The observations I have made in the last chapter and other related observations on linguistic diversity and language acquisition compel me to take an excursion into the realm of two notions in which modern linguistics is deeply enmeshed. These are the notions of linguistic community (or speech community) and competence. My primary objective in taking such an excursion would be to examine if diglossia and its problems as narrated in the proceeding chapters could throw any light on our understanding of the individual in relation to his community and, more particularly, if there could be a significant relationship between the individual's competence and 'the language of a community' as understood by the theoretical linguist.

The subject matter of theoretical linguistics has been categorically defined by Chomsky (1965) in his now famous assertion:

Linguistic theory is concerned primarily with an ideal speaker-listener, in a completely homogenous speech-community, who knows his language perfectly and is unaffected by such grammatically irrelevant conditions as memory limitations, distractions, shifts of attention and interest, and errors (random or characteristic) in applying his knowledge of the language in actual performance.(p. 3)

Idealization is thus taken as a major prerequisite to theoretical investigation. Other linguists follow suit. Notice following statement by Lyons (1968), again on the idealization of data:

When we say that two people speak the same language we are of necessity abstracting from all sorts of differences in their speech. These differences, reflecting differences of age, sex, membership of different social groups, educational background, cultural interests, and so on, are important and, in principle at least, are to be accounted for by the linguist. However, in the speech of my persons who are said to 'speak the same language' there will be what may be described as a 'common core' - a considerable overlap in the words and sentences. The possibility of communication depends upon the existence of this 'common core'. For simplicity of exposition, we shall assume that the language we are describing is uniform (by 'uniform' is meant 'dialectally and stylistically' undifferentiated: this is, of course, an 'idealization' of the facts …) and that all native speakers will agree whether an utterance is acceptable or not
(p. 140-41)

These two quotations beg a number of questions which can take up many pages of discussion. One is, for instance, at a loss to comprehend why Chomsky bundles together memory limitations, distractions etc. and errors; while I am able to appreciate the meaning of the phrase 'random errors', I cannot see the implication of the term 'characteristic errors', for characteristic behaviour being in some sense regular for the individual, the branding of such phenomena as errors must be motivated not by the individual's own judgement, etc. but by the observer's use of some external yardstick. Here, we straight away resort to the communal or homogeneous usage, irrespective of whether we can observe it or not. Lyons mentions speech differences reflecting differences in age, sex, group membership, education, etc. These, according to Lyons, are relevant (unlike Chomsky's irrelevancies), but he is prepared to advocate that the theoretical linguist may ignore them all the same. What all this actually means is that any difference from a hypothetical norm is unimportant for the theoretician. This norm is, for Lyons, the 'common core' or the 'overlap' 'in the words thy use, the manner in which they combine them in sentences and the meanings which they attach to the words and sentences'. Lyons's statement may be taken to mean that there is a dialectally and stylistically undifferentiated grammar and lexis which everybody knows and which is subsequently differentiated by the introduction of features reflecting the social, educational and such other characteristics of the individuals. Or, alternatively, given a body of linguistic data, it is possible to abstract an idealized common core by subtracting from its totality all features reflecting social, educational and such other considerations. Such an attitude to linguistic data presupposes, to my mind, the availability of a theory of language-use which must be far more refined than any theory of language-use one can visualize namely, a theory of language-use which can predict precisely and exhaustively all linguistic correlations of age, sex, social group membership, education and cultural interests. Without such a theory by means of which the additions and subtractions may be performed, we can never see the extent to which the linguist's common core is a true reflection of the grammatical and lexical common core that might be held by all members of the community.

Linguists have thought for a considerable length of time that the content of this common core, or the idealized, homogeneous, communal aspect of language can be tested by referring to the native speakers' intuition. Lyons asserts in the above quotation that all native speakers will agree whether an utterance is acceptable or not; Chomsky holds the same view. Much of modern linguistics, indeed, depends on the reliance on intuitive judgements as a valid testing procedure. It has, however been shown conclusively by such linguists as Labov (1970 etc.) that, while 'generative grammar is the best discovery procedure we have', 'the search for homogeneity in intuitive judgements is a failure' (1970, p. 39). It is not necessary to go over the ground that Labov and others have covered adequately. Suffice it to say this: if the idealization the linguist requires is abstracted form the linguistic goings on in communities, then we cannot, in the present state of our knowledge of language and language-use, abstract an idealization which is anything like complete, exact or testable. Grammarians who do not wish to take note of diversity (as done in, say, Bickerton 1975) should gear their grammars to a highly abstract level: the use of terms like 'community', however, pulls them down from such heights into conflict with issues relating to language-use.

Idealization, then, is based on unprovable assumptions. Idealization has, however, been taken for granted in linguistic science at all times. Every grammar book that has been written from Panini to the present day is an attempt to describe this idealized system at various levels. Behind the stipulation of the idea 'one language, one system' (remember Meillet: 'Chaque langue forme un système où tout se tient') there is always the notion of homogeneous linguistic community. Some grammarians have imposed homogeneity upon languages by making their grammars prescriptive in character. Following the argument adduced by S.K. Chatterji (1960), it may be said that Panini's AÀadhyayi is a prescriptive work of this sort. According to Chatterji, Panini's grammar seems to be the culmination of, or at least an important landmark in, an effort to restore the 'purity' of the Sanskrit language which had been 'corrupted' differently in different parts of the Aryavarta (Aryan India), in the eastward movement of the Aryan immigrants. It is understandable why a prescriptivist would choose to represent the language deliberately as homogeneous: his aim would be to eradicate the 'vulgarities of the vernaculars'. Such puristic efforts are known in many parts of the world even today. It is one thing to impose homogeneity upon languages for prescriptive purposes; it is, however, an entirely different thing to assume homogeneity, at the expense of contrary evidence, in what are seemingly non-prescriptive, dispassionate analyses of human languages. 'Simplicity of exposition' (see Lyons above) is not in itself a valid goal, for achieving which facts about language must be sacrificed. Where norms are clearly and unmistakably specified, homogeneity (or near-homogeneity) should be comparatively easy to achieve, with the help of the educational system and so forth What one sees in the diglossias under consideration, however, is that in spite of the cumulative effect of all these circumstances, the individuals' performances are not always similar, and a cleavage persists between the specified norm and the individual's ability or competence as seen through his performance. I use the phrase 'competence as seen through his performance' purposely; for it is not possible to brush aside the irregularities in literary usage, on which I have commented previously, as products of memory limitation, distraction, and such other factors. All situations in which specified norms serve as models show us that, while the force of such norms is considerable, the people's behaviour continues to fluctuate, rarely reaching the norm in its precise dimensions.

The notion of homogeneity is very much related to the notions of monosystemicity and internalizability as characteristics of each language. Being homogeneous, each language, in its idealized version, has one system. It is known in theoretical linguistics that the speakers 'internalize' this system. Homogeneity, monosystemicity and internalizability ensure that those same qualities perpetuate. However, if the language in some speech community were homogeneous, and, being homogeneous, this language contained a system that the speakers would internalize, then these three qualities of homogeneity, monosystemicity and internalizability might be seen as having the power to block any change. The belief in homogeneity, etc. is, however, refuted by the fact that the speech habits in no given geographical area have ever remained identical in the history of that area.

It is, of course, true that people who live in the same community and amicably interact with one another, speak in a manner, which greatly resembles one another's speech. This is indeed to be expected from the nature of the learning resources available to them. First of all, being of the same species their biological learning capacities are similar; secondly, being in the same culture and environment, their world view tends to be similar; and thirdly - this is very important - they share the same linguistic data pool. In their acts of identity and communion people tend to make unconscious attempts to be like those with whom they wish to be identified (Le Page, 1968. etc.), and so, even if each speaker brought into the common pool his own personal brand of data acts of sharing one another's features would emerge in the normal course of events. An assimilated form of speech behaviour characterizing the entire community is, therefore, a natural thing to expect. It is, thus, an accident of history that people living in the same ethnic, political and social communities, where communication with one another is essential, come to learn rules that greatly resemble one another's. The similarity thus created would enable the members of the community to interpret one another's rules with some success. If varying speech acts, pooled together as a collective data source, by people living in the same community, can in this way increase the tendency towards identical behaviour rather than decrease it, that the members of a group should have similar rules should not surprise us; for the same reason it might also be said that it should not be terribly interesting to us. What is interesting is that, despite the nature of the data resources, dissimilarities continue to prevail, distinguishing one individual from another. As history has shown us, the linguistic habits in any geographical area can change beyond recognition in the course of time. No norm, in however much detail it may be specified, can block this tendency to maintain diversity which characterizes all group linguistic behaviour.

A person's ability to 'understand' another person's speech is not in itself proof that they share the same system. Before talking further about this point, let me hasten to add that if having the same system ensures understanding, there could be no acts of misunderstanding in communities which are allegedly homogeneous in their speech behaviour. Understanding depends on a variety of factors, but system equivalence need not be one of them. It is reasonable to assume that communicability rests on the listening participants' natural skill to construct simultaneous sentences of their own, against which they may map the speaking participants' sentences. A one-to-one correspondence between these two is rarely achieved. That we construct our own sentences simultaneously is proven by the way in which we often fail to hear things in the utterance which we do not expect to hear, and by the manner in which we often complete the speaker's utterances for him, particularly when he is slow and deliberate (sometimes with disastrous results!).

This idea that simultaneous sentence construction is a necessary act in communication episodes seems more acceptable than the belief that communication depends on the availability of a shared common core. It is a truism that the most important requirement in verbal communication is the availability of a shared lexicon which contains sufficient clues for contextualization. In speech communities people develop their lexicons as verbal references for things, events, actions, etc. in order to talk about them. While, in the nature of language as we know it, grammar and lexis are inseparably welded together, people can get much farther with words alone than with grammatical patterns alone. Where the matter that is being communicated - that is to say, the topic of conversation - is straightforward enough not to lead to more than one interpretation, it is often possible to get by with the lexicon, without regard for language - specific grammatical organization. I use the phrase 'language - specific' advisedly' if the nature of human reasoning and the human ability to think in construction, etc. are reflected in the organization of human language, it would be impossible to utter a completely ungrammatical sentence, that is totally contrary to the humanness of the human speaker: the features we refer to as errors are language-specific. (Language, here may mean the individual's language in my sense or a communal language in the grammarian's sense: this does not matter for the present argument.)

'Bus-tell-where', 'where-bus-tell', and 'where-tell-bus', are "sentences" which are unlike any corresponding sentences spoken by any Englishman. They are, therefore, ungrammatical with reference to the speech behaviour of the English. These "sentences", however, are not communicatively inefficient, for it is conceivable that the addressee would "understand" the implication of such an utterance. A totally unambiguous message is one which is contextualizable only in one way. Such a message may be regarded as one which has the fullest amount of understandability. 'Bus-where-tell', etc., in this sense, contain sufficient information with which the hearer may construct a situation. Such a situation need not be physically present; if it is not physically present, it must be constructable on the strength of the situational information in the message. The sentence that the hearer may construct simultaneously with the hearing of this utterance will be his own, and if the hearer is an Englishman, his sentence will hardly be like the one he receives from the speaker. This is an instance which exemplifies the irrelevance of a shared, language-specific grammar in order to engage in communicative acts.

In order to expand some notions I hold with regard to understanding, and more particularly, to understanding sentences which are deviant from the point of view of the hearer, I wish to repeat here very briefly the results of a rather crude test I have already talked about in De Silva (1970). Out of some two hundred people questioned, everyone understood the utterance 'me-town-going-bus-where-stop-tell' to mean that the speaker wanted to go to town and was asking for the bus stop. To the same people was put the utterance 'me-brother-rat-kill'. All but four people interpreted this to have an SVO order as "my brother-kill-rat" (Tense, number, etc. were left vague). These persons had been required to give their responses immediately after the utterance had been made. The 'brother-kill-rat' interpretation was received in terms of this requirement. The four people who gave different interpretation took longer time: three of them took about forty-five seconds and interpreted the utterance as 'rat-kill-my brother'; the fourth one took about three minutes and gave the fascinating interpretation 'my brother is a rat. Kill him!'

What is the moral of this story? Where the utterance contains sufficient information, given via the shared lexis, etc., for the hearer to make suitable situational constructs, the simultaneous sentences he constructs would provide unambiguous understanding. In this act of understanding he disregards or ignores the deviances in the speaker's utterance. Where the speaker's sentence is deviant from the point of view of the hearer's grammar, the perceptive hearer may, depending on how obvious the deviances are, recognize the presence of such deviances; he may, if his memory is as good as his perception, also identify the exact places in the phonetic sequence which are phonetically, phonologically or grammatically deviant. Not every person can, however, repeat a deviant sentence as he hears it, particularly if the deviances are more subtle than the ones I have talked about. The hearer's inability to reproduce all deviant sentences does not depend entirely on the subtlety of the deviances; it stems, at least partly, from his preoccupation with the simultaneous construction of his own sentences as a necessary requirement for understanding.

Unlike the 'bus stop' utterance, 'me-brother-rat-kill' does not contain sufficient situational information to facilitate a unanimous interpretation. Where situational information is negative or insufficient, a great deal of similarity between the input sentence and the hearer's simultaneous construct would be essential to yield an acceptable interpretation. Any sentence that an Englishman would construct to correspond to the utterance 'me-brother-rat-kill' would be of the SVO order, with the subject-object relationship reflected by their positions on either side of the verb. People who are used to interpreting NVN as SVO would naturally find NNV an ambiguous construction. Why, then, did ninety-eight per cent of the people tested give an interpretation which would reflect the SVO order as 'my brother-kill-rat'? It seems to be the case that, in making simultaneous sentence constructions in order to interpret deviant sentences, people only make the minimum distortion to the input sentence: one might call this the politeness consideration in sentence interpretation in so far as the hearer does the least 'damage' to the speaker's sentence; more seriously speaking, however, it might be all that the hearer can do in simultaneous interpretation, due to lack of time for more considered and thought out interpretations. The minimum change that can be made in 'me-brother-rat-kill' to provide an SVO order is to move 'rat' to the position after 'kill'. Notice that those four persons who gave other interpretations took longer time: 'rat-kill-my brother' took about a minute and the one person who gave a very complex interpretation took much longer.

Although the hearers' interpretation favoured 'my brother-kill-rat', we do not know what the speaker actually meant. Where situational information is lacking, some language-specific grammatical similarity between the input sentence and the hearer's construction would help greatly. Where situational information is present, such grammatical similarity would not be of paramount importance. When, for instance, the two hundred persons that were tested as above were told that 'me brother-rat-kill' was uttered with a dead rat in hand, they all interpreted the SVO order to be 'my brother-kill-rat' unambiguously. Most verbal behaviour takes place in situationally definable circumstances. Uncontextualizable utterances are very rare indeed. The need for grammatical identity as a communication requirement is, therefore, never very great.

I hope I have succeeded in showing in the above paragraphs that the presence of a language-specific 'common core' (outside of lexis) is not always a necessary requirement of communication. (let me hasten to add that my hypothesis is not based on just two funny utterances; it is not relevant to produce all my test examples in this essay.) This is perhaps a convenient point to examine the term 'competence' as introduced by Chomsky (1965) and developed in various ways by others (e.g. Campbell and Wales 1970; Hymes, mimo; Le Page, 1973). I have said above that, in closely welded societies, individuals, by virtue of the availability of the same data resources (and other factors), learn to speak more and more like one another, and that, yet, diversity persists. The total competence that each person has for producing, interpreting and assessing utterances is not necessarily the same as anyone else's in that community. Quite rightly, no one who has described the notion 'competence' has found the need to invoke the hypothesis that competence is communally derived or shared; for all of them competence means noting more than the individual's ability. By invoking the belief that languages in specific communities have homogeneous bases, they do, however, imply that each person's competence is, by and large, representative of the competences within the community as a whole. Chomsky's competence is, unlike the Saussurean concept of 'language', an individualistic one; and as one writer puts it,

The distinction between competence and performance in language is sometimes thought to be the same as, or similar to, de Saussure's between 'language' and 'perole'. But this is not so. His distinction was between one's stock of linguistic materials and the utterances that could be composed out of them. To suppose that this is similar to the distinction between competence and performance is like treating distinction between dough and loaves as similar to that between the ability to bake and baking. (Cooper 1975: p. 28)

Chomskys' definition of competence is very non-Saussurean in that competence has been shown every-where to be the individual's ability rather than a collective, social ability. Notice a few of his definitions: (My italics)

Every speaker of a language has mastered and internalized a generative grammar that expresses his knowledge of his language. (Chomsky 1965: p. 8)

A distinction must be made between what the speaker of a language known implicitly (what we may call his competence) and what he does (his performance). (Chomsky 1969: p. 9)

On the basis of a limited experience with the data of speech, each normal human has developed for himself a thorough competence of his native language. (Chomsky 1964: p. 8-9)

Despite this individualist character of competence, Chomsky himself does, however, draw our attention, although with some qualification, to the similarity between de Saussure's langue-parole dichotomy and his own dichotomy of competence and performance (Chomsky 1965, p. 4). It is a legitimate question to ask why Chomsky felt any need for seeing some comparability between de Saussure's sociological notions and his own mentalistic notions. The answer to such a question, it seems to me, is a simple one. Although the notion of competence may imply an individual's mastery of his linguistic rules, Chomksy obviously wanted to see a social basis or social justification for it which would incorporate the notion of homogeneous community - a notion that is very important in his theory. More regard has been paid to the homogeneity of the community than to the individuality of competence in all Chomskian grammars (and any other grammars) that have hitherto been written. Had the individual character of competence been the theme of linguistic investigation, the difference between competence and performance might not have emerged as very significant, and, there might not have been a necessary connection between the individual's competence and the so-called deep structure of 'the language' which his competence is said to portray.

The question has not yet been answered as to how far it is reasonable to delve in order to discover the deep structure of a language. Lyons's use of the term 'common core' is perhaps synonymous with the term 'deep structure'. In order to arrive at the deep structure of, say, Tamil, it would be necessary to set up a neutral base in which the vertical and horizontal, or stylistic and dialectal, differences would be greatly reduced. Just as spatial diversity has to be, in some sense, got rid of, one might argue that temporal diversity ought likewise to be wiped out in this exercise. A base or deep structure, from which all Tamil dialects may be derived by the addition of variable rules of a more superficial kind, may equally be the base from which several historical strata of Tamil might be derived. Despite this situation, it is the case that, while all dialects are said to be of the same language, the basis of which all speakers possess, not all historical strata are said to be derivable from the same base (unless this base is a very abstract one indeed). The question that has not been answered satisfactorily is, how far back is it legitimate to go in the history in setting up a deep structure for a language that would be meaningful in relation to the speaker's competences?

It so happens that the languages of which the deep structures are sought after by the linguist are languages identified as a whole, by the people who do so far a variety of reasons. Labels such as 'they speak the same language' are not necessarily linguistic labels but repercussions of the social, political, cultural, economic and religious thinking of the people. The linguist must, indeed, respect the people's wish to identify their verbal behaviour and label them; he must not, however, interpret the people's beliefs as scientific statements, which are linguistically factual.

The question as to how many strata of 'the language' the competence of the 'native speaker' should be able to generate is an important one, particularly when we consider the diglossic situations in question. It so happens that, as I shall show in more detail later, the high varieties in all three communities under survey have been resurrected from the linguistic behaviour of several centuries ago. The natives of each of these communities believe that the high variety is part of their language; some of them, in fact, believe the high variety to be the real language of which the spoken counterparts are inadequate renditions. If we attempt to set up our deep structures with a view to accommodating the beliefs of the native speakers (- this being what the grammarian seems to do all the time), we have to write our grammars to generate several centuries of linguistic usage (or, more precisely, the usage of the thirteenth, fourteenth and twentieth centuries) simultaneously from the same deep structure. This is precisely what William Bright (1970) suggested as a treatment of Kannada phonology. Bright's suggestion, if brief, is this: the high variety is more complex in grammar and phonology. If the underlying structure were taken to be more similar to this complex structure, then the low variety could be derived from it by simple deletion rules. Simplicity would not be the only achievement; such a statement would not be the only achievement; such a statement would have a historical validity as well. This approach that Bright proposes have several drawbacks. If historical justifiability is a valid consideration in setting up deep structures, why stop at the thirteenth century? Why not make the grammar generate all known strata of 'the language' from the earliest historical beginnings of the community? In addition to this arbitrariness of choice (notice that this corresponds to the arbitrariness of choice in the creation of diglossias) there is the further, and even more significant drawback, namely, that such an approach somehow seems to favour languages whose historical antecedents are known: would, for instance, a deep structure written for a language whose history is not clearly known, be comparatively more tentative? There is the other theoretical consideration, namely, that a deep structure which may predict several centuries of linguistic behaviour in a community may also, by the very nature of its rules, be able to generate linguistic phenomena belonging to quite different communities. From the rules written for, say, Tamil, it would be possible to derive Malayalam and other Dravidian languages of South India as well as some non-Dravidian ones.

If the membership of a community shared an idealized grammar for their speech production, speech interpretation and speech assessment, the competence of each member must resemble that of every other member of that community. The individuality of competence, however, contradicts such a hypothesis, and renders competence, and homogeneity two concepts whose relationship cannot be taken for granted in linguistic studies. In spite of the much-repeated assertion that individuals master the correct grammar from exposure to limited and often incorrect data, the nature of an individuals' competence seems to depend, to a large extent, on the amount of exposure he has had to linguistic data. Let us take one simple example consisting of the three sentences 'Give it to me', 'Give me it' and 'Give it me'. All so-called native speakers of English can say 'Give it to me', and may be said to have a rule to generate it. Many can say 'Give me it' and it may be said that most, but not all, have a rule to generate it. Only some can say 'Give it me'; a good many English speakers, then do not have a rule for it. After living in areas where 'Give it me' is a possible sentence, those whose rules could not generate it earlier learn to say it and thereby extend their rule repertoire. It may be argued that this is a surface rule, but when we are not sure how deep we should delve (see above) the deep surface dichotomy becomes less clear-cut. I think this was, in part, the linguistic essence of Bernstein's original position also.

What does all this mean for diglossia and literacy? In the diglossias under survey, the mere availability of a norm specified in great detail does not prevent performance diversity at all levels. My observations in the previous chapter and elsewhere (De Silva 1974a) would illustrate this. There are several lessons we can learn from these diglossic communities. When a person who has learnt some mode of verbal behaviour, or, if preferred, one kind of rule schema, is exposed to further or different dimensions of behaviour, he acquires those new modes, at least partly, in a form affected by the mode which was already known to him at the time of learning; this does not account for all types of diversity that prevail in linguistic behaviour bu5t it contributes to the perpetuation of diversity in a significant way. This is, of course, the old notion of interference, but I am using it here in a wider sense, to suggest that interference of the known (tentative) behaviour is a factor that affects the individual's learning, even within his own community. Writing is perhaps one of the most careful acts of linguistic performance. It may, therefore, be suggested that diversity in the grammar etc. in the written usage, which would reflect varying degrees of distance from the specified norm, is closely linked with competence diversity. Where specified norms or models have functional limitations such that the behaviour episodes in terms of that model could never match in quantity or frequency the performance phenomena in terms of functionally more productive varieties, a performance behaviour which would be an exact replication of the specified norm would be very difficult to acquire.

The individual's rule repertoire is an ever-evolving thing. As I have said earlier, such evolution depends on the amount of exposure to linguistic usage. If a particular type of usage does, by its very nature fail to provide sufficient data to outweigh in bulk other data to which the individual may have access, the individual can rarely use it to its fullest extent. The imposition by the society of such a usage which, by nature, is delimited in scope, can have an inhibiting influence on the learner, making him feel insecure in the domains, for the free expression in which, this alien form is the one to use.

What I have attempted to show is that, outside of the social pressures, there is no justification, particularly linguistic justification, for the maintenance of diglossia. The high variety, in its specified norm, rarely matches the individuals' competence, and the reasons for this, which I have briefly outlined above, are justifiable ones. Should the society demand a norm-like accuracy from the learners in such a situation? A careful dialogue on this subject would be sociologically and sociolinguistically a worthy one to conduct.