The choice of one form of language for formal usage and another, or some others, for informal purposes does not depend on the availability of a written literature alone. In a society, however primitive it may be all speakers may, as an established tradition, choose to speak on formal occasions like the richer members, the more aristocratic members or the members of the ruling class, thus reserving the language of their best fluency for daily informal behaviour. In doing this, all speakers do not achieve the same standard in the formal usage: what matters are that they believe that they shift their registers appropriately. The participants in a formal situation expect in such societies that the speakers should use the form of language reckoned to be prestigious. In this regard, I must relate an anecdote to prove my point. Tamils make a distinction between formal and informal usage. It was reported to me by two Tamil university lecturers that when they were invited on one occasion to speak on some vital maters (perhaps relating to agricultural economy) to a gathering of illiterate peasant cultivators, they decided to speak to them in the colloquial Tamil language which the whole audience knew rather than in the formal variety of Tamil which would be alien to them. They obviously attempted to communicate with these people in the most effective way. To their dismay, however, everyone in the audience got up and walked away, laughing and commenting. Upon investigation, they discovered the mood of the audience which may be best summarized as "How can they help us with our agricultural persuits when they cannot even speak our language!" It is highly unlikely that, being illiterate, and so forth, any of them would have been able to speak High Tamil despite their claim for an 'our language'. Many are the morals that can be drawn from this for our understanding of language in relation to its communicative potential, systemicity, belief-governed (that is, rather than entirely system-governed) acceptability, concepts of perfection, and so on and so forth.

The availability of a written literature and a literature segment in the population does, however, reinforce and, in some sense, stabilize cleavages such as formal versus informal, prestigious versus ordinary, and the like. Particularly where the varieties of usage involved are believed to be forms of the 'same language', as in the South Asian languages examined in this essay, the prestigious variety is invariably supported by some form of 'classical' literary norm accepted within the community as representative of the glorious heritage of the nation or race. The stabilization of diglossia, to the extent that it is possible to speak of a stabilization in changing linguistic behaviour, and at times the rise of diglossia, stem from the availability of literature and literacy in the community. The simple reason for this is the relative permanence of the written record and its effect as a model for those who aim at safeguarding and accomplishing the marks of excellence in usage. I propose, therefore, to examine at this point the types of writing devices which can in themselves be instrumental in the creation of diglossic usages. Fergusons' own definition of diglossia presupposes a literate society (although all members of the society need not be literate). The converse, however, is not true, that is, not all literate societies are necessarily diglossic communities. It is, therefore, expedient at this point to examine the relationships that exist between written and spoken representations of languages and which of those relationships qualify as capable of causing diglossic cleavages.

Where a language with no previous history of writing is analysed and reduced to writing for the first time (by a linguist or some other person), and where the resultant phonemic (or any other) script is used to record faithfully what is spoken by the people in question, a one-to-one relationship naturally emerges between the spoken language and its written representation. Such a correspondence is an ideal one to have, but it is often inexpedient and inefficient for practical purposes and is, therefore, generally short-lived. The inexpediency arises out of two factors. For social, political and cultural reasons, people regard diverse forms of linguistic behaviour as components of the same language; such diversities are then referred to as dialects and registers. A written representation which fits one dialect in a one-to-one relationship would not necessarily fit another dialect in the same way. This is especially so if the writing system is alphabetic. In order to endow the entire language with one written form, therefore, exercises toward standardization are undertaken. This results in the drifting away from the one-to-one relationship to a more distant relationship between writing systems and speech behaviour. This distance might be reflected not only in the symbolic representation of sounds but also in the grammar and the lexis. Secondly, people, as a rule, change their spoken linguistic behaviour more rapidly and more frequently than their written language. The reason for this is obvious. Because of the availability of a more stable model to which reference can be made for authority, the written usages are maintained more conservatively. It is true that, when too wide a gap is caused by the changes in the spoken language and the conservatism of the written, the written form is modified in the direction of the spoken: this accounts for the differences between, say, written Old Tamil and written Modern Tamil or written Old English and written Modern English; however, such innovations are exercised on the written language with much care and deliberation in order not to alter the form of the written representation any more than is absolutely necessary for maintaining a certain degree of comprehensibility. Because of the tendencies towards standardization and conservatism a gap begins to appear between the spoken languages and their written representations almost from the time of the first codification.

The nature and intensity of the resultant divergence between the written and spoken varieties of language are not always the same. Broadly speaking, it is possible to distinguish between four types of divergence.

First, there are the writing systems like Chinese. Here the writing represents language not alphabetically or syllabically - i.e., not as linear successions of sounds or syllables, but ideographically (or logographically or morphographically or lexigraphically) - i.e., using one symbol for one idea (or one word or one morpheme or one lexeme). Being in some sense stylized pictographic representations of concepts, the symbols used in such wiring systems have an air of timelessness about them. Spoken language change and diversify; and consequently different persons (in different areas and so forth) may acquire different linguistic expressions for the same concept; but the ideograms being symbolic representations of concepts, diversification of speech will not necessarily promulgate alterations in the writing system. Different persons may read ideograms each in his own pronunciation and each imposing upon them his own grammar. The semantic values of the ideograms are equally shared by all irrespective of their dialectal differences which are reflected in their speech habits. In these instances the acquisition of literacy is a cumbersome process: it is said that a Chinese has to master as many as three thousand symbols to become literate; but the number of symbols is kept in humanly manageable proportions by representing the world view through a complex organization of concept-symbols devised stylistically to portray a subtle system of categorization. Form the point of view of the relationship between speech and writing this kind of writing appears to be an ideal one, for it can always be said that the writing system is, in a sense, a true representation of each person's verbal behaviour. Here, the written form does not exercise a direct influence on the spoken language; nor are major changes promulgated in the writing system by the natural changes in spoken verbal behaviour. This is a case where the writing system does not motivate the creation of a diglossic situation.

In contrast to these ideographic or logographic writing systems there are the syllabic and alphabetic wiring systems. Syllabic and alphabetic systems have specific implications of utterance. An Old Tamil text, for instance, cannot be read as if it were modern; a passage of Old English cannot be read as if it were Modern English. In such cases, older texts cannot be read or understood by people who have not had a particular training in the respective older languages. In some situations like this where the disparity between the old and modern forms of language is clearly and unmistakably represented in writing, there is a general belief that the older written language is better and purer. The prestige thus accorded to the older form has religious and cultural implications. Arabic is an instance of this. The older language, or Classical Arabic, has been associated with the codification of the teachings of Islam and is, therefore, regarded with respect and reverence by all Arabic speakers as the pure representation of their language. Different regional varieties of Arabic are not mutually intelligible, but the standard literary language is shared equally by people of all regions irrespective of the mutual unintelligibility of their vernacular speech habits. This is a classic case of diglossia: the day to day conversations are impracticable in the prestigious variety and are, therefore, conducted in the vernaculars; all prestigious activities are carried out in the high or prestigious variety, which is, in the minds of the native speakers of all varieties of Arabic, purer, better and more beautiful (Ferguson 1959a). This high or sacred form of language is taught in schools in order to train the pupils to write correctly and speak suitably on formal occasions. Tamil is another instance of this kind. As in Arabic, the high or formal variety and low are vernacular variety in Tamil have mutually exclusive uses; high Tamil has been preserved for cultural reasons.

Then there are situations like the English one. Here the written language is not a prestigious model preserved from the past for any religious or such other reasons. In English writing there are archaic characteristics in so far as the spelling system is reminiscent of an older phase of the language. There are also pedantic lexical items which are used rarely, if at all, in speech. However, the grammar and vocabulary as well as the phonology that the spelling represents can be used in day to day affairs if the speaker so wishes. There is, thus, at least for a reasonable segment of the population, no clear separation between the functions of the varieties. Except where it becomes necessary to use dialectal features for some effect-for instance, for dialogues in novels-the language represented in writing is the product of some standardization. It is well nigh uniform and is universally acclaimed as the standard. In situations such as this, it is possible to discern a close similarity between a nationally recognized standard speech and the written from (i.e., the grammar, etc. it represents). It is not possible to state a general rule for whether the standard speech conditions the writing or vice versa, or even whether the relationship between them is one of inter-dependence rather than one of cause and effect. This is an aspect which must be examined for each language by carefully analyzing the various vicissitudes of its history. In English, there is a striking similarity between Standard English or the educated speech of the southeast and the language that is generally written. Because the standard speech may be used in all situations, there are no mutually exclusive settings for the prestigious and local varieties. For this reason, English and other languages with similar characteristics do not normally qualify for diglossia as defined by Ferguson.

We now come to the last type where the divergence between the vernacular and literary varieties is of a different order. We have observed that in Arabic and Tamil the high variety is used on formal occasions as well as for writing. In the type we are presently discussing, however, the high variety is never used for speaking even ou the most formal occasions: it is used only for writing. Thus, this type is characterized by the availability of distinctly different spoken and written usages. Two examples of this are Sinhalese and Telugu. I shall have occasion to talk about certain new innovations in Telugu usage later on, but in Sinhalese, the high or written variety is not used for speech at any time, and there are many features in the spoken usages that would not be permitted in the orthodox written language. Sinhalese (perhaps like Telugu before the recent events which I shall refer to later on) is a clear case of diglossia, for it shows mutually exclusive domains of use for the two varieties; it does, however, differ slightly from Ferguson's defining languages in that many normative literary rules are not allowed even in formal speech. The high variety here has, in this sense, a more restricted function, and consequently the concept of prestige, too, is different here from that in Arabic and Tamil situations.

On the basis of the foregoing discussion it is now possible to say that any linguistic situation which is collectively characterized by the following three principles might be viewed as an instance of diglossia:

a) There is a superposed variety; the characteristics of this superposed variety have been specified by local grammarians and therefore the divergences between the extreme form of this superposed variety and the local or natural speech habits are statable fairly clearly.

b) The two varieties function in mutually exclusive settings, graded in terms of prestige; these settings are clearly definable for the most part.

c) Both varieties are socio-culturally recognized as functionally distinct, and may be contained in the linguistic behaviour of the same individual in the following manner: a person who is a native to such a community and has a mastery of the prescribed use of the high variety also knows the correct social use of the low variety; although the converse is not true, the education systems in such communities aim at teaching the high variety to all their members.

It will have become clear from our discussion so far that diglossic communities are those which hold rigidly codified values with regard to prestige and propriety and which, in the choice of models of excellence, have created a renaissance of classical cultural and/or religious models are brought ot bear upon the society the impact, if only for certain formal purposes, of the values inherent in a supposedly superior or more forceful community. Language being the most dominant means of communicating cultures and values, the impact of these notions of prestige and propriety are most noticed in the people's linguistic behaviour. Linguistic choice is, in this sense, the most significant symptom of this kind of community organization, whence the term 'diglossia' (rather than 'di-' anything else!).