it can only be settled by considering much more of the discursive and extra-discursive context than LI contemplates. When this is done, the force of the textual utterance under discussion surely becomes more determinate - even allowing for the fact that we are making simplifying assumptions that Burke's English is at least as transparent to us as to his contemporaries, and that the categories of illocutionary force available to Burke are identical to or a proper sub-set of those available to us.
To establish the force(s) of an individual textual utterance we need, first of all, to know something about the speaker (writer) and 'hearer' (addressee and audience) (the latter a more complicated matter, as we shall see in a moment). Of the man, Burke, we know his biography, and, most relevantly, his status in English politics when he wrote the Reflections: 'a rather isolated figure, unpopular, frustrated, hard-pressed by exhausting labours, and to some extent already estranged from his old parliamentary friends and colleagues [the Whigs - TP]' (O'Brien in Burke 1968, p.80). He did not occupy a post of political power, and was not himself an English grandee. In the primary sense, he could felicitously command only his wife and servants.25 As for the 'hearer' side, Burke adopts the common literary device of casting his reflections into the form of a letter 'intended to have been sent to a gentleman in Paris' (Burke 1968, p.83), further described as 'a very young gentleman', (p.84), who had sought the author's opinions on events in France. This primary addressee may or may not have existed; no matter - we know, as did Burke's original readers - that the addressee is not the illocutionary target for his text.26 The target is his intended readership ('the implied reader' of literary theory); other readers participate as 'overhearers' (Levinson 1982) or 'eavesdroppers' (the term used in mass-media research). Kress, Hodge and myself are overhearers; hence our difficulties sometimes in understanding what is going on.
Understanding the rhetorical situation 27 Burke sets up helps clarify the possible force of his textual utterances. He says his opinions have been solicited, and in a primary sense what he can and does give in theReflection are his opinions. In speech-act theoretic terms, we should expect the bulk of his textual utterances to have the primary illocutionary force of Expressives or Representatives, though they could also felicitously have the primary force of weak Directives:28 advisings, for example - if it is accepted that when a young man solicits an old man's opinions, he may legitimately be given advice: I am sure that was accepted in the 18th century. The point also allows us to see the advantage to Burke of the rhetorical strategy he has chosen: because the target audience is not addressed in the text, its readership position is formally the same as that of overhearers or eavesdroppers. Burke can consequently express his opinions and give his advice in a situation where the risk of being charged with presumptuousness (i.e., lack of authority) is minimised. This might be regarded as mystificatory indirection, was not the rhetorical strategy a conventionalized one.
But why is it important for Burke to avoid presumptuousness? It is because his target audience is 'his Masters' and not at all 'the people' (i.e., the propertyless and powerless). Burke's target is surely not the latter. That being the case, his textual utterances cannot have the illocutionary force of instructings, though they may be meant to indicate to his masters, what he thinks they should instruct the people in. As for his masters, Burke is their inferior, as the biographical indications above establish. He could felicitously command them only in the name of God (but he is not a clergyman) or universal morality (but his ideology is not universalistic, unlike that of Paine's Rights of Man).
As it happens, the paragraph in Burke from which Kress and Hodge extract the quoted passage makes clear enough that Burke's target audience is 'his masters'. In the light of what has so far been said, there is not much difficulty in understanding the force of this piece of co-text:
Statesmen, before they valued themselves on the relief given to the people, by the destruction of their revenue, ought first to have carefully attended to the solution of this problem:- Whether it be more advantageous to the people to pay considerably, and to gain in proportion; or to gain little or nothing, and to be disburthened of all contribution? (Burke 1968, pp 371-72)
Having put this question into his readers' minds, Burke immediately answers it: 'My mind is made up to decide in favour of the first proposition' (ibid., p.372), he writes, and goes on to say that 'Experience is with me, and, I believe, the best opinions also'. It is two sentences later that the passage Kress and Hodge quote occurs. Considered globally, it is Burke's argument for his opinion, but takes the form of a statement of his solution to the 'problem of order' (the problem of political theory from Hobbes onward). In this co-textual context, I read the primary or dominant force of Burke's utterances asrepresentative , and consequently disambiguate the modals as having in this instance their epistemic sense. For what Burke is doing is set out what he takes to be the necessary conditions or essential components of good order: it is a necessary condition or essential component of good order that the people reverence the magistrate, and so on. What does differentiate Burke's claims is not at this level their epistemic or deontic character but their degree of contentiousness or the degree to which they are threatened. Reconstructing the context we can see that Burke's final claim reflects his opinion that the Whig Lords had failed to appreciate the role of the Church in securing their own economic survival, and had been rashly sympathetic to the confiscations of Church property in France (O'Brien in Burke 1968, p.42; see also pp 363-65). In context, it is the most important of Burke's claims and the one he immediately follows up with a condemnation which ranges over confiscators and atheists, lumped together:
Of this consolation, whoever deprives them, [the people - TP] deadens their industry, and strikes at the root of all acquisition as of all conservation. He that does this is the cruel oppressor, the merciless enemy of the poor and wretched; at the same time that by his wicked speculations he exposes the fruits of successful industry, and the accumulations of fortune, to the plunder of the negligent, the disappointed, and the unprosperous. (Burke 1968, p. 372)
And, of course, from any claim about the conditions of possibility or components of good order, there can be derived either prudential or moral imperatives about what ought to be done to promote good order. Today we would insist on the necessity of a major premiss to the inference from the nature of things to what ought to be done, but Burke writes within the context of a 'naturalism' common to conservative and radical varieties of 18th century thought. (Cf. Habermas 1974, chapter 2). Hence, there is a sense in which Burke's textual utterances can do double duty as both representatives and directives, which is rather different from saying that the primary illocutionary force is representative and the indirect force directive. But far from representing any concealment, this fact about the force of the utterances (if it is a fact) reflects the workings of an inferential scheme which we may see as fallacious, but which was quite acceptable to Burke and his readers. It would be an anachronism to see concealment in a use of modals which, if anything, is an index of an episteme not yet in crisis.
This last point also allows a move away from the view that recovery of speaker's intent is necessary and sufficient to the understanding of illocutionary force towards an acceptance of the view that aspects of their thinking and speaking may escape the intentions of contemporaries (cf. Graham 1981 and Foucault 1966 and 1969). But what I hope to have shown is how the understanding of the force of utterances has to be approached via context and co-text,29 and how consequently the operative senses of words (here epistemic versus deontic senses of the modal 'must') is co-decided from decisions about the possible force of utterances. We recover form, force and sense jointly, and in no way recover force and sense from form (cf. my essay, 'How is understanding an advertisement possible?' elsewhere on this site)
1. This paper is a revised version of Pateman 1981a, which in turn was revised and expanded from Part II of Pateman 1980b. I am grateful to William Outhwaite for comments on material now in section 2, and Dell Hymes for comments on material now in section 3 (though I have not expanded this into a full treatment of issues around 'Whorfianism', the ramifications of which can be gauged from Hymes 1966).
2. I owe this way of proving the point to Barry Cooper.
3. The problem set out here has an analogue in the problem of specifying a plausible inference mechanism in theories of human information processing. If humans processed incoming information (utterances) using a device representable by any standard logic, that device would crank out an infinite or at least very large number of inferences from very limited input. This is an obviously impossible account of the actual inference mechanism(s) used by humans; hence it is necessary to describe a different kind of mechanism altogether, or else to place 'relevance' constraints on the operation of a standard logic.
4. The language here is that of Chomsky 1965.
5. Though as subsequent discussion in this section indicates, it is a contestable analysis of passives to say that they involve agent deletion, as opposed to simply not specifying the agent.
6. As expressed for example in the following quotation, 'we can hypothesize and suggest that, since language functions to deceive as well as to inform, every component of the grammar will contain one set of forms which allow the speaker to avoid making distinctions which are primary and another set where these distinctions have to be made sharply and with precision' (LI, p.125).
7. Hymes 1966, pp 120-21 distinguishes four theories about the relation of language and culture, and four about language and the individual, resulting in sixteen different possible positions. In this section, I am exclusively concerned with the position which treats 'language as primary (source, cause, factor, independent variable, etc.') in culture' (p.121).
8. The stereotypical examples by which differences between languages in the distinctions which are lexicalized or grammaticalized are illustrated, such as 'the n Eskimo words for snow', encourage the commission of this fallacy. For they lead us to overlook the fact that a language can lexicalize or grammaticalize distinctions which are tendentious, or more relevant to some speakers than others - the best examples are undoubtedly provided by gender distinctions. See Spender 1980 for a survey.
9. In Hodge and Kress 1974 the authors write, 'The view of language put forward here is not meant to include the proposition that language (or rather the models fixed in language) determines thought: this extreme position could not be attributed to Whorf, and cannot be attributed to our view ... we seriously wish to keep open the possibility that there are other than linguistic models available to the member of a society' (p.18). It may be that despite the quotations from LI in this section I have misunderstood the tendency of that book, or else that Kress and Hodge have changed their position between 1974 and LI.
10 Tony Trew and the other authors of LC and LI have in common that they approach ideology in language in terms of the transformation of one linguistic form into another. But Tony Trew is mainly concerned with the rewriting of forms in time, for example, the changing presentation of the 'same' story in newspapers. He is concerned with practices which can be diagrammed like this:
LF1 => LF2
where 'LF' stands for linguistic form; the arrow represents a transformation; and the horizontal axis is a time axis. Here ideology is a function (a transformation) of surface forms into surface forms.
In contrast, the Rossian approach of the other authors is a psychological theory of the generation of surface forms from underlying (deep, basic) forms by means of transformations. In this case, the underlying form should be thought of as represented by an English sentence, rather than as necessarily realized as an English sentence. Diagrammatically:
indicating that a surface linguistic form (LFs) is generated from an underlying form ( represented by LFd), and where the vertical axis represents psychological depth. Ideology is then a function from underlying to surface forms. It is this theory with which I shall be almost exclusively concerned.
The critical conclusions I arrive at suggest a third model of ideology in language which one might call 'base generated ideology', because ideological forms are directly generated without transformations. This does not, however, directly bear on or affect the validity of Trew's very interesting analyses.
11 LC and LI depend on a theory of agent deletion for passives which is disputed. Thus Bresnan 1978 (to which LI refers) develops a lexical-interpretive grammar which yields separate lexical entries for passive verbs, the lexical entry showing the logical argument structure. On this theory, 'the agentive by-phrase that optionally appears with passives [e.g. The cat was eaten by the tiger - TP] ... can be analyzed simply as an optional prepositional phrase that functions semantically to identify the logical subject of the passive verb' (Bresnan 1978, pp 20-21). However, Bresnan's theory does allow for the possibility of making claims like those LC and LI wish to develop, for she says, 'The set of lexical relations is not fixed: it can change to embody those concepts that become important for communication' (ibid., p.21). See also Coetzee 1980 for a survey of theories of the passive, and many interesting remarks bearing on what follows.
12 'Transformation' in LC and LI does not have the sense it has in Chomsky 1957 or 1965, as the authors point out (LI, p.9; see also Hodge and Kress 1974 for discussion). That is to say, transformations in LC and LI do not map phrase markers into phrase markers, nor are they meaning preserving - specifically, they are not constrained by recoverability conditions. In Hodge and Kress 1974, to which LI refers us (p.9), transformations are real psycholinguistic processes which operate on semantically-organised basic models or Gestalten. If in some cases they correspond one-to-one with Chomskyan transformations, this is simply fortunate coincidence. Given their more general character, I shall regard the representations of transformations in LC and LI as rewriting rules, and argue below that they are not psychologically plausible.
13 In an uninterpreted Aspects-type syntax, (1) would be derived from a deep structure which contained no more than a dummy unspecified agent (' '), which the passive transformation would delete, but which would be held to be recoverable after deletion. See Chomsky 1965, pp 128-31 and compare Bresnan's position as sketched in note 11.
14 To convert (7) to the source of (1) as a system-sentence, it would be regarded as legitimate in Ross's theory to replace 'tell' with an abstract performative verb, yielding (7a): (7a) I VERB you that I feel that adoption of the proposal is inadvisable.
15 (8) contains thinly disguised internal sentence boundaries, but this will only matter in a theory which insists on the recoverability of deletion transformations. As LC and LI do not wish to adhere to such a theory, I think (8) is a candidate for a psychologically-real source of (1) compatible with the theoretical position of LC and LI.
16 Thus when it is true that (10), a newspaper story using (9) would be misleading, since it would probably license the inference that the police fired randomly into the crowd, though (9) is also true:
(9) The police fired into the crowd.
(10) The police fired into the crowd aiming at Joe Smith.
Note that these sentences don't use the passive construction. Stanley (now Penelope) writes that 'The passive voice provides us with a syntactic construction for deceiving our readers into believing that we're giving them information when we're not. It is a construction that allows us to lie without overtly lying, and only the careful, analytical reader will notice that information is missing' (Stanley 1975, p.30). In the light of (9) and (10), the active voice can be used just like the passive - material can be left out (in (9) the logical object of the police's action; in passives the logical subject (the actor)) without this being noticeable. Indeed, whereas (9) as a newspaper report could licence the inference of random firing, and does not contain a trace of its logical object, at least in the case of passives we always know that somewhere there must be an actor knowledge of whose identity could be relevant to us! (A fact exploited by ironic uses of the agentless passive. See Coetzee 1980, p.214.)
17 Which is the better newspaper headline, (11) or (12)? Isn't (12) likely to make readers vainly search their minds for the referent of 'Simon Chapman', a person they have no reason to know?
(11) John Lennon murdered.
(12) Simon Chapman murders John Lennon.
In this connection, I imagine that even if (11) is agreed to be less misleading than (12), people will still want to know the information (12) adds to that provided by (11). But though when asked to choose between pairs of less informative and more informative statements, I take it we will prefer the more informative, it does not follow that this preference is generalisable over all pairs (it would be a fallacy of composition so to generalise). For generalised, our preference for more over less information would produce an information overload with which we could not cope. I think this point not only connects to familiar dilemmas in rational choice theory, but is of some practical and political significance, though I cannot explore that here.
18 Though this is false, as the following exchange indicates:
S: Go home
H: Is that a request or an order?
S: I request you to go home.
19 'Picketing' is listed in my third (1944) edition of the shorter OED, though nothing hinges on this fact in what follows.
20 Or of words and cognition, since if the speaker does indeed order someone to shut the door by saying the door is open, they must in some sense, know that that is what they are doing. It may be that the works under consideration are happy to use a generative semantics just because they equate language and thought, whereas the whole tendency of this chapter is to argue the importance of keeping them apart.
21 This is Searle's position (Searle 1975a; see also Searle 1975b). Lyons 1977, p.786 says that 'the conventionalization of such formulas as Can you ... ?Would you mind ...? may reach the point where is is no longer reasonable to treat the utterance in which they occur as being, even incidentally, anything other than requests', but I think this view is falsified by the conversational facts. (Levinson 1980 calls Lyons' position the 'idiomatic' view of indirect speech acts, and also says it does not stand up).
22 The use of 'afraid' explicitly indicates that H recognizes the interrogative as an indirect request, and serves to express regret that the request is being refused by means of an answer to the question expressed by the interrogative.
23 See Brown and Levinson 1978 whose theory (using the concepts of power, distance and imposition) supersedes that sketched at LC, p.29 (using the concept of power).
24 The problem here for speech-act theory is that a deduction is from one argument to another. If this means there are two utterances involved (and two sentences ?) then 'deduction' is a discourse feature not analysable atomistically. (For many other problems in speech act approaches to discourse, see Levinson 1981).
25 Except insofar as he could command in God's name or in the name of universal morality, that is as a relayerof commands of a higher authority. See note 26.
26 In using 'relayer', 'addressee' and 'target' I conform my usage to that of Levinson 1982.
27 I take the term 'rhetorical situation' from Ellrich 1969.
28 Using the taxonomy of Searle 1975c, which though the most adequate taxonomy actually handles opinions unsatisfactorily. See Atelsek 1981.
29 Of course, the 'hermeneutic circle' operates here: in Dilthey's original formulation, 'The totality of a work must be understood through its individual propositions and their relations, and yet the full understanding of an individual component presupposes an understanding of the whole' (quoted from Outhwaite 1975, p.34). In other words, we have to take the force of some utterances as 'always already' (immer schon) established as we set out to decide the force of others, our decisions about which may entail revision of our initial assumptions, and so on. What I have offered in this Appendix is a mere fragment of an analysis, meant to illustrate a method of interpretation rather than definitely solve a question in reading Burke.
Originally a 1981 review of the books Language and Control and Language and Ideology - listed as Pateman 1981a in the Bibliography above; revised into a chapter of my 1983 doctorate; further revised in the late 1980s for an unpublished book, and published on this website for the first time in 2002. The Bibliography above is the consolidated Bibliography for the whole unpublished book