Chapter 3 - The Sociolinguistic Codes

In looking at the sociolinguistic codes I shall begin with a historical account of the development of Bernstein's ideas of the two codes. I shall then report his sociological speculations about the kind of family in which the codes arise, which will lead on to a sketch of the part played by the codes in the explanation of educational failure. After these expository sections I shall turn to consider two of the contrasts (implicit versus explicit and particularistic versus universalistic) that are central to Bernstein's various accounts of his theory and I shall set out my own reinterpretations of these contrasts and some of the issues to which these reinterpretations lead. These discussions should lead into the sketch in the following chapter of what an empirical research programme inspired by my renditions of Bernstein would look like in a Jamaican context, a sketch that will involve some examination of the bearing, or rather lack of bearing of Craig's work in a more traditional sociolinguistic vein.

Bernstein's Sociolinguistic Codes

The core of Bernstein's original "intuition", "an obstinate idea in me which I could neither give up nor properly understand" (1971b = 1974, p. 1) as he calls it, is that there is a difference between working class and middle class speech, and that this difference is fraught with educational consequence. The labour has been to refine the characterization of this difference.

We have, in Bernstein's research programme, a small example of a problem that has been exploited by some philosophers of science: the commensurability of differing theories - did Newton and Einstein talk about the same things, or different things? If you insist on sticking to what they, or others, might have offered as definitions of their terms, and what they said about their various entities, as exhausting the meaning of their claims, you seem forced to admit that Newton's "mass" is a different quantity from Einstein's "mass", and so on with the other terms in their theories. But if that is so, how are they in conflict, any more than a theory about kinship is in conflict with a theory about electrons? To regain what non-philosophers rightly assume to be conflicts, we need a different construal of the meaning of their expressions. Part of this construal may well involve us in saying that both Newton and Einstein meant by "t" whatever it is that underlies or causes such and such phenomena. Their accounts of what it is that does underlie the structure of the solar system, or apples falling to earth, or whatever, may well differ; but such a construal allows that they are both trying to make conflicting statements about "the same thing".

Any such approach to explicating a theory obviously rejects the idea that an operational definition is any real sort of definition. In our present context this means that I shall not take Bernstein to be literally defining his codes when he says things that suggest that he may be, and I shall charitably allow him to be trying to talk about the same difference, whatever it is, between typical working class and middle class speech, at the different stages of his research, when he offers rather different accounts or criteria for the codes.¹ While it might be best to move immediately to his latest reflections upon the contrast, both its intrinsic interest and the possible suspicion that Bernstein's has been a "degenerating" research programme, in Lakatos' phrase, - cf. the claim that there is a "growing body of evidence that Bernstein's programme has moved into a degeneration phase" (Toft and Kitwood, 1980, p. 54) - suggest the need to look at some of the historical development of his characterizations.

In Bernstein's earliest papers, his distinction is between "two forms of linguistic expression", two forms of "language-use", that he labels "public" and "formal" (1959 = 1974, p. 42). The paper I have been citing also contains two notorious lists of the characteristics of these two different forms of language use. For a public language they are:

(1) Short, grammatically simple, often unfinished sentences, a poor syntactical construction with a verbal form stressing the active mood.

(2) Simple and repetitive use of conjunctions (so, then, and, because).

(3) Frequent use of short commands and questions.

(4) Rigid and limited use of adjectives and adverbs.

(5) Infrequent use of impersonal pronouns as subjects (one, it).

(6) Statements formulated as implicit questions which set up a sympathetic circularity, e.g. 'Just fancy?', 'It's only natural, isn't it?', 'I wouldn't have believed it.'

(7) A statement of fact is often used as both a reason and a conclusion, or more accurately, the reason and conclusion are confounded to produce a categoric statement, e.g. 'Do as I tell you', 'Hold on tight', 'You're not going out', 'Lay off that.'

(8) Individual selection from a group of idiomatic phrases will frequently be found.

(9) Symbolism is of a low order of generality.

(10) The individual qualification is implicit in the sentence structure, therefore it is a language of implicit meaning. It is believed that this fact determines the form of the language (ibid., pp. 42-3).

(1) Accurate grammatical order and syntax regulate what is said.

(2) Logical modifications and stress are mediated through a grammatically complex sentence construction, especially through the use of a range of conjunctions and relative clauses.

(3) Frequent use of prepositions which indicate logical relationships as well as prepositions which indicate temporal and spatial contiguity.

(4) Frequent use of impersonal pronouns (it, one).

(5) A discriminative selection from a range of adjectives and adverbs.

(6) Individual qualification is verbally mediated through the structure and relationships within and between sentences. That is, it is explicit.

(7) Expressive symbolism conditioned by this linguistic form distributes affectual support rather than logical meaning to what is said.

(8) A language use which points to the possibilities inherent in a complex conceptual hierarchy for the organizing of experience" (ibid., p. 55).

I include these sorry rag-bags of often ill-expressed features (I wouldn't have believed it is hardly formulated as an implicit question; there is no reason to suppose that Lay off that telescopes any statement of a reason) because for all their faults they seem to connect with popular misconceptions about language use. Education students regularly regurgitate them (and not just students: Craig, 1984, once more reminds the world of the "public" language list), and I do not believe this is simply because Bernstein has already gone to the trouble of listing them. People, or at least teachers, do seem to believe that their own speech is grammatically accurate and exploits the resources of the language to the full, while that of lower status children and adults is often unfinished, repetitive, limited, and semantically confused. But even at this stage, Bernstein recognized that these differences did not go along with a difference in range of vocabulary, at least for younger children, (1959 = 1974, p. 43) though working class children are said to be likely not to see the need to increase their vocabularies as much as middle class children (ibid., p. 44, and 1958= 1974, p. 34). He also noted, but only to dismiss it, the possibility that he was characterizing a difference between spoken and written language (1959 =1974, p. 56) or between dialects (1958 = 1974, p. 27).

But having already said that I would not press the attempts to operationalize the contrast, let me turn to what Bernstein has to say about the point of the contrast. Part of this is already contained in the emphasis in his tenth feature of public language use, the implicitness of this sort of language. Much is left unsaid because it can be taken for granted; social relationships are stably given. "The public language is ... a language to be used between equals (from a middle class point of view)" (1958 = 1974, p.34). On the other hand, formal language use seeks explicitness and requires the giving of reasons and explanations. Bernstein also associates a predilection for the concrete, the here and now, with his public language and conversely a focus on the abstract and general with the formal language (ibid., p. 29 and p. 33; 1960 = 1974, p. 61). As we shall see, he sees these features flowing from the kinds of personal relationships and typical concerns of families in different social class strata.

Given the widespread assumption that Bernstein was at first, if not later, adopting a "deficit" model of working class speech,² it is perhaps worth noting that at this early stage Bernstein made the point that changing a form of language use involves a lot more than may be apparent at first sight:

A public language contains its own aesthetic, a simplicity and directness of expression, emotionally virile, pithy and powerful and a metaphoric range of considerable force and appropriateness.... To simply substitute a formal language (which is not necessarily a logical, impersonal, emotionally eviscerated language) is to cut off the individual from his traditional relationships and perhaps alienate him from them. This is the old problem of Gemeinschaft and Gesellschaft in another guise" (1959 = 1974, p. 54).

When Bernstein began to study linguistics, the inadequacy of the earlier lists of characteristics became very obvious - no one, for instance, usually speaks complete and grammatically correct sentences. But in the papers which he groups in 1974 as "Developments" he retains a fairly positivistic operationalization of the distinction, which was now however formulated in terms of restricted and elaborated codes. So the codes were introduced in this way in 1962:

Two general types of code can be distinguished: elaborated and restricted. They can be defined, on a linguistic level, in terms of the probability of predicting for any one speaker which syntactic elements will be used to organize meaning. In the case of an elaborated code, the speaker will select from a relatively extensive range of alternatives and therefore the probability of predicting the pattern of organizing elements is considerably reduced. In the case of a restricted code the number of these alternatives is often severely limited and the probability of predicting the pattern is greatly increased. On a psychological level the codes may be distinguished by the extent to which each facilitates (elaborated code) or inhibits (restricted code) the orientation to symbolize intent in a verbally explicit form" (1962a = 1974, pp. 76-7).

A remark made in a postscript in 1961, which I have not found reprinted, might bring home the absurdity into which it is easy to fall with such operationalizing: "an actor, also, would be using a pure form of a restricted code, although from the point of view of the audience it would be an elaborated code" (1961, p. 259). Even ignoring the fact that some audiences know their Hamlet as well as the actors, this is a particularly blind following of one's criteria. One also wonders how useful it is to conjoin pure rituals, such as the canon of the mass, with the tired opening gambits at a cocktail party or the perennial English conversations about the weather, as Bernstein does in these papers (1962a = 1974, p.77; 1965 = 1974, p. 126).

That this is no mere antiquarian issue can be seen from Atkinson's (1981, 1985 ch. 5) gloss on restricted code. He usefully introduces Saussure's terminology of paradigmatic and syntagmatic choices to explain Bernstein's selections. In a phrase such as a bottle of rum the word bottle shares the same paradigm with words like glass, tot, barrel, etc while having syntagmatic relations with a, of, and rum. But while this explanation is uncontentious, Atkinson goes on to suggest that "the restricted code can be illustrated most aptly by ... the construction of oral epic verse"(1981, p. 89). Homer's formulae are compared with the rule governed structures for ritual insults in Black American 'sounding', as described by Labov (1972). Certainly both exemplify a limitation on allowable paradigmatic or syntagmatic selections, and so give some basis for greater predictability in some respects at least, but ritualized speech is not anyone's typical kind of language use. The idea, central to Bernstein's position, that some people are confined to restricted code cannot plausibly be understood as confining them to formulae of these types. Again, if restricted code is to carry weight with respect to certain cognitive abilities it is difficult to see how an exposition in these terms will help: many of Homer's stock phrases are referring expressions like any other; the variable part of a Labovian ritual insult allows a vast range of syntactic forms and their associated kinds of cognitive functioning.

While Bernstein toyed with these notions, his emphasis was usually more on syntactic predictability than the total (lexical and therefore syntactic)predictability of pure rituals. While some of what Bernstein said might suggest that the codes were somehow two languages, generating different numbers of alternative structures,³ his further descriptions stress semantic matters, as for instance in this remark that the restricted code's major function "is to reinforce the form of the social relationship (a warm and inclusive relationship) by restricting the signalling of individuated responses" (1962a = 1974,p. 78).

It is in these papers that the contrast between particularistic and universalistic is developed. We shall be looking at this in more detail later, but for the time being we should note that the contrast was introduced looking in two or three directions, as it were. Thus we are told that a restricted code is particularistic with reference to its meaning (it leaves a lot unsaid, so only people sharing the context will be able to fully understand) and with reference to "the social structure that controls its inception" (1962a = 1974, p. 78) which, I take it, refers to the "closely shared identifications self-consciously held" (ibid., p. 77) by the people using it (not anyone can be mother or best friend); but on the other hand it is possible for the speech model for this code to be universalistic "as its use depends on the characteristics of a form of social relationship which can arise at any point in the social structure" (ibid., pp. 78-9). On the other hand, the speech model for elaborated code in our societies is particularistic - "access to an elaborated code will depend not on psychological factors but on access to specialized social positions within the social structure" (ibid., p. 79) - while in its meaning, in particular, it is universalistic since "it summarizes general social means and ends" (ibid., p. 79).

But Bernstein then goes on to stress a special case of restricted code, which is, he thinks, the most important for English society, in which not only the meaning but also the speech models are particularistic. In an emphatic sentence he says, "in this situation the individual is wholly constrained by the code. He has access to no other" (ibid., p. 79). The point is, I believe, that while restricted codes arise at almost any place in the social structure, in class societies not everyone has access to all such places so that my restricted codes will not be shared by any and everyone. If I have no elaborated code I am then stuck within a parochial boundary of those people who share my social position. And I may, to take a fairly banal example that Bernstein offers, be unable to get off with a girl at the cocktail party, for "the ability to switch codes controls the ability to switch roles" (1965 = 1974, p. 129).

Whatever the underlying intention may have been, the so-called "definitions" in the papers between 1962 and 1965 explicate the codes directly in linguistic and largely non-semantic terms. As the extensive empirical work of the Sociological Research Unit developed, and as Bernstein reformulated his initial insights, codes were eventually characterized in semantic terms, as had always been the intention, and at a fairly high level of abstraction from concrete speech occurrences. Bernstein, in need of a linguistic theory to use in the empirical research, had adopted and modified Halliday's approach, and in particular had picked on four crucial contexts (as Adlam notes [1977, p. 7], context here does not refer to the surroundings of particular utterances but to types of speech situation):

(1) The regulative context - these are authority relationships where the child is made aware of the rules of the moral order and their various backings.

(2) The instrumental context, where the child learns about the objective nature of objects and persons, and acquires skills of various kinds.

(3) The imaginative or innovating contexts, where the child is encouraged to experiment and re-create his world on his own terms, and in his own way.

(4) The interpersonal context, where the child is made aware of affective states - his own, and others (1971c = 1974, p. 181).

if the linguistic realization of these four contexts involves the predominant use of restricted speech variants, I shall postulate that the deep structure of the communication is a restricted code having its basis in communalized roles, realizing context-dependent meanings, i.e. particularistic meaning orders. Clearly the specific grammatical and lexical choices will vary from one to another.
If the linguistic realization of these four contexts involves the predominant usage of elaborated speech variants, I shall postulate that the deep structure of the communication is an elaborated code having its basis in individualized roles realizing context-independent universalistic meanings" (1971c = 1974, pp. 181-2).

One consequence of this development was that Bernstein could easily allow that the purely linguistic resources of the language might be shared by all speakers, that they could all understand and produce any grammatical sentence, so that persons who only had access to a restricted code could well produce and understand what he would label elaborated variants. But given the different speech policies, the different codes, one could expect (and as it turned out, find) differences in the relative frequency of various structures, grammatical elements, etc. which produce "an overall difference in the patterning of the speech" (1971b = 1974, p. 13). So Bernstein and his team continued to rely on largely non-semantic criteria for their observation schedules.

In his later work, the centrality of meaning for the codes is underlined in his usual formula: "a code is a regulative principle, tacitly acquired, which integrates relevant meanings, the form of their realization and their evoking contexts" (1977a, p.180) and he has begun to characterize the distinction between restricted and elaborated codes in terms of their respective distances from "a material base" (1981, pp. 331-3). This procedure makes very obvious the relative nature of the distinction; it is matter of more or less, rather than some absolute intrinsic nature of the kind of language use.

Bernstein's operationalizing at this point can be seen from the work by Holland that he refers to. Holland experimented with two groups of eight year old children from different social class backgrounds; she showed them coloured photographs of 24 food items and asked them to group them any way they liked and then to try to regroup them; in another interview the items had been grouped already in terms of context independent principles (such as coming from the sea rather than context dependent ones such as what we had for breakfast) which the child was asked to guess and then some extra items were introduced which the child was asked to add to the existing groups; finally the first procedure of grouping at will was repeated. Holland summarizes some of the important findings thus:

All the children recognized context independent principles more readily than they produced them in their own groupings, although the middle class children showed greater facility in this than the working class children, but it is important to note the point at which the working class children revert to context dependent principles. It is when they are asked themselves to produce a principle for the allocation of extra items to the interviewer's four groups, which have been discussed in some detail and for which the children by now have available context independent principles of organization. In their own production the mode favoured by the working class children is context dependent (1981, pp. 14-5).

The simpler the social division of labour, and the more specific and local the relation between an agent and its material base, then the more direct the relation between meanings and a specific material base and the more restricted the coding orientation.... If agents become specialized categories of the social division of labour, and their location is fixed and so nontransposable, then coding orientations become specialities of position within the social division of labour.... These coding orientations are in no sense inevitable consequences of any position. Coding orientations are not intrinsic to different positions. Whether they become so depends upon the distribution of power. Thus the distribution of coding orientations depends upon the distribution of power created by the principles regulating the social division of labor" (1981, pp. 332-3).

So we have seen the crude outlines of the development from public versus formal language use through syntactically "defined" codes to the most recent re-affirmation of the abstract and semantic nature of codes as rules selecting and governing the kinds of linguistic activity to be engaged upon, rules embodying alternative forms of "communicative competence" (a phrase associated with Hymes which was used to characterize the codes by Hawkins [1977, p.193]),⁴ and rules that are fundamentally sociological in that they transform social relations into (spoken) action and thereby help to reproduce those social relations. I shall come back later to suggest an alternative rendition of the codes, or rather of their specification in later Bernstein; but for now I wish to turn to Bernstein's sociological account of the family environments within which the codes are meant to arise.

Bernstein's Sociology of the Family

While there have been sweeping changes in at least the operationalizing of the codes, Bernstein's views about the social background in which they arise have been much more consistent. In looking at them I shall start by showing that he has always regarded restricted code as common, in one form or another, to everyone. This has been mentioned in the preceding discussions but it is perhaps worthwhile demonstrating Bernstein's consistency here since Craig for one seems to think that it was a late concession on Bernstein's part (cf. "the fact explained in Bernstein's later work that some upper-class social situations can also give rise to restricted codes" [Craig, 1974, p. 6]).

But in fact from the very beginning Bernstein has insisted that everyone will have access to restricted codes. In his first paper, he said

The middle-class child is capable of manipulating the two languages - the language between social equals (peer groups), which approximates to a public language, and a formal language which permits sensitivity to role and status. This leads to appropriateness of behaviour in a wide range of social circumstances (1958 = 1974, p. 30).

Children socialized within middle-class and associated strata can be expected to possess both an elaborated and a restricted code, whilst children socialized within some sections of the working-class strata, particularly the lower working-class, can be expected to be limited to a restricted code (1965 = 1974, p. 136);

one of the effects of the class system is to limit access to elaborated codes (1971c = 1974, p. 176);

class regulates the distribution and realization of elaborated codes but only the realizations of restricted codes. In other words, all individuals possess a restricted code, but the significance of its realizations vary according to whether it is the major code (1975 = 1977a, p. 27).

To see the continuities and developments in Bernstein's view of the family background I shall expound his account in the very first, 1958, paper and then that given in what is perhaps the fullest recent paper, 1971d (though as he notes in 1971b [= 1974, p.9], many of the ideas in this paper were first formulated as early as 1962).

According to Bernstein's first paper, the typical middle class child

grows up in an environment which is finely and extensively controlled; the space, time, and social relationships are explicitly regulated within and outside the family group. The more purposeful and explicit the organization of the environment with reference to a distant future, that is the greater the rationality of the connections and interrelations between means and distant ends, the greater the significance of objects in the present. Objects in the present are not taken as given, but become centres for enquiry and starting points for relationships" (1958 = 1974, p. 29).

The working-class family structure is less formally organized than the middle-class in relation to the development of the child. Although the authority within the family is explicit, the values which it expresses do not give rise to the carefully ordered universe spatially and temporally of the middle-class child. The exercise of authority will not be related to a stable system of rewards and punishments but may often appear arbitrary. The specific character of long-term goals tends to be replaced by more general notions of the future, in which chance, a friend or relative plays a greater part than the rigorous working out of connections.... By implication a more volatile patterning of affective and expressive behaviour will be found in the working classes (ibid., p. 32).

Notice here that middle class access to restricted code (at this time, of course, a public language use) derives not so much from their rearing by their mothers as from peer-group interaction, though it would be a rare middle class family in which the mother never let up on the drive for formality and explicitness. The public language is seen primarily as a means of reinforcing group membership, for indulging in "phatic communion", in Malinowski's phrase:

as the structure of a public language reinforces a strong inclusive relationship, the individual will exhibit through a range of activities a powerful sense of allegiance and loyalty to the group, its forms and aspirations, at the cost of exclusion and perhaps conflict with other social groups which possess a different linguistic form which symbolizes their social arrangements" (1959 = 1974, pp. 47-8).

In these early papers, attention shifts from the kinds of adult language use in question to some of the consequences of its use in the context of parent-child interactions, but the two issues are not clearly separated and there is a tendency to delineate a blanket stereotype (excused, perhaps, by being acknowledged an "ideal" type) of middle and working class life.⁵

As we shall soon see, the later view involves amore detailed specification of parent-child relationships for mediating the codes, and it also severs the definitional bond between social class and code. Class is still the final determiner, but it works through different types of family, examples of which can be distributed across class boundaries.

As for the codes themselves, the basic idea is still the same:

A restricted code will arise where the form of the social relation is based upon closely shared identifications, upon an extensive range of shared expectations, upon a range of common assumptions. Thus a restricted code emerges where the culture or sub-culture raises the 'we' above the 'I'. Such codes will emerge as both controls and transmitters of culture in such diverse groups as prisons, the age group of adolescents, army, friends of long standing, between husband and wife. The use of a restricted code creates social solidarity at the cost of verbal elaboration of individual experience. The type of social solidarity realized through a restricted code points towards mechanical solidarity, whereas the type of solidarity realized through elaborated codes points towards organic solidarity (1971d = 1974, pp. 146-7).

it would follow that the greater the reduction in the range of alternatives, the more communal or collective the verbal meanings and the lower the order of complexity and more rigid the syntactic and vocabulary selections - thus the more restricted the code (ibid., p. 148).

Since he claims that "the connection between social class and linguistic codes is too imprecise" (ibid., p. 152) Bernstein proceeds to postulate two types of family - positional or person-oriented - depending on the nature of the family role system with respect to decision-making and its procedures for social control.

In a positional family, decision-making depends on a member's formal status; there is a clear separation of roles. We should expect close relationships between parents and grandparents and either, in the middle class, close supervision of the child's peer-relations or, in the working class, little interference with the peer group. Such a family would produce a closed communication system.

"By contrast we could consider a family type where the range of decisions, modifications and judgements was a function of the psychological qualities of the person rather than a function of formal status" (ibid., p. 153). In such person-oriented families, peer group relations would be matters for family discussion rather than legislation; an aspect of the open communication system that such a family fosters. More alternatives are open to all members of the family; particular sorts of discussion become necessary; reflection on principles, consequences, and in particular motives and dispositions, is encouraged.

Bernstein is particularly interested in the differing modes of social control associated with the two sorts of family: positional - power or the norms regulating statuses; person-oriented - much more arbitration, persuasion, "linguistically elaborated meanings rather than ... power" (ibid., p. 155) which is, however, still the ultimate basis of authority. Social control in positional families should then manifest itself in explicit imperatives, or what Bernstein calls positional "appeals", whereas person-oriented families will display much more use of personal appeals. An example Bernstein offers is (positional) Children kiss their Grandpa versus (personal) I know you don't like kissing Grandpa, but he is unwell, and he is very fond of you, and it makes him very happy. Positional appeals are claimed to lead to the development of shame; personal appeals to guilt.

With these distinctions Bernstein now has a considerably complicated taxonomy on hand: positional and person-oriented families may have either restricted or elaborated codes, and each code might be person or object focussed; finally there are three modes of control available. Not all of the possible categories are likely to be found - Bernstein only deals with eight as live options, since positional goes along with object code orientation and personal with person orientation and finally positional families are not likely to use personal means of social control. But he gives a few examples that display the potential utility of his taxonomy. Thus he suggests that the core of the educability problem consists of

positional families who were deeply embedded in their community operating essentially with imperative modes of control and where the children were socialized through unsupervised age peers or mates. Here we could expect the development of restricted code (object) (ibid., p. 160).

In very broad terms Bernstein suggests that the English working class family is traditionally positional, with the girls being given complex mediating roles that incline them to acquire a "more differentiated, more individualized use of language" (ibid., p. 161); middle class families both in England and the USA are seen as moving from an elaborated code, positional type to elaborated code, person-oriented type, especially in those families engaged in managerial, service and media occupations (a group of "new" middle class members who are very significant for Bernstein's attempts to relate educational changes to wider social pressures). He concludes the discussion with some speculations about the "basic organizing concept ... whose ramifications may be diffused throughout the culture or sub-culture" corresponding to four of the possible families:

Positional - Restricted Code (object)
The basic organizing concept here would form around the concepts of authority or piety.

Personal - Restricted Code (person)
The basic organizing concepts here would be authority/identity in a state of unresolved tension. By 'identity' I simply mean a preoccupation with the question of 'who am I?'

Positional Elaborated Code (object)
The basic organizing concept here would centre about the concept of rationality.

Personal Elaborated Code (person)
The basic ordering concept would refer to the concept of identity (ibid., p. 165).

The Explanation of Educational Failure

It is perhaps worth noting initially that Bernstein is not attempting to provide a total explanation of educational failure. As Halliday says,

he is offering an interpretation of one aspect of it, the fact that the distribution of failure is not random but follows certain known and sadly predictable patterns.... Even here Bernstein is not trying to tell the whole story; what he is doing is to supply the essential link that was missing from the chain of relevant factors (1973, p. ix).

Just as Bernstein's later accounts of the class stratified types of family environment that produce restricted or elaborated codes of different orientations are mainly elaborations of his earlier sketches, so his explanation of working class educational failure has remained basically the same through changes in formulation. In barest outline it is probably obvious from the preceding section: only some families promote elaborated codes; elaborated codes are the currency of formal educational systems, the precondition of success therein; so some children start at a tremendous advantage in the race for educational certification. Those children who lack a family orientation to elaborated codes may, of course, be lucky enough to pick them up from the educational system (though Bernstein does not devote the kind of attention bestowed by Bourdieu to the problems and more subtle discriminations such persons face in trying to move up the educational cum social ladder), but such children are initially at a loss, they are strangers in what for others is not far from home, and there are strong pressures for them not to see the point, not to appreciate how far they are misunderstood, not to feel wanted in the school system.

I will now illustrate Bernstein's own accounts from different stages of his writings. The very first paper concludes with four and half pages on how working class public language and its attendant mode of perceiving and "structuring of receptivity conflicts with and induces a resistance to formal education" (1958 = 1974, p. 34). One point Bernstein makes is that the public, peer group language lacks the signs of deference required in speaking to teachers, and so may well be interpreted as rude; conversely the formal elaboration of feeling may come across as impersonal. As we have noted above, since the point of language use is different (between public and formal), working class children may see no need to extend their vocabulary in many of the ways endorsed by the school. The failure to attend to principles in public language is said to reflect itself in an inability to follow the principles in mathematical calculations or to make any sense of formulae such as are used in algebra.

Further Bernstein stresses the differences between the structuring of working class life and that of the school, with its slow plodding to distant and scarcely conceptualized ends. Formal language users find cues in the school to respond to, ways of enhancing their self-esteem; the public language user misses these cues, he is a fish out of water, and so finds little to be proud of in his achievements within the school. Noting the comparative failure of working class children in grammar schools, Bernstein notes that even when a working class home endorses middleclass values, its own modus operandi may put the same obstacles in the way of its children's appropriation of what the education system offers.

It is perhaps worth stressing that it is not just language behaviour that is involved in the way the codes help to shape educational outcomes, or perhaps rather we should not forget that language is part of a much larger whole. Thus Bernstein on one occasion stressed the "differential awareness of mothers to the educational functions of play and toys" and made the point that

Working class children often have to learn at school what is part of the experience of the middle-class child. Some middle-class mothers understand, even if they do not always approve, the classroom world of the infant school. Many working-class mothers are at a loss to see what it all means (Bernstein & Davies, 1969, p. 59).

In the more abstract terms of his later code theory, these points appear thus:

the school is necessarily concerned with the transmission and development of universalistic orders of meaning. The school is concerned with the making explicit and elaborating through language, principles and operations, as these apply to objects (science subjects) and persons (arts subjects). One child, through his socialization, is already sensitive to the symbolic orders of the school, whereas the second child is much less sensitive to the universalistic orders of the school.... The school is necessarily trying to develop in the child orders of relevance and relation as these apply to persons and objects, which are not initially the ones he spontaneously moves towards... Orientations towards meta-languages of control and innovation are not made available to these children as part of their initial socialization. (1969 = 1974, p. 196)

In a recent account, Bernstein links a middleclass mode of questioning, in which parents may test a child's ability to answer and from which a child may learn "the confidence to manage a social relationship where the presuppositions of everyday relationships are temporarily suspended" (1977b, p. xiii), to the expectations made almost with out thinking in the classroom.

Because of the correspondence between features of middle-class familial transmissions and formal educational transmissions, middle-class children have the means to appropriate such transmissions, and their experience is legitimated by the school, which in turn enables the child to legitimate the school, irrespective of the immediate relevance of what is to be acquired or the level of the child's performance (ibid., p. xiii).

1. The dependency relation translates into the motivation of the child with respect to school, irrespective of its immediate relevance or interest to the child/pupil.
2. The relative autonomous relation translates into an orientation towards the school's context-independent meanings, and to specific performance rules and texts these create (ibid., p. xiv).

I have quoted fairly extensively from Bernstein so that we can get a clear idea of the kind of claim he is making. As I have said already, I am not going to look at the empirical support for these claims, but I think it is worth exposing the prima facie irrelevance of the kind of work that is often adduced against Bernstein's explanation of educational failure.

It is crucial to the explanation that pupils need elaborated codes for educational success. So "researchers" went into classrooms and listened to what the teachers said. And they found that typical teachers spoke a pretty extreme form of restricted code (I leave aside the qualifications to that sort of statement that need to be made in terms of Bernstein's actual theory). So Strivens claims that one such researcher "suggested that it was highly misleading to associate the use of the elaborated code with requirements of the teacher/pupil roles. He demonstrated convincingly that the ordinary language of teachers in schools is much closer to restricted code use" (1980, p. 99). Musgrove (1979, pp. 49-51) appeals to similar evidence and makes the same points rather more caustically; but there seems to me to be here a perverse blindness to the issue Bernstein is raising.

We have noted already that Bernstein admits that not all schools devote their time to trying to transmit elaborated codes to all, or perhaps any, of their inmates; and it is important here to remember that much of the detailed observational, ethnographic work that has been done in English schools at least has focussed on the average or the lower than average pupils, those whom in many cases Bernstein admits society has given up trying to educate. But even without this possible bias in the evidential basis for these criticisms, the more important point is that children do not merely have to survive in schools, putting up with the gruff commands and physical assaults of laconic teachers; they have to read books, they have to write essays, take examinations, even, if they are lucky, follow educational films or interact with a personal computer. No doubt a researcher armed with a tape-recorder would fail to note the written words pupils were trying to deal with, but surely these written words would play a very significant role in such pupils' achievements, or lack of them. The idea that one reads for one's degree is not totally alien to the lower reaches of the school system. And indeed, there is some evidence that the place of literacy in the family home can play a significant role in the differential achievement of children in the earlier years of schooling, which are for many the most fraught with longer term consequences ("it seems possible that the relationship between family background and educational attainment in the early years may be quite largely mediated by class- associated differences in the relative salience that is given to activities associated with literacy in interaction between parents and children" [Wells, 1981, p. 197], reporting recent very detailed English work, with which one may compare Reid, 1976, for some suggestive data from Jamaica).

There are other criticisms that could be made of the relevance of such empirical research to Bernstein's claims, but my main point now is simply to highlight the crudity of what counts as significant work in this area, and as significant findings.

Code and Role

Before going back to further explorations of what the two codes might be like, I wish to examine one criticism of the way Bernstein has linked codes and roles at different times in his theoretical development. This will also allow us to note the ways in which in his later work he has utilized the ideas of classification and framing to supplant traditional role theory.

The criticism I am concerned with is one of many in a well-known article by Jackson (1974). Rather than deal explicitly with all of Jackson's points, I shall here tackle one of his major criticisms and will elsewhere discuss some of the others. In other cases the exposition of Bernstein's views should constitute an implicit reply to his critique.

Jackson claims that in the course of Bernstein's development he has switched around the supposed causal connection between role and code. In the previous exposition I have used the two quotations Jackson picks on: "The ability to switch codes controls the ability to switch roles" (1965 = 1974, p. 129) which Jackson paraphrases as "if you can't handle the code, you can't manage the role", and later "if you cannot manage the role, you can't produce the appropriate speech" (1971c = 1974, p. 177). Jackson goes on to say that the second version renders codes otiose. The appearance they might give of playing a causal role in the explanation of behaviour is due to mistaking tautologies for genuinely informative claims: "we would have to say: X prefers the imperative mode of control, he has learnt restricted code, and gives commands. In fact, of course, to prefer the imperative mode of control means to prefer giving commands" (1974, p. 80).

I shall first attempt to deal with the question of code and role in isolation from the causal structures postulated in the theory, and then turn to that question. Codes are largely Bernstein's private property, but roles are part of sociology's patrimony. This does not make them any easier to tie down, but perhaps the core of the notion is captured in the following rough definition of what counts as a role:

a description of a person allocates him to a social role or position in the measure that the attribution to him of some rights and/or duties is inseparable from the application of the description, unless that a man falls under the description follows from the mere fact that he is a man (Cohen, 1966, p. 56).

I shall treat roles as sets of normative expectations attached to social positions or what F. H. Bradley called in better English the duties of a station. Positions are the static and roles the dynamic aspect of a normative classification of social actors (Hollis, 1977, pp. 70-1).

In so far as these accounts are typical of what sociologists understand by roles, Bernstein's examples are striking in the degree to which the normative and stable components of such roles have evaporated. His own characterization of social roles displays this weakening and extension: "a social role ... is a constellation of shared, learned meanings through which individuals are able to enter stable, consistent and publicly recognized forms of interaction with others. A social role can then be considered as a complex coding activity controlling both the creation and organization of specific meanings and the conditions for their transmission and reception" (1971d = 1974, p. 144-145). In any event, Bernstein's example when he spoke of "if not code then not role" was of a man chatting up a woman at a cocktail party, which is certainly placed in a normatively constrained context but where the chatting up itself is only weakly a role. His examples for the "if not role then not code" phase are a discussion after seeing a film and children describing a picture to a stranger (one of the standard experimental situations in the SRU work), where the social positions are even less fixed and normatively determined.

Another significant feature of all these examples is that the actions expected of the role incumbent are primarily speech acts (I discuss this term in a later section, but for now it is clear enough - the main things expected in a discussion and even in an initial encounter at a cocktail party are words). It is this feature that can, I believe, help us avoid Jackson's strictures - for these 'roles', performing or playing the role simply is a matter of using the appropriate language (or almost wholly so). You don't have to be anyone in particular or to have any particular history, and in general you don't have to do anything else besides use the appropriate kinds of language.

But it might appear that this manoeuvre has hardly saved Bernstein's causal claim, since I have virtually identified knowing the code and knowing the role. But there is surely room for the earlier type of relation in that one can know, in general terms, what a role requires, what an incumbent is meant to do, without knowing how to go about doing it. And it is easy enough to be thrust into occupying such roles, whether or not one can perform in them. If the main kind of action required is linguistic then one can be said to be unable to perform in the role through ignorance of the appropriate language. (I know for instance that one is meant to express condolence to a dead person's relatives, but I don't know how to go about doing this.) One can occupy the role of mourner, willy nilly, but one may not be able to perform in the role, and at least a partial cause of this inability may be linguistic.

Can we similarly salvage Bernstein's later causal sequence? I take it Bernstein is saying something to the effect that if a child cannot easily see himself in the position of explainer to an adult stranger then he is not likely to sustain the sort of language appropriate, by that stranger's norms, to such a relation. Roles have got to be recognized by their incumbents. This seems a perfectly possible account, as indeed Jackson recognizes. But why should it bethought to be in conflict with the previous version? I know what is required of a person explaining the details of a salary negotiation to fellow employees, but when the same querulous question is asked for the fifth or fiftieth time I tend to find my annoyance preventing me from managing the role of union delegate. We need to explain, on some occasions, why a person occupying a role is unable to perform at all in the appropriate way (and this may sometimes be because they don't know how to use language appropriately), and on other occasions, we need to explain why a person who knows how to use the language or more generally knows how to perform in a certain role nevertheless fails to sustain the performance(and this may be because they are unable to see themselves in that role in that context). If Bernstein ever thought that there was only one direction of causation, from code to role in his earlier work or from role to code in his later, he would then be wrong; but we need not suppose him ever to have been so committed. Clearly the focus of interest has changed from cases (like my mourner case) where a person simply does not know the "right thing to say" for a particular role, and thus can occupy but not play that role adequately, to cases (forced upon him perhaps by various critics) where the person does know the right sort of thing to say but seems unable to produce it on what are, by some observer's lights, the appropriate occasions.

Having argued that Jackson's criticism here is not a problem for Bernstein's account, it is worth noting that Bernstein has now attempted to restructure his appeal to roles by using his concepts of classification and framing. "We have found that it is possible to transform positional and personal family types into the language of classification and frames" (1975 = 1977a, p. 9). This is but one more example of the way in which he tries to construct his own categories in preference to using the ready-made categories that the standard sociological literature offers.

Given the earlier sketch of classification and framing as set out in their original educational context, as focussing on the strengths of different boundaries, and thus on what may go with what and what may not, and given also the characterization of roles as typically normatively oriented - you may do this but not that - it is perhaps not too difficult to see how the characterization of a role could be translated into the language of classification and framing. We can see such a translation in the following remark from one of Bernstein's own papers: "strong linkages (positional families) are families which create rather strong boundaries, whereas person-centred families tend to create weak or blurred boundaries" (1973c = 1974, p. 249).

One advantage that has been claimed for such translations is that they allow for much greater delicacy of characterization than typical talk of roles. Thus Walford, after reviewing a variety of relationships between students and supervisors in the comparatively narrow domain of postgraduate physics, says that:

it is clear from these few examples, where it might be thought that the similarities of research experience would be more evident than the differences, that the rather simple concept of supervisor's role is an inadequate base for understanding the complex relationship between student and supervisor.... The concepts of classification and framing will thus probably be useful to help clarify some of these tensions and to highlight two dimensions which have a part to play in the successes and problems that occur in supervision (1981, pp. 156-157).

Implicit/Explicit and its Metamorphoses

Certain continua, certain contrasts, have permeated Bernstein's thinking from the very beginning- content/form; implicit/explicit; concrete/abstract; context dependent/context independent; particularistic/universalistic; unspecific/specific. In each of my examples, restricted codes have been associated with the first feature, elaborated codes with the second, although in some cases the dimensions are to some extent independent. But these are all ways in which Bernstein has sought to articulate the fundamental intuition behind the codes. They are also mostly comparatively common sense notions. One of the main points of this study is to demonstrate by example that one cannot afford to leave crucial explanatory notions in this state. I shall be borrowing ideas,

mostly from recent work in philosophical logic, to show that real explanations require these more detailed and specific accounts. If I am successful, it will serve to support the programmatic claim announced at the outset that explanatory sociological theory will require more cross fertilization from outside and a greater willingness to respect the distinctions made in other disciplines.

While all of the contrasts mentioned might be worth careful examination, I shall be concentrating on two of them - in this section, implicit/explicit, and in a later section, particularistic/universalistic. These are perhaps the two most stressed by Bernstein himself, but I focus on them also because I wish to offer glosses on them that are not, I believe, a grave distortion of Bernstein's intention though they do perhaps suggest alternative directions for investigation. (I doubt that one should offer so sympathetic a gloss on the concrete/abstract contrast, and I have nothing to add to the voluminous literature on content versus form; the other contrasts are virtually synonymous with the pair to be discussed.)

We saw earlier that in the first papers on a public language the fundamental notion was that of implicitness; public language use was one that leftmost things unsaid. In describing the steps towards his later views, Bernstein says that making a distinction between relatively context dependent and relatively context independent meanings was a return "to the basic ideas of implicitness and explicitness and to the central idea that the form of the social relationship acts selectively upon the meanings to be realized, which in turn activate specific grammatical and lexical choices" (1971b = 1974, p. 14). A recent formulation is in terms of a distinction between

forms of speech where the referents of the speech are explicit in the text, and forms of speech where the referents are not in the text, but in the context. In the latter case, unless the listener has access to critical features of the context in which the speech is embedded, the meanings are not clear (ibid., p. 13)

Both Bernstein's odd examples and the emphases in the SRU monographs which are replete with counts of exophoric versus anaphoric usages would suggest that this sort of implicitness is central to Bernstein's thought.⁶ As Hawkins notes (1977, p. 204), it is a lot easier to use exophoric items than to hunt around for le mot juste to express one's meaning, so children speaking a more context independent language are doing something rather unexpected. (Gellner has recently suggested an even larger wonder:

it is a very puzzling fact that an institution, namely human language, should have this potential for being used as an 'elaborate [sic] code'... as a formal and fairly context-free instrument, given that it had evolved in a milieu which in no way called for this development, and did not selectively favour it if it manifested itself [1983, p. 33 ]

While Bernstein's account of what is going on may confuse, as most non-philosophers do, questions of reference with questions of meaning (you know as well as the mother on the tape what Stop that immediately! means; what those of us outside the context of utterance do not know is to what action that referred), his point can be easily made, as I just have. But while it is acceptable as it stands, I think there are further ramifications of implicitness that would be worth pursuing.

Bernstein's favorite exophorics are what some logicians have called "token reflexives", because what they refer to is given by knowing the particular token of the word in question - if I said It's very hot today on September 5th, 1983, then that token of the word type today referred to September 5th, 1983. If we heard it on a tape, we would have to know which day the utterance occurred. Token reflexives often require further help from the audience; our supposed mother with her imperative requires us to direct our attention to likely activities in her neighbourhood, that relates something to her, but does not pinpoint the activity in question.

The main point I would make is that while token reflexives are important parts of language they are by no means the only mechanism in language that encourages implicitness. All over language there are devices that allow us, if we wish, to leave out things we could perhaps have said, and which are in fact necessary parts of our message, and yet still to bespeaking perfectly grammatically. The term I shall use for this phenomenon is "ellipsis". It can arise from the operation of several different syntactic rules, but, as with Bernstein's codes, its importance is semantic; bits of meaning are left unsaid. Bernstein and his associates argue that the restricted code tendency to greater use of exophoric items puts obstacles in the way of pupils in schools, where explicitness is at a premium. I would argue that if this is so it is all the more so when we consider some of the other ways in which things can be left unspecified and so, in many cases, unthought of.

The fundamental and original notion of ellipsis is of parts of a sentence that have been left out. While traditionally the notion has often been criticized because of the arbitrary ways in which theorists have claimed that something has in fact been left out of the surface structure of a sentence, I think there is a use of the notion to be found among some philosophers that largely avoids this arbitrariness and which yields an important tool for logical analysis. (I have set out some of the history and the philosophical employment of this notion of ellipsis in my paper, 'Ellipsis: History and Prospects', to appear b.) The crucial aspect of this more specialized notion is that something has been omitted from a sentence which is vital for the sentence to be conveying a determinate sense and thus to be saying anything true or false. Thus, to take a topical example, we have recently been told in Jamaica that it is a breach of national etiquette to fly any other flag to the right of the Jamaican flag. But now, what does that actually mean? Does it mean, to the right of the Jamaican flag as you sit on your verandah, or does it mean, to the right as viewed from the road? No doubt I am betraying my ignorance of good form by even asking such a question, but that perhaps makes my point - we have a way of talking which encourages us to omit mention of a vital part of the message. It is an important part of Bernstein's explanation of educational failure that schools trade on unspoken messages; as Bourdieu describes what he calls traditional education, it leaves unsaid the principles you need to know to succeed - if you are the right sort of chap (or otherwise one of the "gifted" few) you will know already; if you are not, you will not be told.

Now while left and right may not usually lead to any serious pedagogical consequences (though I have no evidence, and it might be better not to suppose so) I think there are other cases which are educationally more interesting.

Besides the simple sort of ellipsis we have just examined, there are also transformations in language whose effect is to drop out information, and which thus are similar to simple ellipsis. They are similar also in that this information can be restored, although sometimes at the cost of a certain inelegance. To give an example, consider the nominalizing transformation that takes us from sentences such as Those clouds look dark and heavy to noun phrases such as the look of those clouds. Or another nominalization that takes us from the simply elliptical Johnny needs to learn to read (this is elliptical since it requires supplementation with something that he needs it for in order to make a determinately true or false claim, cf. my 1980) to talk of Johnny's needs. In both cases we could restore the information lost, the dark and heavy look of yonder clouds or Johnny's need to learn to read in order to get a decent job when he grows up, but not perhaps with much stylistic grace.

We are glad to talk of the look of clouds because it is often difficult to find a reasonable description; there are many areas of life for which our descriptive vocabulary is utterly inadequate. We may prefer to talk elliptically about needs because the issues left unmentioned are controversial (why does Johnny need to read in fact?). This last is, I think, among the most important cases for sociological attention to ellipsis since it sometimes generates a false appearance of consensus, as in the example just given or it can be used to try to impose a particular set of values by appeal to what appear, and properly understood are, factual matters about what someone or some group needs. So our motivations to go in for such ellipses and similar content reducing linguistic options may well vary from case to case, and so their pedagogical importance may also vary. In some cases, we may have examples such as Bernstein himself illustrates of a linguistic way of avoiding the greater explicitness demanded in many school subjects (Bernstein mentions the preoccupation in some subjects, such as literature, to find ways of replacing the ubiquitous that's nice or I don't like that of everyday life). But in others we may have amore interesting phenomenon, a potent transmitter of ideology. I shall expand on this suggestion in the next section, but for now we can note one aspect that is likely to be of some importance at least within the more advanced sections of an educational system. An openness to the sorts of ellipsis found in talk of needs, which can also be uncovered in our talk about freedom, responsibility, and equality, to name but three, allows for a much more nuanced judgment, for a discrimination of respects, for the "distinguishing" that is one mark of the quick-witted lawyer. As the cognitive demands of schooling call for greater subtlety of thought anyone commanding an elaborated code with respect to this sort of ellipsis ought to be signally advantaged.

Before moving on to look at some further aspects of ellipsis which tend to undermine the cognitive effectiveness of elaborated as well as restricted codes, it is perhaps worth making one small point about ellipsis to counter frequent misunderstanding. The notion I have been using is not identical with the idea simply of things that have been omitted. Everything we say leaves out some things that could have been said, so a notion of ellipsis that was as wide as that would be useless. The point of my notion can be seen if we contrast two sentences: John is walking and John needs to walk more. Both leave out a vast amount; the first does not tell us how fast, or with what gait John is walking, nor an unlimited number of other things that could be said about his perambulation. But what little the sentence does convey is sufficient for us to have something true or false (when we interpret the token reflexivity of the tenses and when we have tied down the proper name). The second sentence, however, still leaves us unable to begin investigating its truth or falsehood even when we have similarly tracked down the time and the person it is about; we must supply or be supplied with some goal or end or state with respect to which John's increase in walking can be evaluated.

Ellipsis and Ideology⁷

The distinction between restricted and elaborated codes is a matter of degree. Bernstein's focus is on the differential consequences he supposes to follow from positions at different points on this continuum, or these continua. Just as the pins in a pinball machine magnify and dichotomize small differences in initial inputs, so schools magnify and dichotomize (to oversimplify somewhat) small differences in cognitive and attitudinal inputs, so that the output is, more or less, in accord with prior familial positions. Looking at things from such a perspective it is easy to give the impression that while restricted code imposes limitations and liabilities, elaborated codes are as good as they could be. This is not of course Bernstein's view, but there is little mention of the shortcomings of normal elaborated codes while the emphasis is on the opportunities for success they open up for their possessors.

But if the dominant groups dominate in an elaborated code their thinking is no less distorted and misled by ideology than any one else's.⁸ What is it about elaborated codes that permits such confusion in the minds of its users? A tendency to greater use of anaphoric expressions would seem to be neither here nor there. At the risk of offering an explanatory panacea, I suggest that the kind of semantic ellipsis described in the previous section is well suited to perform this task and to do so in ways that preserve a modicum of good sense in the operations of the ideologically slanted mind.

We have noted already that students of society have not in general been overmuch concerned with processes of transmission or reproduction. This is equally true of those many who have concentrated on ideology. In a fairly recent review of the literature Kellner remarked that "the sociology of knowledge ...showed astoundingly little interest in any detailed study of the structure and mechanisms of social communication" (1974, p. 325), and Garland's forthright claim, "there are two central questions one can ask of ideology: what are its origins?; what are its effects?" (1981, p. 128), suggests that most sociologists are not likely to focus on the question of how ideological beliefs are transmitted among comparatively sensible people.

But questions of transmission are, however, of some interest, not only in their own right, but also, in the context of social explanation, because the easy reproduction of ideology seems to clash with the rationality assumptions built into such explanations, and with the perhaps more extensive rationality assumptions many theorists would like to make. The whole question of the degree and kind of rationality implicit in social explanation is too large and controversial to tackle here, but I think it would be generally agreed that we assume other people share with us a physical environment, some of whose ways of working they grasp as well as we do, and that we assume they act, in general, to achieve their purposes. However weird and wonderful some of their ideas and practices may be, there must be some such core (not necessarily the same in every case) of shared world and minimal rationality. This further implies that people can, in some measure, distinguish truths from falsehoods and better from worse arguments or reasons. But of course it does not mean that they must be able to pontificate in the second-order style of logical appraisal, although in fact many people can do so. More needs to be said to square these remarks with the truths contained in talk of "theory-laden" observations and elsewhere that beckon the unwary into the quagmire of a general relativism, but this is not the place to say it. All I need now is agreement that we must attribute some minimal good sense and rationality to our fellows in order to understand them at all. How much more we attribute seems as much a matter of temperament as of empirical evidence – it may be worth noting that good sense in one area is compatible with utter incredulity in another - but the problem I am currently addressing arises the more insistently the more rationality one accords people and is probably only worth discussing if one thinks that people are fairly sensible.

Outside of the religious area, or rather the theological portion thereof, beliefs are within reach of evidence, proposals can be seen to have consequences. So falsehood, gross distortion, and gross partiality ought not to survive, especially when their contents have a central importance in people's lives. But error, distortion, and partiality are precisely the raison d'être of secular (and of course also religious) ideologies. There is a strong temptation to see people who wholeheartedly accept such ideologies as radically incapacitated, dupes of a false consciousness from which they can only be awakened, if at all, by the stern voice of a miraculously clear-sighted authority. But if it can be shown that the beliefs in question are held in accordance with general principles of evidence and inference then we can perhaps hope that a man who cannot, in the miasma of ideology, tell a hawk from a handsaw, could yet do so when the hidden has been made visible, and do so on the basis of intellectual skills he already possesses.

It would take us too far afield to substantiate in detail the possibility just mentioned that the patterns of reasoning that support ideologically distorted beliefs can be given a clean bill of cognitive health (for moves in such a direction see my paper referred to in footnote 7 to this chapter). But I think one key feature of such an attempt would be the kind of semantic ellipsis we have already mentioned and the fact that it permeates many, if not most, of the concepts we use for thinking seriously about our social and political life. Since these terms can be used with no grammatical impropriety while leaving unsaid crucial elements of a determinate sense, it is little wonder that people can be swayed by forms of argument that are formally quite respectable, but which when applied to such elliptical claims can easily lead to equivocation and fallacy. The modes of reasoning can then be quite satisfactory while allowing the transmission of ideological distortion without attention being drawn to its biases.

These programmatic claims can be given a certain amount of support by looking at some typical accounts of what ideology involves. Bernstein has not written very much about ideology as such (though he has had a lot to say about particular ideologies such as the child-centred approach to schooling), but we can start with an author whose work is typical of a lot of recent sociology of education, Michael Apple. In a collection of essays published under the title Ideology and curriculum (1979) he notes at one stage that hegemony requires fundamental agreements at a tacit level, which are transmitted by his version of the "hidden curriculum":

The controversies usually exhibited in schools concern choices within the parameters of implicitly held rules of activity. Little attempt is made to focus on the parameters themselves. The hidden curriculum in schools serves to reinforce basic rules surrounding the nature of conflict and its uses. It posits a network of assumptions that, when internalized by students, establishes the boundaries of legitimacy. This process is accomplished ... by nearly the total absence of instances showing the importance of intellectual and normative conflict in subject areas. The fact is that these assumptions are obligatory for the students, since at no time are the assumptions articulated or questioned (p. 87).

In several other places Apple notices the prevalence of tacit assumptions (ibid., p. 34, p. 83, p. 125f) and our resulting tendency not to question the framework of debate (a topic we shall take up again in a subsequent section) but he does not tell us how the cognitive resources we use aid and abet these restrictions. Semantic ellipsis surely helps. It contributes to an explanation of his claim that: "The orientations which so predominate curriculum and educational theory ... effectively obscure and often deny the profound ethical and economic issues educators face" (ibid., p. 149). It also helps to explain, and make consistent with the stress on the tacit nature of ideology, another point Apple makes: ideological rhetoric is fairly explicit and systematic about what can be agreed upon (ibid., p. 21f). Ellipsis allows an audience to find their own particular supporting instances of a claim but without ever having to face the question whether my way of accepting it is consistent with yours. Semantic ellipsis is yet one more mechanism whereby when "claims are vacuous, vague, imprecise, ambiguous and generally unclear, conflicts are likely to be minimized and go unnoticed" (Naish, Hartnett, and Finlayson, 1976, p. 99).

One other point Apple makes is that ideological categories tend to be "essentializing" (ibid., p. 135). It would seem possible to link this to the way in which ellipsis can encourage false belief in absolutes when the realities are relative. Apple's examples again, slow learner, discipline problem, may not all involve ellipsis, but it is worth noting that a slow learner is presumably a slow learner of A and B but not necessarily of everything the school might have on offer.

Another writer who has focussed on the transmission of ideology is Trevor Pateman (1980). One of his more controversial suggestions is that many people simply lack the conceptual structure required to handle political or social debate (ibid., ch. V). One might hope to side step the general question of who has what conceptual resources by noting that many of the interrelated terms in question (freedom; responsibility; authority;...) are usually used elliptically. The result is a very diffuse conceptual foam, in which these terms connect with each other but with many gaps. It is not surprising that people easily get lost and find it hard to hold on to occasional insights. It is noticeable for instance how much contention the general question of a supposed conflict between equality and liberty is capable of generating among people who are not obviously Pateman's incapacitated workers or children when in fact there are only a range of determinate questions about the relation of equality in this respect to certain people's freedom to do certain things (and often also the existence of various kinds of obstacle to these actions).

The fluidity of the standard terms of political debate is a constant theme in Nigel Harris' discussion of ideology (1971). His metaphors and comments continually invite a gloss in terms of semantic ellipsis. Thus in talking of the move from equality to equality of opportunity he says:

Very little remains of a concept thus subject to the acid of political debate in which at least one side finds it useful to redefine the other side's basic demands so they become unobjectionable. What does remain is a blur, in which anyone can identify with the concept on nearly any grounds, even though the grounds on which two people identify with the same concept are mutually contradictory (ibid., p. 24).

Again, in discussing the intellectual sin of taking remarks out of context, Harris adverts to the very feature that encourages both grammatical ellipsis at one level (leaving out parts of a sentence as in a reply to a question) and ideologically powerful semantic ellipsis at another:

The logic of debate disciplines, reshapes and, indeed, creates the positions of an ideology, even though the essence of those positions derives from a given social situation. That social situation is also crucial for understanding the ideology, for people do not at all have the same perception of events: a flat disc may appear circular from one position, but a thin strip from another (ibid., p. 54).

These examples merely show that semantic ellipsis could be a useful adjunct and extension to typical sociological accounts of the transmission of ideology among otherwise intelligent people. It provides a mechanism to do the trick. No doubt there are many other factors involved - no more than Bernstein am I claiming to give an exhaustive explanation - but I think these examples should support the claim that it is one mechanism.

In the previous section I introduced semantic ellipsis as another area in which the possessor of an elaborated code could shine - by recognizing some of the ellipses and making them explicit - in this section the emphasis is on the other side of the coin, which is, I suspect, socially the more important -that typically such recognition is a very temporary affair. The simple elliptical phraseology of needs or freedom or equality seems almost to exert a tremendous pressure to insulate people from their occasional insights that a need is a need only with respect to certain ends and so on - that insight may be made explicit at one moment but the discussion continues as if it had never surfaced. And of course, when the insight does not surface, the elliptical language works very smoothly in concealing the controversial values that drive most such discussions.

While we have gone some way away from Bernstein's obvious concerns, I hope that this discussion has shown a way in which a possible gloss on his work can perhaps give some added depth to his own speculations, by moving us into a somewhat more intellectually demanding area than his own exophorics, and has also shown a way in which this notion borrowed from a tradition of philosophical logic can be of use within sociological explanation.

Particularistic/Universalistic

We have already examined the basic ideas behind this contrast and its rather different aspects. The terms are applied on the one hand to the speech models for a code and on the other to the kinds of meaning embodied in language. One can see both aspects as, crudely, matters of numbers: if a code has particularistic speech models, it can only be learned from a restricted number of people; if its meanings are particularistic it can only be fully understood by a restricted number of people. What actually occurs will reflect the different role relationships of the people using the language. But while Bernstein certainly stresses these matters and the close connection with the question of context dependence or independence that we looked at in a preceding section, perhaps the distinctive aspect of the contrast, and the one that has surely the most bite in the explanation of educational failure, is the cognitive:

Where meanings are context-independent and so universalistic, then principles may be made verbally explicit and elaborated, whereas where meanings are context-dependent and so particularistic, principles will be relatively implicit, or, as in regulative contexts, simply announced (1971b = 1974, p. 14).

While there is little to query in the link between exophoric and limited audience, anaphoric and wider audience, it does seem somewhat risky to assume a further link between exophoric (context dependent)and absence of principles, explanations, etc. and between context independence and the presence of such things. The restricted code of astrophysicists is surely unintelligible to most of us, but one presumes its content is highly general explanatory theory. Names are more context-dependent than descriptions(compare Edward Seaga with the present prime minister of Jamaica) but we have any number of such names for general principles themselves (Pythagoras' theorem, the Marrison v. Bell presumption, etc). This is just one instance of the general problem of how far the semantic content of what is said constrains or is constrained by its syntax. Bernstein's general approach has been to look for fairly strong correspondences here. But while Bernstein can definitely point to some sort of patterning indifferent people's speech which reflects different orientations to communication, the general point I would make, which is I think likely to be endorsed by Bernstein's group,⁹ is that one should be looking for semantic patternings as directly as possible, since it is all too easy to go astray by presuming a closer connection between syntactic structure and semantic notions than actually obtains. Somewhat greater sensitivity to logic or semantics might well be in order than the research procedures adopted by the SRU actually afforded.

The comparatively hazy grasp of these issues by Bernstein's researchers can be illustrated by a couple of their discussions. Firstly, Hawkins claims that there is a code difference in the use of indefinite pronouns, especially someone and something, and he goes on to say that "these pronouns are substitute nouns and can be seen as a means of expressing uncertainty" (1977, p. 184). While he suggests a qualification to this claim, it is open to question as it stands. To say Someone's at the door may well be to convey uncertainty as to the identity of the person at the door, but it is perhaps worth remembering that in the formalism of modern logic such an existentially quantified sentence is somewhat more complicated than ordinary subject predicate sentences such as John is at the door. And of course, often one simply is uncertain. Presumably Hawkins is concerned more with those cases where the indefinite pronouns are used despite the fact that a more specific designation ought to be available to the speaker; but these situations are not countable on the basis of the syntax alone. Another point that can be made is that some such sentences (like those offered by Adlam in a discussion I shall be looking at next, e.g. Someone's got to hide in a good place) can in fact be performing a cognitively different role - here something like Take anyone you like, let her hide in a good place. Hawkins' results may then be lumping together not only a possible tendency to avoid specification but also a mastery of other and possibly more complex logical operations.

The second example of the confusions of the SRU researchers is afforded by a lengthy discussion in Adlam (1977, pp. 13-6) of the relations between context dependency and universalistic/particularistic meanings. She takes some sample sentences (though she recognizes that code categorizations can only properly be made at the level of the whole text) and tries to explain their categorization in her schema. Thus Someone's got to hide in a good place is universalistic but no more explicit than the particularistic Little girl goes and hides in the cupboard. (For my own part, I am not sure why this second sentence should not be treated logically on a par with the first: take any little girl, let her etc. There is nothing in Adlam's text to suggest that the speaker was describing one actual game of hide-and-seek in which a particular little girl had gone off to hide. Though it must be admitted that in the cupboard is a lapse into unwanted particularity.) She then claims that the following pair cannot be described in terms of universalistic/particularistic, though the first is implicit and the second explicit: He's doing that, The guard man's pushing a luggage cart. Adlam's account commits her, however, to saying that He's doing that must be particularistic. But it might be functioning as an explanation, or doing any of the other universalistic cognitive jobs. Her argument here is that if we take the first pair of sentences and transform them, in a regulative context, to implicit form, they both become She's got to hide there:

making the realization implicit, making an utterance implicit at the level of lexes and grammar, automatically renders the meaning more context-dependent. This would seem to imply that context-independent meaning must be realized explicitly. But the converse does not hold - context-dependent or particularistic meanings can be realized explicitly or implicitly (1977, p. 14).

While the syntactically based patternings of speech, predicted on Bernstein's approach as set out above, can be found, my inclination is to try to separate the two dimensions(particularistic/universalistic and implicit/explicit)in a more clear cut manner. I have already accepted and tried to extend the implicit/explicit distinction in terms of explicitly stated semantic content, which includes exophoric versus anaphoric usage as a component. For the particularistic/universalistic contrast I think we should concentrate on the cognitive issues that Bernstein has often invoked, and which are surely the most potent in the educational relevance of the distinction. We have already seen the association of elaborated codes with "general social means and ends" while the restricted code is concerned with local matters, the degree of restriction being a "function of the parochialness of the social means and ends" (1962a = 1974, p. 79); or again elaborated codes with principles and explanations in school subjects, or with a mother who wishes "to elaborate for the child certain rules and the reasons for those rules and their consequences" (1969 = 1974, p. 195). What is crucial to universalistic meanings is the making explicit of principles, explanations, justifications, causal connections; particularistic language leaves all this unsaid.

I have borrowed one notion, that of ellipsis, from philosophy. In this context I suggest we can profitably borrow another, and much more prominent notion, that of speech acts (introduced by Austin,1962, discussed by many, including Searle, 1969, and Graham, 1977). The basic idea here is that using language is not just uttering noises or making marks; such behaviour is always a doing (or often many doings) of some other action - warning, explaining, diverting attention, putting someone down, promising, etc., etc. And like other intentional actions these occur typically in a complex context, with what we can regard as different levels. The original intuition behind Bernstein's codes was that people do not all use language for the same sorts of purpose, which is to say that their routine dispositions to speech acts are different. One person may be disposed to use language to give explanations or justifications, another to sustain personal relationships. The fundamental difference is at the level of action, and we have already seen that it is none too easy to seek for it through non-semantic aspects of the language used - though some differences there may well show up.

While for personal reasons I find talk of speech acts fairly congenial, others have offered glosses on Bernstein that amount, I think, to pretty much the same thing. Hasan (1973, p. 283), for instance, gives a tabular account of code in which my alternative sets of preferred speech acts appear as an alternative focus of interest: on practice or on underlying principle (cf. also Fletcher's gloss in terms of "perceptual and conceptual and ideational knowledge" [1984, p. 277]). Perhaps closer to my speech acts is Hawkins' suggestion that we should think in terms of alternative speech strategies: "given a particular speech function, or context, different speakers, by virtue of differences in their social origins, or experience of role-relationships, etc., may employ different strategies of communication" (1977, p. 195). While Hawkins goes on to illustrate his idea by reference to possible alternative modes of answering the telephone, and links his idea to the counts of exophoric and anaphoric usages, he also suggests that a preference for "appeals" or "questions" rather than imperatives or threats in expressing social control verbally would also count as an example of alternative strategies. Exercising control by threatening rather than by appeals to conscience is clearly a case of using different sorts of speech act in pursuit of the same end.

This example also reveals one of the reasons I think some sociologists might be inclined to dismiss the subtleties of speech act theory. As was mentioned in the earlier characterization, speech acts, like any other acts, are performed in a typically complex context and can be correctly described in wider or narrower terms or seen as occupying higher or lower levels. To borrow an example from Pateman (1980), a sergeant who shouts Get your hair cut is doing the same thing, at one level of analysis, as one who instead says Who do you think you are? Mozart?; but at a subordinate level they are doing different things. If one's primary concern is with the level at which these two remarks are functionally identical then one may well regard the insistence that they do differ as an irrelevance. But such differences may well have consequences elsewhere; and certainly it is central to Bernstein's view that the differing cognitive demands of different linguistic means of social control do have important spin offs. The apparatus of speech acts helps, I think, in allowing us to grasp the complexity here, while still allowing us to recognize levels of functional identity as well (cf. here Jackson's criticisms, 1974, p. 74).

It is perhaps worth noting here that if the approaches here and in previous sections are on the right line, as interpretations of the codes, and fruitful as tools for future work, they indicate the extreme limitations of a view proffered by Brown and Levinson that "much of the difference between Bernstein's ... codes can be assigned to negative-politeness versus positive-politeness preferences in linguistic expression" (1978, p. 251). But while there may be a lot more to the codes than politeness strategies even as Brown and Levinson understand them (though as we have noted Bernstein is aware that they do play a role, 1958 = 1974, p. 34), these authors may give some support to the present suggestion since they argue powerfully for an approach to the sociological study of language use focussing on strategies to satisfy "face" wants - the aspect of language up for sociological explanation is then the rational actions it embodies, that is to say speech acts.

The Yet To Be Thought

In the previous section I have argued that much of what has gone on under the umbrella of a contrast between particularistic and universalistic orders of meaning has been misconceived. Instead it would have been better, I suggested, to focus on one aspect that Bernstein certainly invokes but which the empirical work did not adequately grasp: the use, or rather the differential use, of explanations or justifications, of appeals to general principles.

In this section I shall attempt to show that such a focus could fruitfully incorporate into Bernstein's schema an account of the middle class nature of much radical thought and protest action. Whereas my suggestions with respect to ellipsis might be moving in directions Bernstein had not foreseen, I think the present section is very much in line with some of the things he has recently been saying, in to me very obscure fashion, about the fact that elaborated codes give one access to "the yet to be voiced". In a note he says, for instance, that

elaborated orientations ... are ... always subject to strong regulation and surveillance; for these orientations have the potential of creating alternative realities, possibilities, and practices. Elaborated orientations are potentially dangerous, and those acquiring them have to be made safe (1981, p. 355)

Even if this is not a correct interpretation of Bernstein's thought, it is not in itself implausible. Many people have noticed that political or social visionaries and revolutionaries have very often come from a social group somewhat better placed than the poor and oppressed they seek to help, and in the aftermath of the student protests of the later 1960's ideas of middle class radicalism have become commonplace. Of course this does not mean that the poor and oppressed never themselves articulate their position or a framework for change, but it might suggest the fruitfulness of following Bernstein's suggestion that "it is a matter of some importance to distinguish between the reactions to, or the challenge of, the realization rule imposed by a given framing and the level we have called the 'yet to be realized'" (ibid., p. 346). He is here contrasting a mode of rebellion that stays within the terms of the status quo and his deeper mode of resistance that confronts that status quo with new possibilities. No doubt it is not a clear cut distinction, nor one that can be easily made in all cases, but one supposes that utopias and some forms of isolationist communities might exemplify Bernstein's deeper kind of resistance.

If we stress the fact that elaborated codes give access to principles and justifications, we may then have part of an explanation for educated commitment to utopian or morally based radicalism. This explanation requires us to note that morality is a pretty mixed up affair. We need to distinguish at least between morality as thought about and morality as lived (a distinction that is itself part of common sense reflection on everyday life - Practice what you preach). Once distinguished, we should note some important differences between them. In particular, and at least in our cultures, morality as thought about is in general a very high-minded, universalistic affair; or at least such is the ideal type of moral reflection to which educated people aspire. (Kohlberg has devoted much time and effort to following people, mainly in the USA, through what he conceives of as the different stages of moral thinking, and he certainly sees them as moving upwards from a crude, external view to the sublime heights of the Constitution or of universal justice. But it is worth noting that in his experiments in which people reflect on fairly concrete situations, he is very hard put to it to find anyone actually using either of the highest levels in his hierarchy.)

When we turn to morality as lived we can see that the very general precepts and principles that are the currency of moral reflection and moral suasion are very rarely put into practice. Those few that try are picked out as saints or lunatics. The usual best that we find in practice can be labelled (as by Broad, 1942) "self-referential altruism" - one treats one's family, friends, or even compatriots rather better than one ever thinks to treat other people's friends, family or tradesmen. The distinctions such moral practice recognizes are deeply embedded in the contingent social situation. Actual morality as practised is largely still positional or ascriptive, rather than universalistic as moralizers or planners like to suppose. It is a kind of restricted code, in Bernstein's terms; but when we talk about it, we usually presume it is a comparatively simple, general system of principles.

I have elsewhere argued (1979) that this schizophrenia in our moral life constitutes a serious intellectual problem since the principles we typically use to justify our practice or our proposals cannot in fact reach down to the socially contingent detail of that practice. (This point has been known for a long time - Godwin shocked his contemporaries by arguing that as a utilitarian he ought to rescue Archbishop Fénelon from a blazing building in preference to his own mother, since the Archbishop was likely to contribute more to the sum of human happiness. Less ruthless utilitarians can of course find subtle objections to such arguments, but I think they were better admitting that what we want to live by is something too self-referential to be generated by universal principles.)

One important way in which this can be seen is in the fact that in practice, if not always in philosophical theory, moral thought remains inside a framework of unquestioned assumptions. Within such socially given parameters, moral reflection is expected to play its part, but it is not expected to turn its attention to the parameters themselves. And for the very good reason, at least in the eyes of those benefitting from things as they are, that those parameters could not be underwritten by the kinds of moral principle people apply within them. As I have written elsewhere:

it is a feature of children's - and philosopher's - discussions that the principles are given much more of the authority they claim, and are turned upon the contingent structuring of the field itself. So it was obvious to Plato, as it is to some children, that boys and girls should exercise together, that family riches should not be allowed to interfere with their education, and that both should be eligible for the highest responsibilities (1982, p. 29).

Moral thinking also insinuates an image of its own authority as objectively given; as ordinary language moral philosophers do not tire of telling us X is morally good does not mean I approve of X. As Mackie argued (1977), everyday moral language intends to convey an objective prescriptive force for its injunctions or judgments; metaphysical argument, however, suggests that this presumption is in error.

Putting these facts together, we find that everyday practice cannot be justified by the normal principles we use; those principles, however, are the source of objective support - they and they alone can underwrite our morally acceptable practice. If that is the situation you face, it is not surprising that you should seek to manipulate reality, everyday practice, to fit. But the manipulation cannot be a matter of minor adjustment - the principles make no distinction of Jew or Gentile, man or woman, slave or free - so one seems forced into utopian or extreme radical directions.

There has been a great deal of speculation and research on the student movement that culminated in May 1968 in Paris and in the anti Viet Nam War movement in the USA. I cannot go into this literature here, but it is perhaps worth noting that the suggestion that it involved an exercise of moral reflection on the standard practice of political agents is fairly well supported. Thus Hirst (1973),for instance, discussing the European experiences, argued for the centrality of a perception of hypocrisy, a mismatch between the ideals espoused in pulpits and public platforms and the somewhat less edifying reality of social and political life. And in a discussion of the American scene, Miles remarked that "politically, these students do not rebel against their parents' values but extend and apply them" (1974, p. 442). Both these writers are, of course, concerned to explain further why particular groups of students should have reacted in these ways – after all, most educated people get to think at least a little about their moral and social beliefs without espousing either Plato's or Tariq Ali's view of a better world - but for my present purposes these more detailed questions are not central. The point, simply, is to show that given the state of morality in thought and action, exposure to elaborated codes as purveyors of principled argument and justification can lead to radical rejection of the status quo or to moralized reformism. The transmission of such codes constitutes then a standing threat to the social order and so, as Bernstein says, requires careful attention to making its possessors safe.

This context might be the place to try also to make use of a contrast Bernstein has introduced, though with little explanation or exemplification, between two modes of elaborated codes.¹⁰ The distinction is phrased in terms of individual versus personal orientations, and the latter are associated with the 'new' middle classes whose place depends on the possession of the means of symbolic control rather than on the possession of more traditional forms of capital. My reason for introducing the issue here is as a partial contribution to the question noted above of why some people and not others get bitten by the discrepancy between profession and practice that is all too easy to discern. Bernstein suggests that an individual orientation encourages idiosyncrasy, or at least promotes a belief in the value of isolated and quite possibly eccentric individuals. On the other hand, a personal orientation seeks a less individualistic togetherness. In English terms, the one is epitomized by slogans about the home being one's castle, the other by community action and claimants' unions. If such contrasts can be substantiated (cf. Runciman's varieties of orientation to reference groups, 1972, ch. 2), I would suggest that they would go along with a different approach to moral thought as well. The individual orientation would focus more on the individual's personal responsibility (in good protestant work ethic ways)than on questions of general distribution. It would also tend to see such individuals as particular socially positioned people in a web of statuses and duties rather than as abstract utility maximizers. If such tendencies exist, then they would contribute to the greater likelihood of radical consequences flowing from moral reflection for members of the 'new' middleclass with their personal orientation than for members of more traditional groups.

Footnotes

1. In agreement with Hasan who similarly thinks that the aim has always been to "characterize the varieties of code by reference to the differences at the level of meaning" (1973, p. 291).

2. As noted previously, Bernstein has himself commented on this widespread ascription to him of "verbal-deficit" or "cultural deprivation" models of working class failure and he has published several versions of a critique of "compensatory education" (e.g., 1969 = 1974, ch. 10). He fails to convince everyone, however. Dittmar (1976) entitles his first chapter "The Deficit Hypothesis: Bernstein's assumptions on the correlation between speech and socialization"; and a recent commentator on the compensatory education paper concludes that "Bernstein is unable to combine his theory of codes with the progressive comments he makes on educational provision except at a superficial level" (Demaine, 1981, p. 40).

3. Counting structures or kinds of structure requires some specification of your grammatical theory, but that is no reason for some of the baseless criticisms that have been made. Dittmar, for instance, endorses a variant of a criticism by Labov that "the question of greater predictability ... can only arise when the speakers have command of a grammar capable of generating only a finite number of sentences" (1976, p. 91), but this is simply wrong. Peano's axioms generate an infinite number of natural numbers, but we could have a physical number generator which we rightly predict will produce a certain percentage of one kind of number (even numbers, for instance, or primes), and switching its control mechanisms might lead to a different percentage thereof. Just because a set of grammatical rules will generate an infinite set of structurally different sentences, we do not have to say that there are an infinite number of significantly different kinds of sentence or that there is no way to make a testable prediction about the output of actual employments of those rules.

4. Bernstein himself had accepted the phrase in the discussion reported after his 1967 on the understanding that he was not thereby committed to a Chomskyan contrast between competence and performance (1967, p. 37). But that association is not so easy to avoid, so he has reiterated that he is not to be interpreted in a Chomskyan way - most recently in an appendix to his 1981:

Theories that operate with a concept of competence (linguistic or cognitive) are theories in which the conditions for acquisition of the given competence require some innate facility together with interaction with a culturally nonspecific other who also possesses the competence.... Of course, no other who possesses a given competence can be culturally non-specific.... Be that as it may, and it inevitably is, theories of competence necessarily abstract the non-culturally specific from the culturally specific. Code is transmitted and acquired in interactions that are culturally specific. Codes therefore presuppose specialized others.... The concept code presupposes competencies (linguistic/cognitive) that all acquire and share; therefore it is not possible to discuss code with reference to cognitive/linguistic deficiencies located at the level of competence. Code refers to a specific cultural regulation of the realization of commonly shared competencies (1981, p. 356).

5. Such stereotypes were endemic at the time, according to one reputable contemporary: "revolutionary ideas were middle-class 'constructs'; they could never be engendered of their own accord within the soil of working-class culture, where all ideas and relations are dense, local, particular, and inarticulate" (Thompson, 1960, p. 24). For one criticism of Bernstein's caricatures of working class life, one out of many, see Fletcher, 1984, pp. 271-272.

6. Hasan's authoritative exposition explains implicitness in terms of exophoric reference, of which she says "the use of exophoric reference generally implies an assumption on the speaker's part that the hearer knows what the communication is about in general" (1973, p. 263). She goes on to link it to some of the other word counts of the SRU monographs by noting that "the structure of the nominal phrase will be predominantly simpler ... fewer modifying parameters will be selected to set aside the entity modified" (p. 264). The idea here can be seen by contrasting They're playing football (whose nominal phrase is the highly exophoric they) with Three boys who are dressed in red are playing football (where the nominal phrase is somewhat more complicated and simultaneously less context dependent, more explicitly helpful as far as identifying the reference goes).

7. The material in this section draws heavily on my paper 'Ellipsis and Ideology' (to appear, a) where the logical side of the issue is somewhat more fully discussed.

8. Conceptions of ideology are legion and provoke much argument. Rather than enter into these debates, I shall simply record the conception of ideology which I shall be using:

An ideology is a system of concepts, beliefs, and values which is characteristic of some social class (or perhaps of some other social group, perhaps even of a whole society), and in terms of which the members of that class (etc.) see and understand their own position in and relation to their social environment and the world as a whole, and explain, evaluate, and justify their actions, and especially the activities and policies characteristic of their class (etc.). Thinking in terms of this system unites and strengthens that class and helps to maintain it and to advance its interests. This system is determined by the social existence of the class (etc.) of which it is characteristic. It is not in general deliberately invented or adopted (though it may be deliberately fostered or propagated). At least some of the beliefs and concepts in the system are false, distorted, or slanted, and at least some of the activities sustained and guided by the ideology have a real function different from that which, in the ideology, they are seen as having (Mackie, 1975, p. 185).

9. Thus Hawkins sums up his study by saying "our study emphasizes the importance of defining one's linguistic categories according to both functional and structural criteria rather than on structural criteria alone" (1977, p. 185), and Adlam offers a case in point: When the coding frame was evolved it was intended to look at the extent to which children gave rationales for what they said. Somewhat naively, it was considered that rationale statements could be identified mainly by the inclusion of 'because' and less frequently by conjunctions such as 'so that' and 'in case' (1977, p. 71).

10. Bernstein's own remarks on this contrast are as follows:

Durkheim .... did not foresee, although his conceptual procedures make this possible, a form of organic solidarity based upon weak classification and weak frames.... Durkheim's organic solidarity refers to individuals in privatized class relationships; the second form of organic solidarity refers to persons in privatized class relationships.... These two forms arise out of developments of the division of labour within class societies.... personalized organic solidarity, it is suggested, develops out of increases in the complexity of the division of labour of cultural or symbolic control which the new middle classes have appropriated.... Whereas the concept of the individual leads to specific, unambiguous role identities and relatively inflexible role performances, the concept of the person leads to ambiguous personal identity and flexible role performances (1973a = 1977a, p. 125).