An Encyclopaedia of Language_09 docx

of mankind at a correspondingly primitive stage of development. From these beginnings language typology emerged. The
problem of the origin of language itself suffered an interesting fate: the prize offered by the Berlin Academy in 1771 for an
essay on this subject attracted some 31 entries, and, despite Herder’s winning contribution (Abhandlung über den Ursprung
der Sprache), continued to prompt lively discussion to the end of the century and beyond. That the Société de Linguistique de
Paris found it necessary to proscribe contributions on this subject as late as 1866 shows as clearly that popular interest in the
subject was still very much alive as that the academic mainstream despaired of ever finding a solution.
But in the meantime an unexpected new element was introduced: the Sanskrit language. The question of language origin
was first briefly diverted into the assumption that Sanskrit was the parent language, and then forgotten as energy went into the
detailed formal comparison of Sanskrit and cognate languages. A few scholars took advantage of the breadth of perspective
afforded by the discovery of Sanskrit and still more exotic languages to formulate far-reaching hypotheses about the nature
and role of language. Wilhelm von Humboldt, brother of the scientist and explorer Alexander and a friend of Goethe’s,
viewed language as the organ of inner existence, the way to understanding—or manifesting—thoughts and feelings (Über die
Verschiedenheit des menschlichen Sprachbaues und ihren Einfluss auf die geistige Entwickelung des Menschengeschlechts,
1836). The human spirit manifests itself in different forms of civilisation and culture among different peoples, and also in
different languages. Each is an attempt, an approximation, a contribution to the universal need to develop a people’s
intellectual and spiritual powers and to unfold their own mode of relating to the world. Far from representing phenomena
directly, each language articulates that speech community’s perception of the world around it from its own distinctive point of
view. Humboldt’s deep interest in the essence of individual differences between languages is found not so much in the
mainstream of theoretical linguistics in the twentieth century as in anthropological linguistics, initiated by the work of E.Sapir
and B.L.Whorf, and in the work of German scholars from Leo Weisgerber on.
But by the time Humboldt’s great work was published (posthumously) the universal aspect of language study was being
pushed aside by the newly emerging discipline of comparative philology. Although philologists of the stature of Jakob Grimm,
Max Müller, H.Steinthal, and W.D.Whitney concerned themselves with problems like the ultimate origin of language, its
relation to thought, and its position among the sciences, their writings on these subjects were eclipsed by the contemporary
enthusiasm for historical and comparative work. Thus it was that although the ideas taught by Ferdinand de Saussure in his
celebrated lecture course at Geneva (Cours de linguistique générale, published posthumously on the basis of his students’
lecture notes in 1916) were far from novel, and had been in circulation intermittently during the nineteenth century, in the
changed intellectual climate of the post-Great War world they struck the scholarly community as fresh, stimulating, and above
all unfamiliar. The consequence was an over-rigid interpretation of ideas presented schematically in the Cours which had
once been current in subtler and more diversified guises. Such, for instance, has been the case with the doctrine of l’arbitraire
du signe, the arbitrary nature of the relationship between the linguistic sign and what it denotes. Elevated into a dogma in most

branches of linguistics, it contravenes the intuitive feeling of the native speaker (and the literary critic) for the affective value
of certain sounds and sound-groups, a fact recognised by von Humboldt (and by Plato long before). The linguistic sign,
defined as the union between a concept (signifié) and its acoustic representation (signifiant), is yet another manifestation of
Aristotle’s schema, but lacking the refinement of the level of percept. Saussure’s distinction between the synchronic and the
diachronic approach to language was taken as a charter for the liberation of a synchronic descriptive structural linguistics from
historical linguistics, despite Saussure’s own awareness of the symbiotic relationship of the two temporal axes. Perhaps the
most fruitful of Saussure’s insights has been the celebrated langue−parole dichotomy—langue that aspect of language which
is a system of signs existing in the speech community independent of the will of any individual, while parole denotes the
particular utterances of individual speakers; langue is the essential, parole the accidental aspect of language.
Saussure’s importance lies more in his making explicit the implications of a structuralist approach to language than in the
specific tenets of his doctrine, most of which have been modified by both European and American Structuralists—Bloomfield
and his followers, Troubetzkoy and other linguists of the Prague School, Hjelmslev and the Copenhagen School, Martinet,
Chomsky and the many offshoots of Generative Grammar—who all acknowledge some degree of indebtedness to him. But
more important still are the consequences of his work for disciplines on the periphery of linguistics or quite outside it:
semiotics, anthropology, history, psychology, philosophy and literary criticism. Particularly among French scholars—Lévi-
Strauss, Foucault, Barthes, Derrida—Structuralism has undergone a development which takes it far beyond its original
context. Within the linguistic mainstream, attention has recently returned to semantics, and the allied discipline of pragmatics,
heavily influenced by logic, has sprung up. The twentieth-century development of these and other branches of Western
linguistics is covered in Chapters 5 and 6 of this volume. Rather than duplicate this information, let us now shift the focus of
this hitherto Eurocentric account to some equally rich but—to the Westerner—less accessible traditions.
AN ENCYCLOPAEDIA OF LANGUAGE 445
3.
NON-EUROPEAN TRADITIONS
The Greco-Roman tradition out of which modern Western linguistics has grown is only one of several independent traditions
of language study known to have existed in the last three millennia. But it is by no means the most ancient, nor necessarily the
most varied or subtle in its understanding of language. The ancient Near East, India and China can all boast of linguistic
traditions of greater antiquity, while for richness of insight and comprehensiveness of scope both India and the Arab tradition
compete on equal terms with the West. Each of these traditions arose independently of the others and for the most part
developed separately, drawing on the resources of the culture within which it grew. Episodes of contact and mutual influence
were few but significant: the Arab tradition was repeatedly fertilised by insights derived from the works of Aristotle;

Buddhist missionaries took the Indian system of phonetic classification to Tang dynasty China; and Westerners twice received
a vital stimulus from alien traditions, from the Semitic tradition around 1500 and from India around 1800. But to understand
how each tradition came to take its characteristic shape it is more important to consider people’s outlook and way of relating
to the world as they are manifested in all areas of culture, than to go source-hunting. In any case, a borrowed piece of doctrine
takes on a different significance when transplanted into a alien cultural setting, and the very act of borrowing is itself
important.
Although the Arab, the Indian and the Chinese linguistic traditions would each merit discussion on the same scale as the
European tradition, space permits of no more than the barest outline. In the circumstances, it has seemed most useful to
discuss these traditions with reference to the Western tradition, contrasting the course of their development as well as glancing
at those areas of doctrine where contact has taken place. The works mentioned in the suggestions for further reading offer a
more comprehensive view of these traditions, as well as ample bibliographical leads.
3.1
The Arab world
Apparently arising out of nothing, no venerable Classical tradition stretching behind it, Arab grammatical writing makes its
first appearance in the elaborate and detailed form given it by its most renowned exponent, Sibawayhi, a bare 150 years after
the death of Muhammad. This extraordinarily rapid codification was followed by an equally speedy and prolific expansion
into all areas of language study: phonetics, morphology, syntax, semantics and the philosophy of language. In the space of six
centuries linguistics among the Arabs developed in a direction more akin to Western post-Renaissance linguistics than to
contemporary medieval work in the West. Yet by 1400 its momentum was gone. Instead of sharing in the intellectual change
of direction brought about by the new world outlook of the European Renaissance, Arab writers in all branches of
scholarship, scientific as well as linguistic, continued to work out of their old traditions, which had ceased to provide fresh
inspiration. By comparison with European linguistics, linguistics in the Arab world reached a technically much more
‘modern’ level in a remarkably brief space of time—but, like a prematurely blossoming flower, withered away just as its
fellows were catching up.
How the study of grammar developed in its very earliest stages in the Arab world is likely to remain a mystery. Two factors
played into it, one internal, the other external. Within Islamic culture great importance was attached to the Qur’ān, to the
preservation of the text and the exegesis of its contents. As Islam spread to non-Arabic-speaking peoples the need for formal
grammatical instruction became urgent, all the more so since the script did not indicate the vowels, nor, in its earliest form,
did it distinguish among several of the consonants. Without a thorough knowledge of the language there could be no question
of correct oral recitation of the Qur’ān, a vital part of religious ceremony. This was the motivation attributed to the reputed

inventor of Arabic grammar, Abū ’l-Aswad ad-Du’alī (who died c. 688). Many of the most noted grammarians of Arabic,
starting with Siṱ bawayhi himself, were of non-Arab (often Persian) or mixed descent. Had it been simply a matter of
vocalising the text, Arab linguistic scholarship need not have developed further, even as the Massoretes concentrated on the
vocalisation of Hebrew and allied problems connected with recitation of the scriptures, but went no further. But the needs of
converts speaking unrelated languages, notably Persian, and the possibilities presented by the lively intellectual surroundings
at centres such as Nisibis, Jundishapur, and later Baghdad, contributed to the unprecedented speed with which the Arab
grammatical tradition emerged. Greek was still used in parts of the Near East in the first centuries of Islam, and much Greek
material, including a part of the grammar attributed to Dionysius Thrax, had been translated into Syriac, a Semitic language with
widespread currency in the Near East. Although the extent of Greek influence on the foundation of the Arab grammatical
tradition is still disputed, the existence of contact between the two traditions is not at issue, and details such as the choice of a
model verb meaning ‘to beat’ in both traditions—ḍaraba in Arabic, túptō in Grteek—corroborate this. More problematical is
the question of dependence at the systemic level. Without parallel in the other major linguistic traditions is the importance of
the terminology and concepts of Islamic law in early Arabic grammar.
446 LANGUAGE AND ITS STUDENTS
The earliest grammar to have come down to us from the Arab tradition is the Kitāb (‘Book’) by Siṱ bawayhi (who died in
793). Perhaps its most striking feature to the reader familiar with the Greco-Roman tradition is the minor importance of
semantic issues. The introduction begins by setting out the three word classes, names (ism), operations/actions ( ) and words
‘directed towards a meaning’ which are neither names nor operations (ḥarf). Names require no explanation, and are given none;
operations take various forms to specify the actions of the names, and indicated what has passed, or what will be but has not yet
taken place, or that which is and has not yet finished. Much of the discussion which follows, on
, broadly ‘inflection’, is form-centred to an extent unknown in the West. Instead of identifying functional, semantic or logical
categories such as ‘subject’ or ‘nominative’ or ‘first person singular’, and listing the different forms by which they may be
realised, Sībawayhi identifies different formal states and processes, and exemplifies each one as it manifests itself in words of
different classes. Thus, ‘stabilisation’ (or ‘taking an -a(n) ending’) occurs in both the ism and the : ra’aytu Zaydan ‘I saw
Zaid’ and lan
‘he will not act’; but ‘stretching’ (taking an -i(n) ending) occurs only in the ism, as in marartu bi-Zaydin ‘I passed Zaid’, and
‘amputation’ (apocope) only in the , as in lam
‘he has not done’. Had a Roman grammarian wanted to work in the same way, he would have decided not to discuss the parts
of speech in turn, but instead to treat all the forms ending in the same letter together; those ending in -a, for example, would
include verbs in the imperative singular, nouns, pronouns, adjectives and participles in the nominative and ablative singular,

adverbs and prepositions. Although Siṱ bawayhi works with notions like number, person and gender, they are to a large
extent taken for granted: the emphasis is on form.
A natural consequence of a form-orientated linguistics is the handling of syntactic relations in terms of a change in form
effected by an operation (
), whether implicit or explicit. Thus, the sentence yaqūmu Zaydun ‘Zayd stands’ contains two independent elements, neither
entailing a relationship of dependency, whereas in inna Zaydan lan yaqūma ‘truly Zayd will not stand’, ‘Zayd’ is made
dependent by inna ‘truly’ and yaqūma ‘he will stand’ by lan ‘not’. The parallels with modern dependency grammar are more
far-reaching than this simple example suggests.
A further manifestation of the Arab preoccupation with linguistic form is the profound interest in articulatory phonetics
apparent from the very start of the Arab tradition. Both Sībawayhi and al-Hṱ alīl (who died in 791), the compiler of a huge
dictionary, provided a classification of the consonantal sounds of Arabic based on the maḥ rag̣, ‘area of emission’, a
broader notion than the Sanskrit sthāna and karaṇa, involving the specification of the point in the vocal tract at which the
airstream is interrupted. Sībawayhi’s list of sixteen maḥ ārig̣ was gradually reduced to five, which passed into the Hebrew
and thence into the Western tradition. Recognition of nasality goes back to Siṱ bawayhi himself, who offers the time-
honoured test: ‘if you hold your nose, the nasals cannot be produced.’ Ibn Sinā, known to the West as Avicenna (980–1037),
included among his voluminous writings a treatise on the ḥurūf (the Arabic term ḥarf, pl. ḥurūf, functions like the Latin
littera) which contained a detailed description of the anatomy and function of the larynx. The lively Islamic tradition of
phonetic investigation, represented by a large number of writers from Sībawayhi on, was paralleled by a no less active interest
among Hebrew grammarians.
The Arab interest in form and in physical phenomena was a characteristic rooted deeply in their culture, its manifestations
visible not only in linguistics but throughout their intellectual life. It was from outside Islam that the impetus came to consider
language from other points of view. In the second half of the eighth and during the ninth century many Greek texts were
translated into Arabic (sometimes via Syriac), including the Aristotelian corpus. Under their influence, and particularly among
grammarians belonging to the Mu
c
tazilite sect, a more philosophical approach to grammar took root, exemplified in the works
of al-Fārābī (who died in 950) and az-Zağğāğī (who died in 949 or 951). Logic and grammar were compared and their
respective domains delimited, and attention was given to such matters as the correct way to frame a definition, at
approximately the same time as the assimilation of Aristotelian logic was causing a preliminary reassessment of the role of
grammar in the West. Thus, later grammars such as Ibn Aṱ jurrūm’s (the famous Ạ jurrūmiyya) define the utterance as

having form, composite, informative and conventional: the Aristotelian elements are obvious. (A further parallel with the
development of grammar in the West is the popularity in the later Middle Ages of grammars in verse such as the Alfiyya by
Ibn Mālik.)
AN ENCYCLOPAEDIA OF LANGUAGE 447
The importance attached to the formal aspects of language is a natural consequence of the reasoning that, since language
mirrors the world, it must be governed by the same laws. Since all phenomena perceived by the senses could be described in
quantitative terms, so too could language. Grammarians sought regularities in language, particularly through observing qiyās,
‘analogy’, while natural philosophers, most notably the Pythagorean-influenced Jābir ibn Hṱayyān, traced the proportions
found in natural phenomena in language. Whereas Western pre-Renaissance linguistics stressed those aspects of language
which participated in the eternal and spiritual, Islamic scholars devoted equal attention to its earthly side. And yet by the time
Western scholars were ready to appreciate the Arab approach, it had lost its impetus. The awakening to form experienced by
Renaissance Europeans had nothing new to offer the Arabs. Arab linguists had travelled the same path centuries before; aware
of the importance of form from the very start of their tradition, they had exhausted its possibilities. The Renaissance brought
them no fresh tools to use. They were the lenders rather than the borrowers, contributing a few fundamental concepts from
their rich stock to the founding of the Western disciplines of phonetics and morphology. They had much more to offer, but
Western scholars chose to go their own way, arriving at the same insights in their own time, using their own methods.
3.2
India
The study of language in India arose out of the needs of religious ritual, and remained closely linked with religion for most of
its history. At the most profound level the Vedic hymns acknowledge the mysterious nature and power of speech: ‘Speech
was divided into four parts known to inspired priests. Three parts, placed in hiding, mortals do not rouse to action; the fourth
part of speech they speak’ (Ṛgveda I 164.45). And yet, despite their fascination with the spiritual nature of speech, Indians
concerned themselves with its physical aspect far earlier than Europeans did. The immediate impetus came from the practical
needs of religious ritual. The Vedic hymns and prayers used in Brahmin rituals had been given definitive form near the start
of the first millennium BC. Handed down from father to son, the text inevitably began to sound first old-fashioned, and then
downright archaic, as the Sanskrit language gradually underwent change. Instead of updating the text, the Brahmins felt it
imperative to maintain its traditional form as accurately as possible. To counteract the sloppy pronunciation of the younger
generation—the usual response the world over to the effects of language change—they instituted formal training, supported
by lists of the legitimate assimilatory features (sandhi) found in each collection of Vedic hymns, prefaced (or finished) by a
summary of the theoretical notions underlying the practical instruction. Although very brief, usually only three or four pages

in length, the outlines contained in these texts (called prātiśākhyas) amount to an introduction to the principles of articulatory
phonetics. They centre on the Sanskrit alphabet, itself a masterpiece of phonemic analysis and articulatory systematisation:
a ā i ī ū ū rṱ rṱ ṱ lṱ
e ai o au
k kh g gh nṱ
c ch j jh ñ
tṱ tṱh dṱ dṱh nṱ
t th d dh n
p ph b bh m
y r l v
ś sṱ s h
This order facilitates recognition of four basic categories: vowels (subdivided into monophthongs and diphthongs), plosives
+nasals (the ‘five groups of five’), glides and fricatives. These categories were described and subdivided with the aid of
various subsidiary notions: articulator (karaṇa) and place of articulation (sthāna), voicing, aspiration, nasality, degree of
closure. The systematic nature of Indian phonetic analysis—arrived at somewhere in the first half of the first millennium BC—
provided Western scholars from about 1800 with the key to handling features such as voicing and aspiration in an orderly
classificatory matrix.
The same attention to the physical aspect of speech is apparent in the oldest and best-known Sanskrit grammar to come
down to us, Pānṱini’s Aṣṭādhyāyī (‘Eight Books’), written in the fifth or fourth century BC. The Aṣṭādhyāyī is very different
from the hierarchically-structured grammars of Greco-Roman Antiquity, or even from Varro’s more fluid characterisations.
Instead of allowing the structure perceived in the language to dictate the overall shape of the work, as Donatus, for instance,
does in beginning with the smallest units of speech and building up to the largest, Pānini attaches overriding importance to the
presentation, the outer form, of the doctrine. Economy of expression dominates: not only does he use a large number of
technical abbreviations and a telegraphic metalanguage (which has to be mastered by the student at the outset), but he also
employs a dittoing procedure which enables him to avoid repeating a rule which applies to many different types of linquistic
phenomena—at the expense of assembling together rules which have nothing more in common than being affected by this more
powerful rule. Each rule is given equal prominence, down to those which govern a single word; the concept of ‘exception’, so
448 LANGUAGE AND ITS STUDENTS
integral a part of the Western tradition, is absent. Consequently, it is very difficult to form any impression of the structure of
Sanskrit from the Aṣṭādhyāyī without protracted study. Traditionally, the young child would spend eight years memorising it

before being shown how to apply its doctrine, for in order to generate even the simplest verb form, rules from all parts of the
work must be applied. Not surprisingly, both commentaries, like Patanṱ jali’s huge and detailed Mahābhāṣya, (c. 150 BC),
and simplified and rearranged versions of the Aṣṭādhyāyī, like Śarvavarman’s Kātantra (before AD 400) and the Siddhānta
Kaumudī by Bhatṱtṱoji Dīksṱita (early seventeenth century), found large numbers of readers.
Underlying Pānṱini’s work, and indeed that of all Indian grammarians, is the notion of ‘root’ as an abstract entity which
never—or very rarely—manifests itself in actual speech, but provides the raw material which, after the operation of various
phonological changes, becomes a base to which affixes may be added, eventually culminating in the spoken word. So much a
part of the linguistic tradition was this that lists of roots were compiled from Pānṱini on. Pānṱini’s treatment of morphology is
therefore utterly different from that current in the ancient West. He uses no paradigms; instead, he relies on complex
sequences of generative rules. To take a somewhat simplified example, to form the third person singular of the present tense of
the verb ‘go’, ̣ gam, we have to follow a procedure which involves the addition of a conventional marker indicating a present
tense form: laṭ. A condition for the addition of laṭ to verbs is the insertion of an infix between root and suffix. In the case of
roots of the class to whicḥ gam belongs, this infix is -a Replacing the present tense marker with the third person singular
ending -ti gives *gamati, not yet a Sanskrit word. One further rule, prescribing the change of roots of the ̣ gam type when
followed by a laṭ suffix, remains to be applied, producing the actual Sanskrit form, gacchati. Less ponderous—once it has
been mastered—than it sounds, this method provides elegant descriptions of much of the luxuriant morphology of Sanskrit. It
is rather less appropriate for foreign beginners, however (and it is interesting to note that whereas those European grammarians
of Sanskrit who had had the advantage of instruction from a native pandit adhered as closely as they could to the basically
Paninian system, scholars based in Europe—notably Frank (1823) and the enormously influential Bopp (1824–37)—
abandoned much of the apparatus of the Indian system of description, preferring to adapt the traditional framework of Greek
and Latin grammar to accommodate Sanskrit.)
Accompanying Pānṱini’s keen sense for the formal structure of language was a precise awareness of the separate domains of
form and meaning. One feature of Indo-European languages where their relationship has repeatedly challenged grammarians
is that of the nature of case: to what extent can the changes in the form of a noun as it discharges different grammatical
functions in a sentence be equated with semantic or real-world categories? In the traditional Western analysis of a pre-sandhi
sentence like Devadattaḥ odanam pacati, ‘Devadatta cooks rice’, Devadattaḥ is described as the subject of the active verb
pacati, ‘cooks’, and therefore has the nominative, or subject, case-form; but in the corresponding passive sentence odanaḥ
Devadattena pacyate, ‘rice is being cooked by Devadatta’, odanaḥ ‘rice’, is now in the nominative case, even though its
relation to the act of cooking has not changed. Pānṱini circumvents the difficulty by setting up a series of categories, kārakas,
which, although ultimately based on the syntactic relations through the cases, are semantic notions which can be manifested in

many different ways: through a verb (active, passive, in the appropriate mood and tense), a preposition, or by word-formation,
as well as by case endings. In our first example, the agent kāraka is expressed in the active verb form, pacati; the name of the
actual agent therefore need not express agency (since a kāraka expressed in the verb should not be given double expression),
but instead takes the umarked, or default, form, the nominative case. In the second sentence, it is the object kāraka which is
expressed by the verb pacyate, ‘is being cooked’, and therefore ‘rice’, odanaḥ, is now left unmarked (i.e. nominative); the
agent kāraka, on the other hand, now has to be expressed explicitly, this time through the instrumental case form,
Devadattena. Although far more comprehensively elaborated, kāraka theory has much in common with the medieval Western
doctrine of the modi significandi: both attempt to explain how it comes about that the same semantic content can be
manifested in different lexical and syntactic structures within a given language. Parallels may be drawn too with Fillmore’s ‘deep
cases’ and with Tesnière’s dependency syntax. (See Chapter 3, section 5, above).
As in the Western tradition, the problem of sentence meaning was tackled largely by philosophers rather than by
grammarians, although occasionally, as in the case of Patañjali and Bhartrṱhari, the two provinces coincided in one man. The
Vedas again provided the focus: given that their very words were regarded as eternal and uncreated, their nature and status
required investigation. The Mīmāmṱsā school (third to second century BC) sought a philosophical basis for the Vedas in which
linguistic issues played a large part. Its members propounded two opposing views of sentence meaning, both of which took
the word as the unit. According to the first theory, sentence meaning was linear and cumulative, based on the word, an
autonomous unit of thought and sense. Sentence meaning was simply the sum of the individual word meanings. According to
the second, word meaning was defined by sentence meaning. Only when the listener had grasped how ‘cow’ and ‘horse’ stood
in paradigmatic relation to one another in sentences like ‘bring the cow’ and ‘bring the horse’ could he attach any meaning to
the isolated words ‘cow’ and ‘horse’. These two views were in time overshadowed by the much more thorough-going work of
Bhartrṱhari (c. 450–510). In his Vākyapadīya he attempted to dislodge his fellow-grammarians from their complacent
preoccupation with methodological technicalities by reminding them of the fundamental role of language and of grammar:
‘just as all the universals of things depend upon the form of their words for their communication, so is this science the basis of
all other sciences’ (I 15). No knowledge is possible without the organising, sequential properties of language, so closely
AN ENCYCLOPAEDIA OF LANGUAGE 449
linked with memory. Even the child has residual traces of a knowledge of the word from its previous existence, on the basis
of which both cognition and the use of the speech organs can arise—neither of them faculties which can be taught. But what
is the nature of the words that one utters? ‘Just as light has two powers, that of being revealed and that of being the revealer,
similarly, all words have two distinct powers’ (I 55). A particular relationship obtains between individual sounds and that
which is revealed through them, the sphoṭa. The speech-essence, the source of the world of reality, goes through three stages

in becoming manifest in human speech: the stage of indivisible reality or consciousness without extension in time, in which
an idea occurs in a flash, complete; the intermediate stage in which the idea is set out sequentially, becoming a logically-
organised linear thought capable of being verbalised (and already possessing certain features characteristic of a particular
language); and the stage of producing audible words marked by the individual idiosyncrasies of the speaker. The total
meaning of a sentence can no more be explained in terms of its parts than the meaning of a picture can be explained by listing
the different colours used in it. ‘Just as the meaning of the word is not understood from each sound, in the same way the meaning
of the sentence is not understood from each word’ (II 60). Meaning itself is indivisible: it bursts forth from the spoken
utterance in a flash of intuition. Words, roots, suffixes, individual speech sounds, are an analytical convenience, a
grammarian’s fiction. By viewing utterance meaning as unitary and primary, and not dependent on individual words,
Bhartrṱhari provided the basis for a very powerful theory of meaning—one capable of dealing with ungrammatical utterances,
with foreigners’ grotesque but nevertheless comprehended attempts to communicate, with figurative language, and indeed
with the suggestive power of language. Bhartrṱhari’s successors concentrated on figurative language, applying his teachings
and their own to the study of Sanskrit poetry. In the ninth century Aṱ nandavardhana developed a more comprehensive
approach to the suggestive power of language in his Dhvanyāloka, studying factors such as intonation, emphasis, gesture, tone
of voice, and sociocultural elements peculiar to a particular speech community. Traditional Indian semantics is now
undergoing lively investigation and development by Indian scholars versed in both Indian and Western semantics.
3.3
China
In one respect the ancient Chinese world-view was strikingly similar to that of the pre-modern West, and in another
unexpectedly dissimilar; and in each case this is reflected in their views on language. Perhaps the most conspicuous feature of
the Chinese outlook is a belief in order. The universe was an ordered system, and everything within it had its place—heaven,
the gods, the planets, man, the lower orders. To maintain this highly structured, hierarchical system Chinese society had recourse
to the Confucian system of ethics. This system permeated the whole of society: nothing was without its niche, everything had
its own meaning and purpose. It was incumbent upon human beings to discover the meaning of the phenomena around them.
The second essential trait of the Chinese outlook, one which is visible as clearly in medicine as in linguistics, is a desire to
function on two levels: broad, universal principles applicable to every sphere of existence, and individual, particular
phenomena. Intermediate levels—generalisations relevant to a particular class of phenomena, for instance—were not
investigated. These views naturally had their consequences both in the choice of fields of study and in the way in which they
were pursued. If the aim of study is to discover the meaning of a phenomenon, then it will be inappropriate to remain upon the
level of the phenomena themselves. Explanations must be sought on a higher level (unless the seeker is to revert to

reductionism). Hence, in China as in the medieval West, astronomy merged into astrology and biology into the teleological
and moralising remarks of medieval bestiaries. Chinese thinkers observed processes on one level, the physical, but sought
explanations on another.
This attitude naturally had implications for the study of language. There was no notion of studying language ‘in and for itself;
only those aspects were studied which were compatible with the Chinese world-view. So, for example, there are no
comprehensive descriptions of Chinese grammar. Instead, language was used to fortify the ethical bases of Chinese society.
For example, Confucius, when asked about the true nature of a government, replied that it was of the essence of a government
(zheng) to be upright (zheng): the semantic connection reinforces the natural order of things, as in the medieval Western
association of reges with recte agenda. Argumentation of this kind plays a large part in early Chinese ethics and philosophy,
notably in the works of Confucius and Mencius. Essentially, the question asked was ‘what light can language shed on the
world?’ That part of language most likely to illuminate the world was its meaningful aspect.
Consequently, it was the word which was the focus of Chinese linguistics— the word as the bearer or embodiment of
meaning. Levels below the word— morphology and phonetics—could contribute nothing (phonology was a special case
which we will consider later), while the study of meaning at a higher level was carried out by literary critics. Syntax was not
pursued. Where the word as a meaningful entity is the centre of attention, then linguistic study will tend to focus on word-
lists: vocabularies, glossaries, thesauruses, dictionaries. Lexicography and dialect studies, also concentrating on lexis, were
the most highly developed branches of linguistics in pre-modern China.
The earliest dictionary, ascribed to a disciple of Confucius, Zi Xia, is the Er Ya (‘Treasury of Fine Words’, between the
third and first centuries BC). It is divided into nineteen sections, most of which are semantic categories: names, idioms,
450 LANGUAGE AND ITS STUDENTS
difficult words, family relationships, occupations, tools, music, sky (both the calendar and climate), land, hills, mountains,
water, plants, trees, insects and reptiles, fish, birds, wild animals, domestic animals. Any semantic classification suffers from
certain limitations. Many words do not lend themselves to categorisation on this basis, a problem which the Er Ya
acknowledged by reserving several sections for miscellaneous terms and words used figuratively. Perhaps more to the point,
what of the difficulties faced by the reader of an unfamiliar text who wants to look up a strange character? To be told to look
it up in a semantically-classified work puts him a similar plight to that of the Western child who asks how to spell a word and
is told to look it up in an alphabetically-ordered dictionary. A path to a solution was suggested by the most famous of all
Chinese dictionaries, the Shuo Wen Jie Zi (‘The Explanation of Simple and Derived Characters’) (c. 100 AD). As its name
suggests, the focus of classification has shifted from the meanings to the characters. Characters were assigned to one or
another of six categories on the basis of their construction—iconic, symbolic, phonetic and phonetic loan, semantic

compound, modified—and the semantic common factor, or radical, was isolated and used as the criterion of classification.
The original list of 540 radicals was later reduced to 360, and finally to the list of 214 which is still in use today. Dictionaries
which rely on radicals as their ordering criterion still have much in common with thesauruses, for all the characters which
share the same radical will come together. Thus, all the characters with the ‘man’ radical come together, ensuring a bunch of
words denoting human beings, and similarly words denoting liquids are grouped together because they have in common the
‘water’ radical, and so on. Under each radical characters are listed in order according to the number of strokes additional to
the radical they have, making it possible to locate unfamiliar ones. Only in the present century, under Western influence, have
Chinese lexicographers experimented with phonetic principles.
The celebrated distinction between ‘full’ and ‘empty’ words, taken over into Western linguistics as a straightforward
opposition between words with lexical meaning and words serving only to express grammatical relationships, was in its
original context a semantic distinction between words denoting objective concepts or things, and those which were essentially
subjective, expressing the feelings and responses of the individual. This category included intensifies, modal particles and also
the verb ‘to be’ (which has greater assertive force in Chinese than in English).
One problem which Chinese scholars grappled with from very early times was that of indicating pronunciation in a script in
which each character represented an entire word. For dialect studies in particular the lack of any means of phonetic
representation was a significant handicap. Early works resorted to homophony, saying that character x sounded like character
y. Some time before about 600 AD a more satisfactory technique was developed. Called fanqie, it involved the analysis of the
syllable into two segments, (a) the initial vowel or consonant plus (b) the rest of the syllable, including the tone. Another
syllable beginning with the same initial was found to represent (a), and one finishing with the same features as (b) to
represent the second half. Thus, dong could be represented by de hong. The system was further refined to permit the
indication of suprasegmental phenomena such as palatalisation. The fanqie analysis was structural in nature: it works with
relationships between entities rather than with the entities in their own right, a procedure also characteristic of Chinese
medicine. Not until the basic tenets of Indian teaching reached China with Buddhism, from the seventh century, was a kind of
articulatory phonetics elaborated. Initial consonants were divided into five categories according to place of articulation: lips,
tongue (=dentals), incisors (=dental affricates), molars (=velars), throat (=laryngeals). Although degrees of voicing were
recognised, palatalisation—not catered for by the Indian system—was ignored. Interestingly, although a popular version of
Indian grammar was introduced at the same time, it failed to undergo further development. Only with the arrival of Western
grammar at the end of the nineteenth century did Chinese scholars give serious consideration to the grammar of their own
language. The celebrated product of this meeting was Ma Jian Zhong’s Ma Shi Wen Tong (1898), a grammar which sought to
apply ‘universal’ (i.e. Latin) categories as closely as possible to Chinese. Only in the present century—when political

vicissitudes have permitted—have scholars such as Wang Li attempted to devise a more sensitive grammatical structure for
Chinese, a process in which many Western scholars, particularly in America, have joined.
REFERENCES
Where these are simply evaluated within particular approaches they appear in the text above. Where they are more generally
useful they are listed below (with brief references in the text).
GENERAL REFERENCES AND FURTHER READING
Items are presented in relation to the places and eras surveyed. A reader’s further study is best guided by the titles in each
section.
AN ENCYCLOPAEDIA OF LANGUAGE 451
I.
General (including works relevant to several sections)
Amirova, T.A., Ol’chovikov, B.A. and V.Roždestvenskij, Ju. (1980) Abriss der Geschichte der Linguistik, transl. from the Russian (1975)
by Meier, B., VEB Bibliographisches Institut, Leipzig.
Arens, H. (1st ed. 1955, 2nd ed. 1969) Sprachwissenschaft: Der Gang ihrer Entwicklung von der Antike bis zur Gegenwart 2 vols.,
Athenäum Fischer, Frankfurt.
Asher, R.E. and Henderson, E.J.A. (eds) (1981) Towards a History of Phonetics, The University Press, Edinburgh.
Auroux, S. et al. (1985) La linguistique fantastique, Joseph Clims, Paris.
Borst, A. (1957–63) Der Turmbau von Babel: Geschichte der Meinungen über Ursprung und Vielfalt der Sprachen und Völker 3 vols.,
Anton Hiersemann, Stuttgart.
Bynon, T. and Palmer, F.R. (eds) (1986) Studies in the History of Western Linguistics in Honour of R.H.Robins, Cambridge University
Press, Cambridge.
Foucault, M. (1966) Les mots et les choses, Gallimard, Paris, transl. into English as The Order of Things: An Archaeology of the Human
Sciences (1970), Tavistock, London.
Histoire Epistémologie Langage (1979–).
Historiographia Linguistica (1974–).
Hymes, D. (1974) Studies in the History of Linguistics: Traditions and Paradigms, Indiana University Press, Bloomington, Ind.
Koerner, E.F.K. (1978) Western Histories of Linguistic Thought: An Annotated Chronological Bibliography 1822−1976, Studies in the
History of Linguistics 11, Benjamins, Amsterdam.
Parret, H. (1976) History of Linguistic Thought and Contemporary Linguistics, Walter de Gruyter, Berlin and New York.
Robins, R.H. (1st ed. 1967, 2nd ed. 1979) A Short History of Linguistics, Longman, London.

Schmitter, P. (1982) Untersuchungen zur Historiographie der Linguistik: Struktur—Methodik —theoretische Fundierung, Tübinger
Beiträge zur Linguistik 181, Gunter Narr, Tübingen.
Sebeok, T.A. (1975) Current Trends in Linguistics 13: Historiography of Linguistics, 2 vols., Mouton, The Hague.
Todorov, Ts. (1972) ‘Le sense des sons’, Poétique 11: 446–62.
II.
Greco-Roman Antiquity
Allen, W.S. (1981) ‘The Greek contribution to the history of phonetics’, in Asher and Henderson (1981):115–22.
Arens, H. (1984) Aristotle’s Theory of Language and its Tradition: Texts from 500 to 1750, Studies in the History of Linguistics 29, Benjamins,
Amsterdam.
Baratin, M. and Desbordes, F. (1981) L’analyse linguistique dans l’Antiquité classique 1. Les théories, Klincksieck, Paris.
Collart, J. (1954) Varron grammairien latin, Publications de la Faculté des Lettres de l’Université de Strasbourg 121, Les Belles Lettres,
Paris.
Coseriu, E. (1975) Die Geschichte der Sprachphilosophie von der Antike bis zur Gegenwart: eine Übersicht, 1. Von der Antike bis Leibniz,
2nd ed., Tübinger Beiträge zur Linguistik 11, Gunter Narr, Tübingen.
Derbolav, J. (1972) Platons Sprachphilosophie im Kratylos und in den späteren Schriften, Wissenschaftliche Buchgesellschaft, Darmstadt.
Holtz, L. (1981) Donat et la tradition de I’enseignement grammatical: étude sur l’Ars Donati et sa diffusion (IVe-IXe siècle) et édition
critique, CNRS, Paris.
Householder, F.W. (1981) The Syntax of Apollonius Dyscolus translated, and with commentary, Studies in the History of Linguistics 23,
Benjamins, Amsterdam.
Hovdhaugen, E. (1982) Foundations of Western Linguistics: From the Beginning to the End of the First Millennium AD,
Universitetsforlaget, Oslo.
Keil, H. (1855–80) Grammatici Latini, 7 vols+suppl., Teubner, Leipzig, repr. 1981, Georg Olms, Hildesheim.
Kemp, A. (1986) ‘The Tekhnē grammatikē of Dionysius Thrax translated into English’, Historiographia Linguistica, 13:343–63.
Law, V. (1986) ‘Late Latin grammars in the early Middle Ages: a typological history’, Historiographia Linguistica, 13:365–80.
McKeon, R. (1946–7) ‘Aristotle’s conception of language and the arts of language’, Classical Philology, 41:193–206 and 42:21–50.
Pinborg, J. (1975) ‘Classical Greece’, in Sebeok, T.A. (1975):69–126.
Schmidt, R.T. (1979) Die Grammatik der Stoiker, transl. from the Latin (1839) by Hülser,
K., bibliography by Egli, U., Schriften zur Linguistik 12, Vieweg, Braunschweig and Wiesbaden.
452 LANGUAGE AND ITS STUDENTS
Steinthal, H. (1st ed. 1863, 2nd ed. 1890–1) Geschichte der Sprachwissenschaft bei den Griechen und Römern mit besonderer Rücksicht

auf die Logik, Ferdinand Dümmler, Berlin, repr. Georg Olms, Darmstadt 1961.
Uhlig, G. (1883–1910) Grammatici Graeci, Teubner, B.G. Leipzig, repr. Georg Olms, Heidelberg (1979).
Wouters, A. (1979) The Grammatical Papyri from Graeco-Roman Egypt: Contributions to the Study of the ‘Ars grammatica’ in Antiquity,
Verhandelingen van de Koninklijke Academic voor Wetenschappen, Letteren en Schone Kunsten van België, Klasse der Letteren 41.
III.
The Middle Ages
Ahlqvist, A. (ed.) (1987) Les premières grammaires des vernaculaires européens (=Histoire Epistémologie Langage 9.1).
Bursill-Hall, G.L. (1975) ‘The Middle Ages’, in Sebeok, T.A. (1975):179–230.
Bursill-Hall, G.L. (1981) A Census of Medieval Latin Grammatical Manuscripts, Grammatica Speculativa 4, Frommann-Holzboog,
Stuttgart-Bad Cannstatt.
Covington, M.A. (1984) Syntactic Theory in the High Middle Ages: Modistic Models of Sentence Structure, Cambridge Studies in
Linguistics, 39, Cambridge University Press, Cambridge.
Covington, M.A. (1986) ‘Grammatical theory in the Middle Ages’, in Bynon and Palmer (1986):23–42.
Hunt, R.W. (1980) Collected Papers on the History of Grammar in the Middle Ages, Studies in the History of Linguistics, 5, Benjamins,
Amsterdam.
Law, V. (1982) The Insular Latin Grammarians, Studies in Celtic History 3, The Boydell Press, Woodbridge.
Law, V. (1985) ‘Linguistics in the earlier Middle Ages: the Insular and Carolingian grammarians’, Transactions of the Philological Society:
171–93.
Pinborg, J. (1967) Die Entwicklung der Sprachtheorie im Mittelalter, Aschendorff, Münster and Frost-Hansen, Copenhagen.
Rosier, I. (1983) La grammaire spéculative des Modistes, Presses Universitaires de Lille, Lille.
IV.
Since the Renaissance
Aarsleff, H. (1976, repr. with corrections 1983) The Study of Language in England, 1780− 1860 University of Minnesota, Minneapolis and
Athlone, London.
Aarsleff, H. (1982) From Locke to Saussure: Essays on the Study of Language and Intellectual History, Athlone, London.
Alston, R.C. (1965–73) A Bibliography of the English Language from the Invention of Printing to the Year 1800, 11 vols., E.J.Arnold,
Leeds and Bradford.
Anthropological Linguistics 5.1. (1963): History of Linguistics.
Apel, K.O. (1963) Die Idee der Sprache in der Tradition des Humanismus von Dante bis Vico, Archiv für Begriffsgeschichte 8, Bouvier,
Bonn.

Benfey, Th. (1869) Geschichte der Sprachwissenschaft und orientalischen Philologie in Deutschland seit dem Anfange des 19.
Jahrhunderts mit einem Rückblick auf die früheren Zeiten, Geschichte der Wissenschaften in Deutschland, Neuere Zeit 8, Cotta,
Munich, repr. 1965.
Beyer, A. (1981) Deutsche Einflüsse auf die englische Sprachwissenschaft im 19. Jahrhundert, Göppinger Arbeiten zur Germanistik No.
324, Kümmerle, Göppingen.
Brunot, F. (1966–) Histoire de la langue française des origines à nos jours, revised edition, Armand Colin, Paris.
Carvalhâo Buescu, M. (1983) O estudo das línguas exóticas no século XVI, Biblioteca breve, Serie pensamento e ciência 71, Lisbon.
Carvalhâo Buescu, M. (1983) Babel ou a ruptura do signo: a gramática e os gramáticos portugueses do século XVI, Imprensa Nacional,
Lisbon.
Dobson, E.J. (1st ed. 1957, 2nd ed. 1968) English Pronunciation 1500−1700 I: Survey of the Sources, Clarendon Press, Oxford.
Droixhe, D. (1978) La linguistique et l’appel de l’histoire, 1600−1800: rationalisme et révolutions positivistes, Droz, Geneva.
Dubois, C G. (1970) Mythe et langage au seizième siècle, Ducros, Bordeaux.
Formigari, L. (1970) Linguistica ed empirismo nel Seicento inglese, Laterza, Bari.
Gipper, H. and P.Schmitter (1979) Sprachwissenschaft und Sprachphilosophie im Zeitalter der Romantik: ein Beitrag zur Historiographie
der Linguistik, Tübinger Beiträge zur Linguistik 123, Gunter Narr, Tübingen (revised version of their article in Sebeok (1975):481–
606).
Hankamer, P. (1927) Die Sprache, ihr Begriff und ihre Deutung im sechzenhnten und siebzehnten Jahrhundert: ein Beitrag zur Frage der
literaturhistorischen Gliederung des Zeitraums, Bonn, repr. 1965, Georg Olms, Hildesheim.
AN ENCYCLOPAEDIA OF LANGUAGE 453
Hanzeli, V.E. (1969) Missionary Linguistics in New France: A Study of Seventeenth- and Eighteenth-Century Descriptions of American
Indian Languages, Mouton, The Hague.
Hymes, D. and J.Fought (1981) American Structuralism, Mouton, The Hague (revised version of their article in Sebeok (1975):903–1176).
Jankowsky, K.R. (1972) The Neogrammarians: A Re-evaluation of their Place in the Development of Linguistic Science, Mouton, The
Hague.
Jellinek, M.H. (1913–14) Geschichte der neuhochdeutschen Grammatik von den Anfängen bis auf Adelung, 2 vols., Carl Winter,
Heidelberg.
Kayser, W. (1930) ‘Böhmes Natursprachenlehre und ihre Grundlagen’ Euphorion, 31: 521–62, transl. into French by J.Launay, ‘La doctrine
du langage naturel chez Jacob Boehme et ses sources’, Poétique, 11 (1972):337–66.
Knowlson, J. (1975) Universal Language Schemes in England and France 1600−1800, University of Toronto Press, Toronto and Buffalo.
Lehmann, W.P. (1967) A Reader in Nineteenth-Century Historical Indo-European Linguistics, Indiana University Press, Bloomington, Ind.

Lepschy, G. (1970) A Survey of Structural Linguistics, André Deutsch, London.
Lepschy, G. (1986) ‘European Linguistics in the twentieth century’, in Bynon and Palmer (1986):189–201.
Michael, I. (1970) English Grammatical Categories and the Tradition to 1800, Cambridge University Press, Cambridge.
Muller, J C. (1986) ‘Early stages of language comparison from Sassetti to Sir William Jones (1786)’, Cratylus 31:1–31.
Padley, G.A. (1976) Grammatical Theory in Western Europe 1500−1700: The Latin Tradition, Cambridge University Press, Cambridge.
Padley, G.A. (1st ed. 1985, 2nd ed. 1988) Grammatical Theory in Western Europe 1500−1700: Trends in Vernacular Grammar I and II,
Cambridge University Press, Cambridge.
Pedersen, H. (1931) The Discovery of Language: Linguistic Science in the Nineteenth Century, transl. from the Danish (1924) by Spargo,
J.W., Harvard University Press, Cambridge, Mass., repr. 1962, Indiana University Press, Bloomington.
Percival, W.K. (1975) ‘The grammatical tradition and the rise of the vernaculars’, in Sebeok, T.A. (1975):231–75.
Percival, W.K. (1984) ‘The reception of Hebrew in sixteenth-century Europe: the impact of the Cabbala’, Historiographia Linguistica 11:
21–38.
Ricken, U. (n.d.) Grammaire et philosophie au siècle des Lumières, Publications de l’Université de Lille III, Lille.
Rousseau, J. (1984) ‘La racine arabe et son traitement par les grammairiens européens (1505–1831)’, Bulletin de la Société de Linguistique
de Paris 79:285–321.
Salmon, V. (1979) The Study of Language in 17th-Century England, Studies in the History of Linguistics 17, Benjamins, Amsterdam.
Salmon, V. (1986) ‘Effort and achievement in seventeenth-century British linguistics’, in Bynon and Palmer (1986):69–95.
Stengel, E. (1976) Chronologisches Verzeichnis französischer Grammatiken vom Ende des 14. bis zum Ausgange des 18. Jahrhunderts,
nebst Angabe der bisher ermittelten Fundorte derselben, new ed. with supplement by H J.Niederehe, Studies in the History of
Linguistics 8, Benjamins, Amsterdam.
Tavoni, M., ed. (1st ed. 1987, 2nd ed. 1988) Renaissance Linguistics Archive 1350−1700: A First (Second) Print-Out from the Secondary-
Sources Data-Base, Istituto di Studi Rinascimentali, Ferrara.
Trabalza, C. (1908) Storia della grammatica italiana, Ulrico Hoepli, Milan, repr. 1963, Arnaldo Ferrari, Bologna.
Verburg, P.A. (1952) Taal en Functionaliteit: een historisch-critische studie over de opvattingen aangaande de functies der taal vanaf de
prae-humanistische philologie van Orleans tot de rationalistische linguistiek van Bopp, H.Veenman & Zonen, Wageningen.
Worth, D.S. (1983) The Origins of Russian Grammar: Notes on the State of Russian Philology before the Advent of Printed Grammars,
Slavica, Columbus, Ohio.
VI.
Non-European Traditions
Armenia

Adontz, N. (1970) Denys de Thrace et les commentateurs arméniens, transl. from the Russian (1915) by R.Hotterbeex, Imprimerie
Orientaliste, Louvain.
Mesopotamia
Black, J. (1984) Sumerian Grammar in Babylonian Theory, Studia Pohl: Series Maior 12, Biblical Institute, Rome.
Persia
Windfuhr, G.L. (1979) Persian Grammar: History and State of its Study, Trends in Linguistics: State-of-the-Art Reports 12, Mouton, The
Hague.
Judaism
454 LANGUAGE AND ITS STUDENTS
Bacher, W. (1975) Die Angänge der hebräischen Grammatik und Die hebräische Sprachwissenschaft vom 10. bis zum 16. Jahrhundert,
Studies in the History of Linguistics 4, Benjamins, Amsterdam (first published 1895 and 1892).
Hirschfeld, H. (1926) Literary History of Hebrew Grammarians and Lexicographers Accompanied by Unpublished Texts Oxford University
Press, Oxford.
Swiggers, P. (1979) ‘L’histoire de la grammaire hébraïque jusqu’au XVIe siècle, Orientalia Lovaniensia Periodica 10:183–93.
Islam
Diem, W. (1983) ‘Sekundärliteratur zur einheimischen arabischen Grammatikschreibung’, in Veersteegh, C.H.M., Koerner, K. and
Niederehe, H J. (eds), The History of Linguistics in the Near East, Studies in the History of Linguistics 28, Benjamins, Amsterdam
and Philadelphia: 195–250.
Owens, J. (1988) The Foundations of Grammar: An Introduction to Medieval Arabic Grammatical Theory, Studies in the History of the
Language Sciences 45, Benjamins, Amsterdam.
Semaan, Kh.I. (1968) Linguistics in the Middle Ages: Phonetic Studies in Early Islam, E.J. Brill, Leiden.
Sezgin, F. (1984) Geschichte des arabischen Schrifttums 9. Grammatik bis c. 430 H., E.J.Brill, Leiden.
Versteegh, C.H.M. (1977) Greek Elements in Arabic Linguistic Thinking, Studies in Semitic Languages and Linguistics 7, E.J.Brill, Leiden.
India
Allen, W.S. (1953) Phonetics in Ancient India, London Oriental Series 1, Oxford University Press, London.
Cardona, G. (1976) Pāṇini: A Survey of Research, Mouton, The Hague.
Nitti-Dolci, L. (1972) The Prākṛita Grammarians, translated from the French (1938) by P. Jhā, Motilal Banarsidass, Delhi.
Raja, K.Kunjunni (1963) Indian Theories of Meaning, The Adyar Library Series 91, Madras.
Scharfe, H. (1977) A History
o

f Indian Literature 5.2: Grammatical Literature, Otto Harrassowitz, Wiesbaden.
Staal, J.F. (1972) A Reader on the Sanskrit Grammarians, The MIT Press, Cambridge, Mass, and London.
The Far East
Halliday, M.A.K. (1981) ‘The origin and early development of Chinese phonological theory’, in Asher and Henderson (1981):123–40.
Miller, R.A. (1975) ‘The Far East’, in Sebeok, T.A. (1975):1213–64.
Miller, R.A. (1976) Studies in the Grammatical Tradition in Tibet, Studies in the History of Linguistics 6, Benjamins, Amsterdam.
AN ENCYCLOPAEDIA OF LANGUAGE 455
23
LANGUAGE ENGINEERING: SPECIAL LANGUAGES
DONALD C.LAYCOCK and PETER MÜHLHÄUSLER
1.
INTRODUCTION
Language has been compared variously with either bodily organs such as legs or hands, or else man-made tools like axes and
hammers. Related to the view that language is a mental organ is that it is optimally suited to the requirements of human
communication and that it is therefore best left alone: given no human interference it will adapt to the communicative purpose
it is put to. If language is viewed as a tool, however, there is ample room for the possibility of a misfit between language and
the world or its communicative tasks. This constitutes a justification for direct human interference in order to make it a more
adequate tool.
As the language metaphors of linguistics change, so do their views on the legitimacy and indeed feasibility of language
engineering. What was thought to be inadvisable and impossible within American structuralism, for instance, seems a
legitimate concern to those working within the developmental paradigm pioneered by Bailey in the late 1970s.
Our concern here is twofold. First we wish to discuss the theoretical limits within which all human interference with
language must take place. Second, we would like to analyse a number of different types of such interference and explore their
historical roots. In doing this we have opted to combine recent insights into linguistic naturalness with more traditional
terminology.
The parameters of naturalness and artificiality have not always been properly separated in past discussions. In particular,
one can observe a tendency to label as ‘natural’ any solution that draws lexical or grammatical material from an existing linguistic
system. It is for this reason that we shall avail ourselves of the widely used distinction between a priori and a posteriori languages.
A priori languages are constructed from scratch, their roots being selected on the basis of theoretical philosophical
considerations rather than adapted from real human languages. A posteriori languages, on the other hand, typically are

simplifications or adaptations of one or more existing languages. A number of artificial languages, e.g. Schleyer’s Volapük or
the proposals made by Comenius, are of a mixed type.
The attraction of a priori systems is perceived to be that they offer the possibility of encoding language-independent
universal logic. In actual fact, in the absence of such a logic, they tend to be very closely based, in their semantic properties,
on the European languages spoken by their inventors and hence turn out to be a posteriori languages in disguise. It is not
clear to us whether totally language-independent a priori languages can be created in principle. The attraction of a posteriori
languages is often seen to be their ‘naturalness’. In practice, the vast majority of special languages have been derived from a
very small number of Standard Average European (SAE) languages (principally Romance) and hence reflect the culture-
specific semantics of only a small part of the world’s population. In this connection, proposals such as Pham Xuan Thai’s
Frater, an a posteriori language based on Latin, Greek, Chinese, Japanese and ‘other non-Aryan speech communities’ are an
interesting exception.
2.
THEORETICAL CONSIDERATIONS: FORMAL PROPERTIES
Whilst it is possible in principle to add to, subtract from and permutate an existing linguistic structure in an almost infinite
number of ways and/or to invent entirely new languages from scratch, in practice many constraints operate if the outcome is
to be a communication system which human beings can handle. Further constraints arise if the requirement is to create a
language that can be acquired by children and yet others if the requirement is that it should lend itself to acquisition by adults
without formal instruction. Quite different limitations obtain when communication between species or with the inhabitants of
imagined or real other worlds are at issue. Yet other constraints need to be considered when languages are to be used by and
with machines.
Let us explore first the notion of ‘possible human language’. It has been characterised either in terms of a set of abstract
principles, as in Chomsky’s more recent writings, or else in terms of more superficial implicational patterns which lay down
the limited possibilities for the combination of surface properties of languages such as word order, the encoding of causatives
or the formation of relative clauses. Typologists of the school represented by Greenberg and Comrie note that only a small
subset of all mathematically conceivable combinations of features are found in actual languages. Whilst their ultimate aim of
discovering those typological features on which the maximum number of other features depends has been achieved only in
part, such studies as are available nevertheless tell us a great deal about the highly restricted options of human languages in
encoding certain concepts.
Derivations from typological studies include the study of language change, language acquisition, pidginisation and
creolisation as well as the notion of ‘linguistic naturalness’, a gradient concept ranging from the most natural to the most

abnatural. The term ‘unnatural’ is reserved for phenomena that cannot be produced by the human speech organs or handled by
the human intellectual apparatus. Examples of unnatural phenomena include sound which can only be produced if the lips are
enlarged by wooden plates and the effects of knocking out one or more of the front teeth. However, as natural human
languages can be expected to function in conjunction with natural human activities, claims that languages are excessively
affected by, for example, the chewing of the areca nut (as has been said of the Santa Cruz language in the Solomon Islands), or
by excessive teeth-chattering from the cold (Eskimo), or by extreme heat (African and Pacific Island languages), are nonsense.
Languages devised for extraterrestrial communication and computer languages are also unnatural in the technical sense.
‘Naturalness’ is considered to be a property of human language in general, though the degree to which languages approach
the most natural end of the continuum differs. According to Bickerton (1981, 1984), for instance, creoles are maximally
natural, whilst old cultural languages such as French or English are located at the other end of the spectrum. Whatever the degree
of naturalness in an individual language, the following observations hold for all languages:
(i) The presence of a less natural category implies that of a more natural one. Trial, for instance, implies dual which in turn
implies plural (at least with nouns). The reverse is not the case,
(ii) More natural categories are acquired before less natural ones: word final [p] is acquired before [b] and [t] is acquired
before [6] in all positions,
(iii) The maintenance of less natural categories requires more nurture and formal learning than that of more natural ones.
The characterisation of linguistic naturalness is rendered difficult by the presence of two basic requirements to language: that
it be produced and that it be perceived. Unless one is dealing with an extremely limited language, e.g. one consisting of only
the forms [mama], [pupu] and [dodo], it is inevitable that there is a conflict between optimalisation of production
(phonological naturalness) and optimalisation of perception (morphological naturalness). Generally speaking, what is easier to
produce is more difficult to distinguish and vice versa. The requirement of one form=one meaning can cause a high degree of
abnaturalness in production. For example, if the English concept ‘more than one’ was expressed invariably by [s], the
resulting plurals of rat, dog and rose would be a rather natural [ræts] next to almost unpronounceable [dogs] and [rœuzs]. The
requirement of one form=one meaning has other implications for perception; languages that rank high on the morphological
naturalness scale will favour synthesis or polysynthesis such that words consist of many smaller parts, each of them with their
fixed meaning. Polysynthesis, as is well known to typologists, virtually excludes fusion and similar natural phonological
processes. Language engineers have seldom appealed to such principles in a conscious fashion. Instead, in most instances they
seem to have opted unconsciously for strategies optimalising morphological naturalness (a posteriori languages such as
Esperanto being a good example). This is defensible if languages are to be used among second-language adult speakers. Children,
on the other hand, tend to require a considerable amount of phonological naturalness as well and will change languages in the

direction of greater pronounceability.
Whereas the understanding of linguistic naturalness among language engineers has been very limited, the notions of
simplicity and simplification, the former referring to the end product, the latter to the process leading to it, have played a
much more important role. At the most general level simplicity relates to the requirement of expressing the maximum number
of distinctions by means of the most economical linguistic apparatus. In practice, any act of engineering which increases the
coverage of grammatical rules, removes contextual conditions for their application or eliminates competing rules that do the
same job, contributes to the overall simplicity of the resulting system. Above all other things it means that lexical
irregularities are replaced, as much as possible, by grammatical regularities.
Technical simplicity is a great virtue for machine languages and in areas of human language, such as scientific
terminologies. Human processing skills, limitations on time and other human ‘performance’ factors impose quite severe
limitations on simplicity in languages for human use. Thus, whilst a computer can cope with a convention which expresses
the number 17 as two+ two+two+two+two+two+two+two+one, human beings cannot. Instead, more number stems are
needed. It appears that human languages have only very limited leeway for shifting the boundary between lexicon and
grammar. However, this principle is often found to be violated, particularly in apriori languages.
AN ENCYCLOPAEDIA OF LANGUAGE 457
In a language imagined by Tyssot de Patot (1710), for example, the numerals are expressed by the successive consonants of
the language (b is ‘1’, d is ‘2’, p is ‘10’, and ‘21’ is expressed dpb). Even fewrer numerical roots are used by John Weilgart, in
aUI, his ‘language of space’ (Weilgart 1968); the first five numerals are expressed by five nasalised vowels, and the next five
by the same vowels lengthened (represented in his system by a, e, i, o, u, A, E, I, O,U).
Next to violating constraints on the boundary between lexis and grammar in human languages, the promotion of maximum
simplicity also clashes with considerations of naturalness within grammar and grammatical components. Thus, from the point
of view of simplicity, a system which adds a plural ending every time more than one of an entity is mentioned would seem
preferable to one where plural endings occur with some lexical stems only. Applying this principle English should have plural
forms such as traffics, funs, wheats and pronouns such as Is, yous and hes, shes and its.
From the point of view of naturalness, the most natural environment for plural marking are nouns referring to human
beings, followed by animates, countable objects, mass nouns and finally abstract terms. The absence of a form such as funs is
thus natural, but an irregularity in the technical sense.
Other examples are the Esperanto forms patro (stem pair+noun marker -o) and patrino (stem patr ‘father’+in
‘feminine’+noun marker o). Here we are dealing with a violation of the principle that what is prominent in the perception or
experience of speakers of a language should be encoded by simple lexical items. Since the mother tends to be more central to

the experience of children than their father, the direction of the derivation should be the reverse of what it is in Esperanto;
considerations such as the requirement that languages should be free from sexism, racism and so forth also militate against
this Esperanto form.
The principle just mentioned is known as Zipf’s Law and it raises another important issue, namely whether languages are in
essence iconic or arbitrary. This is not the place to go into detail about the history of this debate. Instead, some areas where
iconicity appears to have been regarded as important by language engineers will be briefly illustrated.
Lexical stems can be motivated in two senses, either by being a direct icon of the entities they refer to (onomatopoeia) or
else being made up of sound segments that are iconic of aspects of the real world, e.g. [i] for smallness or proximity, [t] for
abruptness and so forth. Again, this is the system of Weilgart’s aUI, where, for instance, the basic vowels a, e, i, o, u, represent
‘space, movement, light, life, human’; but such a system is proposed as early as 1636 by M.Mersenne, in his Harmonie
universelle, where the vowels represent the following qualities:
a and o=what is great and noble
e=delicate, subtle things, suitable for the representation of sadness and mourning
i=very thin, small things
o=the expression of great passion
u=dark, hidden things
(Knowlson 1975:67–8, which see for further examples).
In George Dalgarno’s philosophical language, presented in his Ars signorum (1661), words tend to have a basic root shape,
with the differentiations being expressed by vowels, or by inserted consonants (r is ‘opposite’, l ‘the mean between two
extremes’). Such minimal distinctions are of course totally at odds with the important principle of redundancy in natural
language, since the slightest interference in the channel (noise of any kind) will obliterate essential distinctions. Pronouns and
demonstratives in Dalgarno’s language, for instance, are (using some Greek letters):
lal ‘I’, lŋl ‘thou’, lei ‘he’, lol ‘this’,
lvl‘itself’, lul ‘who’.
Mishearing the vowel could lead to very serious misunderstanding. Constructing languages entirely out of iconic segments or
lexical stems often means having to give up considerations of the next step of iconicity, morphological or constructional
iconicity (cf. Mayerthaler 1981). With this type the length and the morphological complexity of a lexical item are related to
both the already mentioned cultural prominence and naturalness. Thus, a word referring to great great great grandmother is
iconically encoded if it is longer than that for great great grandmother, great grandmother, grandmother and mother, and
counter-iconically if it is not. Similarly, plural forms of nouns should be longer or morphologically more complex than

singular ones and past tense forms of stative verbs longer than non-past forms. Thus, Esperanto, where equally long and
morphologically complex forms are required for all tenses, is not ideally encoded, a problem which does not arise with
Volapük where non-present forms are morphologically more complex than present ones.
Little is known about iconicity in word order other than the existence of certain general principles, such as the fact that
prominence of a topic should be reflected by its appearance in a syntactically prominent position or sequence of events should
be mirrored by sequence of clauses. No examples of conscious attempts to introduce syntactic iconicity are known to the
authors.
Attempts to rationalise the pronunciation of artificial languages have been either absent or rudimentary. Thus, highly
marked sounds such as [y] and [ø] are found in Volapük, and the distinction between [l] and [r] is found in most of the
458 LANGUAGE ENGINEERING
twentieth-century languages such as Esperanto, Ido, etc. including Frater which is allegedly based on Chinese and Japanese as
well as European languages, and complex consonant clusters are encountered in many of them. Much more attention is
usually paid to streamlining the writing systems of these languages e.g. by eliminating digraphs or special symbols. This, of
course, is in line with the general observation that most artificial languages are derived from a written form of language and
designed to facilitate written in preference to oral intercommunication. Whereas some a priori languages lend themselves to
the former only, artificial languages designed exclusively for oral communication are very rare, a possible example being
Schwörer’s Kolonialdeutsch (1916), designed for the use of the coloured workforce of a German world empire.
Whereas phonetic or phonological naturalness in the sense of optimalising production is not much in evidence, concern for
sound symbolism or onomatopoeic iconicity is encountered frequently, particularly in a priori languages. Thus, the idea that
the vowel (i) should stand for smallness, (a) for largeness and (0) for roundness or universality is encountered in a number of
proposals, though in most instances such iconicity tends to be postulated in an impressionistic manner and be mixed with
culture-specific conventions.
The paucity of information in the areas of syntactic and phonetic naturalness relates to the more general observation that
not all levels of language have been equally susceptible to engineering. Deliberate interference with human languages seems
to be guided by the following hierarchy:
degree of interference
most
least
lexicon
derivational morphology

inflexional morphology
phonology
phonetics
syntax
Language creation often exploits the principle that there are trade relationships between the levels. Thus, the size of the
irregular lexicon is reduced by promoting a more powerful derivational morphology in many a posteriori proposals.
At each level a number of operations can be carried out, mainly addition, subtraction, substitution and rearrangement or a
combination of these (polysystemic). Such processes are common in the mechanical alterations of natural languages called
‘ludlings’ (Laycock 1972), or play-languages of the Pig Latin type. Some examples of each type follow:
Addition: you can talk Gree -yougree cangree talkgree Greegree
you can talk skimono jive -skyou skcan sktalk skskimono skjive
Subtraction: fabulous -fab
snake+shark -snark
sado-masochism -sadie-maisie
Substitution: Ei, da sitzt ‘ne Flieg’ an der Wand -i, di sitzt ‘ni Flig’ in dir Wind
Rearrangement: look at the old woman butterfly -cool ta the dillow namow
-flutterby
Polysystemic: Pig Latin -Igpay Atinlay
-withus youvus govus
Arguably the most important and least heeded property of ‘natural’ human languages is their versatility. Far from being closed
rule-governed systems they are open-ended and only partially governed by rules. The human language ability is rule-changing
rather than (as Chomsky and his followers have it) rule-governed creativity. Thus, an important measure for the success of
artificial languages should be their adaptive creativity. It can be noted that most invented systems, in particular a priori
systems, are ill-suited to change. The history of languages such as Esperanto demonstrates that a posteriori languages also
suffer from this deficiency. Those who are in the business of language elevation (e.g. in the case of Afrikaans, Bahasa
Indonesia etc.) have begun to cater for change, though this remains one of the least researched areas.
The uneven coverage of different levels of language is encountered again with rank or size-level. Engineering typically
concentrates on the rank of word, maximally the sentence. Very little regard is paid to paragraph or discourse organisation in
spite of the fact that languages can differ greatly in their discourse pragmatics; sharing words and grammatical structures is
insufficient in bringing about an understanding between different groups.

AN ENCYCLOPAEDIA OF LANGUAGE 459
3.
THEORETICAL CONSIDERATIONS: FUNCTIONAL PROPERTIES
The study of language functions is still at a pre-theoretical stage and no conclusive findings exist about either the number of
functions or the boundaries between them. For the purposes of this section we shall rely on the work of scholars such as
Halliday and Jakobson. In Halliday’s model (1974), the order in which functions emerge developmentally in child language
acquisition is the following:
directive
phatic
expressive
heuristic
metalinguistic
poetic
cognitive
The general principle that awareness about linguistic matters decreases as developmentally earlier phenomena are involved,
when applied to the above hierarchy, means that one can expect more deliberate interference with linguistic structures at the
cognitive and poetic functional levels.
Any survey of language engineering, both with a posteriori languages derived from existing human languages or a priori
languages designed on the basis of philosophical or theological considerations, will demonstrate the primacy of the cognitive
function in man-made systems.
It is interesting to note that in many early examples of language engineering other, non-linguistic, functions were often
dominant. Thus, in medieval and postmedieval Europe the creation of a new language was seen as imitatio dei, i.e. promoting
the creation of a better world. Figures such as Jacob Böhme (1586–1654) insisted that the creation of a perfect language was
an urgent necessity in preparing mankind for the imminent second coming of Christ, a view which was also held by many of
his contemporaries and successors. Comenius (1592–1670) sketches a three-stage programme for the transition from a
pantaglossic via a polyglossic to a monoglossic mankind, the endpoint being reached when all mankind could praise God in a
single language.
Similarly abnatural functions of language which emerge again and again are attempts to create codes which could provide
access to the secrets of the universe and systems for concealment of information, either for the use of small privileged groups
or, in the case of some forms of glossolalia, for individuals only. Such secret languages also emerge ‘naturally’ in children

(e.g. in twin languages such as Spaka—see Diehl et al. 1981).
The most important distinction in the area of functions is that between artificial languages designed as replacements of one,
many or all human languages and others which are to serve as supplementary interlanguages or languages of special domains.
The desirability of a single language was motivated, in many earlier writings, by the argument that the original (pre-Babel)
state was one of monolingualism. Later writers, particularly those of the nineteenth and twentieth century emphasise the
importance of a single world language for world peace. Thus, Schleyer, the inventor of Volapük, adopted as his motto:
‘Menade bal, püki bal’ ‘to one human race, one language’; others, such as Baumann, the inventor of Weltdeutsch (1916), saw
the use of a single German-derived language of intercommunication as a means of perpetuating German superiority. Whereas
single languages, by definition, would have to fulfil all functions of human languages (plus any additional ones which their
inventors may wish to promote), auxiliary or supplementary artificial languages can be restricted to a small range of functions
and domains.
For the majority of those devising a priori languages the main function of language was seen as reflecting the orderliness
of creation. It was generally felt that the proper representation of things is taxonomy and that words should be isomorphic
with the things they refer to. By bringing about such a fit between language and the world, it was felt that a proper foundation
for human cognition was laid.
The expectations of most authors of a posteriori languages tended to be less ambitious, though again related to the function
of cognition: auxiliary languages were seen as a means of exchanging ideas between members of different communities.
Typically, these ideas are those of educated groups (Schuchardt 1928:371 refers to the exchange of scientific ideas as the
most important and most immediate task) which, at the time when most artificial languages were devised, meant the inhabitants
of Europe and North America. The demonstration that an artificial language also lent itself to the poetic function was
frequently felt to be an important additional virtue. Thus, there has been a stream of poetical work in Esperanto ever since the
first volume of poetry appeared in 1893. Distinct from this poetic function is the creation of languages for stylistic or other
literary purposes, in order to illustrate how the inhabitants of various utopias and dystopias speak. A survey of these
languages (Laycock 1987) draws attention to some salient aspects of such literary creations:
(a) their typological diversity is greater than that of natural human languages, but in quite different ways
460 LANGUAGE ENGINEERING
(b) unnatural types (e.g. speaking backwards or a priori systems) are common among the inhabitants of literary worlds
(c) most a posteriori languages are modelled very closely on a small number of European languages, especially in features
such as word order (SVO) and nominative-accusative grammar
(d) iconicity is high and regularity paramount

(e) the languages are frequently controlled and regulated by some governing body such as an academy
The desire to bring about a fit between language and the world underlies many literary creations. Thus, to mention some
recent examples, the use of Russian and quasi-Russian forms in Burgess’s A Clockwork Orange has been interpreted as an
index of a violent society and the use of numerous West Germanic and Scandinavian roots in Nabokov’s Zemblan is
appropriate to a country like Zembla, a distant northern land (see Krueger 1967).
The poetic function of artificial languages is manifested not only in ‘high’ literature but, to an even greater extent and
across a wide range of cultures, in verbal play and play languages such as ludlings (those involving regular transformations).
Play-languages form part of the linguistic competence of a whole community, but, since their use requires practice, they often
serve as virtually ‘secret’ languages for a particular subgroup in a community, since the other members of the community lack
the fluency that comes from constant use. Such a subgroup is frequently composed of adolescents and pre-adolescents, at a
stage after children have mastered the basic grammatical norms of the language, and now feel free to experiment, in an
environment where their concerns (sexual and individualistic) may already be at odds with those of the adult world.
In view of the fact that language is talked about in terms derived from everyday languages one could expect a concern, on
the part of the inventors of artificial languages, for culture-neutral (universal logic) metalanguage. In actual fact, there is little
evidence of this among planned and artificial languages other than as part of a wider effort to organise human discourse into
taxonomies. Typically, the metalinguistic categories established within the classical tradition are taken over in the
construction of languages. Thus, there is an almost universal subdivision into the traditional parts of speech, inflections with
verbs (tense rather than aspect) and number indication with plurals. Constructions for which there are no labels at the time of
invention, (e.g. ergativity, switch-reference, anti-passives) are not considered by the vast majority of language makers.
Linguistic terminology continues to remain unsatisfactory and better attention to metalinguistics, particularly on a posteriori
sorts of metalinguistics rather than on a prioristic systems, remains an urgent task.
Attempts to construct languages appropriate to the expressive function typically relate to religious experience and tend to
be restricted to individual needs (as in glossolalia) rather than communication between individuals. In the latter case, as
pointed out by Samarin (1976:6 ff), language creation is geared towards promoting the spiritual cohesion of religious groups
as well as the exclusion of outsiders.
Perhaps inevitably, however, such phatic communication is rarely in the form of a fully worked-out language, but is
expressed in non-linguistic modes such as glossolalia, which is merely a form of language-babbling, or glossomimia, which is
the imitation of natural languages in such a way to give the impression that the actual language is being spoken. Glossolalia is
common in many of the world’s religions—not only charismatic Christianity—and serves to indicate the acceptance by the
speaker of a divine spirit. Glossomimia, on the other hand, is more common in dramatic representations, where, for instance,

an actor may wish to give the impression of being a German guard in a prison camp; he will speak a nonsensical ‘language’ with
the phonetic qualities of German, and the nature of the role will thereby be established.
The term phatic communication was introduced in the 1920s, i.e. after the majority of a priori and a posteriori languages
under consideration in this paper had already been constructed. Thus, in spite of the importance of phatic communication (i.e.
keeping the communication channel open) for harmony in interpersonal relations, little formal planning of language for
conflict-avoiding small talk etc. has been produced. Rather, the general assumption has been that the maximum amount of
information should be exchanged through planned languages.
Whilst we have so far observed a neglect of the lower non-cognitive functions in artificial languages, there has been
considerable activity in connection with the last function to be considered here, the directive function. The idea that it was
possible to construct a language for mind control is most closely associated with Orwell’s Newspeak (cf. Bolton 1984), where
it is argued that, by removing words from the dictionary, the associated concepts and the ability to argue about them is also
removed. At the same time, existing words are given new more specific meanings and new terms are coined to suit
ideological requirements. Social control through language use (rather than language creation), of course, has a very long
tradition and languages constructed with social control as a principal motive also predate literary newspeak. Thus, Schwörer’s
Kolonialdeutsch of 1916 was designed for purposes such as:
‘to increase the regional mobility of native workers, thus increasing their reliability’;
‘a language for German masters and colonisers to give orders in’;
‘a symbol of German authority’.
AN ENCYCLOPAEDIA OF LANGUAGE 461
Like Newspeak, Kolonialdeutsch was vastly reduced in its lexical resources, in order to prevent the ‘natives’ from
overhearing their masters’ conversations and from ‘debating controversial topics among themselves’. Since the German plans
for world domination after the First World War did not materialise, this artificial form of German was never implemented,
nor were subsequent attempts to revive a similar form of German for use with guest workers. Most current attempts to make
language more suited to social control concentrate on small areas of the lexicon rather than large-scale structural revision.
Considerable advances have been made in both the language of advertising (e.g. Vestergaard and Schrøder 1985) and politics
(Bolinger 1980).
In sum, most language engineering has been carried out with relatively little regard to the functions of language, and a good
deal of it with deliberate disregard of functions other than the cognitive one. Whilst cognition is an important aspect of
language, its role tends to be overrated by the members of academic institutions. Such progress that has been made in the
areas of logic and grammar still needs to be matched by advances in the rhetorical dimensions of language.

4.
A PRIORI LANGUAGES
A priori languages exist for the most part as ideational constructs, as, in the sense of being languages that display ‘perfection’
in all the possible ways— iconicity and total naturalness—they cannot exist as actual languages, or even be constructed to any
real degree. Nevertheless, the concept of the ideal a priori language has been a common theme throughout all recorded
civilisations, and we deal now with some of the principal preoccupations.
The ideal language has been envisaged as the divine language, the primal language, and the philosophical language. For
some writers these three are all the same; others may wish to distinguish them. The divine language is that spoken by
benevolent superhuman forces such as God or the gods, and angels and other divine messengers; the features are best
summarised by Swedenborg (1758, 1958):
‘In the entire heaven all have one language…Language there is not learned but is implanted by nature with every one,
for it flows from their very affection and thought. The tones of their speech correspond to their affection, and the vocal
articulations which are words correspond to the ideas of thought that spring from the affection; and because of this
correspondence the speech itself is spiritual, for it is affection sounding and thought speaking…Because the speech of
angels proceeds directly from their affection…angels can therefore express in a moment what a man cannot express in
half an hour.’
The language, therefore, is immutable, instinctive, iconic, ultra-brief, logical and natural. These qualities are also attributed to
the primal language, or supposed original language of mankind, which in most religious systems is considered to be of divine
origin.
There are, of course, considerable problems with such a concept, from a linguistic point of view. In such a language it
would be impossible to lie, since the complete match between utterance and reality would mean that only true statements
could be uttered; moreover, nothing new could be uttered, and all possible statements would already be known to hearers, so
that the only choice is between platitudes and silence, or else, as in the imagination of Borges (1970), the uttering of a single
word which will in fact substitute for all possible utterances:
‘In time, the notion of a divine sentence seemed puerile or blasphemous. A god, I reflected, ought only to utter a single
word and in that word absolute fullness. No word uttered by him can be inferior to the universe or less than the sum total
of time. Shadows or simulacra of that single word equivalent to a language and to all a language can embrace are the
poor and ambitious human words, all, world, universe.’
The philosophical language is not usually conceptualised in such an extreme form; it need not be immutable (although change
is not easily incorporable into most of the proposed systems), it is capable of being learned (and is therefore not instinctive),

and it is not ultra-brief or comprehensive (in the sense of a single utterance being able to stand for much longer utterances in
natural languages). Also, as pointed out earlier in this paper, the iconicity and logic of the philosophical language operate in
the opposite direction to naturalness.
From about the sixteenth century on (with a few earlier precursors like Ramon Lull) the idea begins to take shape among
European scholars that it might be possible to create such a universal philosophical language that would be as non-arbitrary as
possible, based on a philosophical categorisation of human experience, and completely regular in morphology and derivation.
It is perhaps not coincidental that this idea grows in force with the gradual decline of Latin as the lingua franca of scholarly
Europe.
462 LANGUAGE ENGINEERING
Credit for one of the first serious suggestions for such an a priori language is often given to Descartes, in a letter to Père
Marin Mersenne (1629). But Descartes admits only grudgingly the possible utility of a multilingual dictionary with an
interlingual key, and has few concrete suggestions to make on the construction of a language based on philosophical
categories, other than to suggest that the philosophy would have to be correct, and that the morphology should be regular:
‘. . car faisant une langue, où il n’y ait qu’une façon de conjuger, de décliner, et de construire les mots, qu’il n’y en ait
point de défectifs ni d’irréguliers, qui sont toutes choses venues de la corruption de l’usage, et même que l’inflexion des
noms ou des verbes et la construction se fassent par affixes, ou devant ou après les mots primitifs lesquelles affixes
soient toutes spécifiées dans le dictionnaire, ce ne sera pas merveille que les esprits vulgaires apprennent en moins de
six heures a composer en cette langue avec l’aide du dictionnaire, qui est le sujet de la première proposition.’
Leibniz also toyed with the idea in the third quarter of the seventeenth century, but, as Knowlson (1975) points out, his ideas
were only theoretical in nature, and many of his notes on the subject were not accessible to his contemporaries, and have
become known only through the work of Couturat (1901, 1903). Credit for one of the most thorough-going attempts to
construct such a language must therefore be given to George Dalgarno, who published his Ars signorum in 1661.
The complexity to which such philosophical languages can aspire is seen also in a language proposed by Thomas Urquhart
in 1953. Urquhart did not create the words of his language, but set down its rules: two hundred and fifty roots, ‘eleven
genders, ten tenses (occurring with all parts of speech), six moods, four voices, twelve parts of speech—and, more
importantly, every letter to have a meaning, and every word generatable in the language will actually occur’, so that
92. Two and twentiethly, In this Language the opposite members of a division have usually the same letters in the words
which signifie them; the initial and final letter being all one, with a transmutation only in the middle ones.
93. Three and twentiethly, Every word in this Language signifieth as well backward as forward; and however you invert the
letters, still shall you fall upon significant words: whereby a wonderful facility is obtained in making of Anagrams.

The whole concept of the a priori philosophical language is satirised by Jonathan Swift, in his account (1726) of the
philosophers of Laputa, who carried around with them all objects that could conceivably be the subjects of discourse, ‘since
words are only names for things’. The strength of the lampoon lies in the fact that in a short while it would occur to such
philosophers that they could carry around pictures of the objects talked about, and that it is only a small step from such iconic
representation to the iconic representation of objects in words, as envisaged in the philosophical language.
Perhaps because of such satirical comment, it is rare to meet with fully elaborated a priori languages later than the eighteenth
century. Their direct descendants, however, are the formal language systems described in textbooks of mathematics and
formal logic—including the interesting Lincos, a language designed for ‘cosmic intercourse’ (Freudenthal 1960 and 1974)—
and also some computer languages. Mention has also been made of Weilgarth’s ‘language of space’ (1968), which is claimed
to be beneficial in imparting concepts to mentally-disturbed children. But such languages also tend to derive a great deal of
their structure from a posteriori principles, and are therefore not philosophical languages in the strict sense.
Because of the iconic nature of the philosophical language, some attempts have been made to present the language in pictorial
form; such ‘sign languages’ are called pasigraphies. Weilgarth’s aUI is one such, since it provides pictorial elements for each
of the units of the language, and two other twentieth-century pasigraphies are Leslie Charteris’s Paleneo (1972) and
‘Blissymbolics’ (Bliss 1965). Such systems have their uses in the signs required for communication across many languages
at, for example, international airports, but tend, in their detailed elaboration, to be far from transparent, and thus to offer little
advantage, if any, over invented languages presented in normal script.
The idea of a pictorial iconic language in general can be traced back to the dawning knowledge, in Europe, of the Chinese
language (Frodsham 1964). Chinese was popularly supposed to have a writing system in which whole words, or ‘concepts’, were
expressed by single characters, some of which were analysable as pictorial representations—and were therefore non-arbitrary,
along the lines of the philosophical language. In addition, the presence of tones in Chinese, and the concept of music as an
‘international language’ (a concept which ignores the extremely cultural bases of all music, and also the very limited extent of
possible communication) gave rise, from the seventeenth century on, to a large number of ‘Utopian’ languages based on
musical themes. By ‘Utopian’ languages we mean languages designed to serve Utopian states, and which, like the states
themselves, are regulated so as to be, in some way, an ‘improvement’ on natural languages—whether they are conceived of as
merely modified natural languages (a posteriori languages), or as philosophical a priori languages, or a combination of both.
Although such fictional Utopian languages cannot be considered as serious linguistic engineering, one continuing theme, that
of the expression of concepts contradictory in meaning, needs to be singled out, since it resurfaces in the construction of a
posteriori languages. Since there is a direct logical relation between a word, or concept, and its contradictory, it follows that
in an ideal language the contradiction should always be expressed in the same way. In Solresol, an a priori language based on

the musical scale, and playable on any instrument, proposed in 1818 by Jean-Frédéric Sudre, contrary ideas are expressed by
AN ENCYCLOPAEDIA OF LANGUAGE 463
reversing the notes—thus domisol is ‘God’ and solmido is ‘Satan’—but as Couturat and Léau (1903) point out, the system
breaks down with dosidomi ‘vegetable’ and midosido ‘sacrifice’. A corollary of this regularised negation, so much a feature
of the a posteriori languages, are the ideas that the same word in a Utopian language can express widely different meanings,
even opposites, or that the ambiguity of the spoken language is so high that speech must be accompanied by gesture; such
ideas also have their ultimate origin in misconceptions about the nature of Chinese—even the notion of gestures may derive
from observation of Chinese signalling written characters in the air. But the Utopian languages of fiction, including science-
fiction, are many, and only some of the prevailing themes have been indicated here. (For a comprehensive treatment, see
Laycock 1987.)
5.
A POSTERIORI LANGUAGES
A posteriori languages are usually defined as deliberately simplified versions of existing human languages or groups of
languages. Contrasting with this narrow definition is the broader view which on the one hand studies instances of language
engineering leading to more esoteric and more complex languages, and on the other simplified languages (pidgins, koinés and
lingue franche) whose simplicity is not the result of deliberate human agency. It is somewhat ironic that perfectly smoothly
functioning a posteriori languages were used by millions of non-literate speakers in many parts of the world at a time when
the academics of central Europe were trying to discover the secret of the optimal language for international communication. It
is interesting to note that the majority of these scholars either did not take note of pidgins or else rejected them; as did
Jespersen in his Introduction to An International Language who argues that Simplicity does not mean that the language we
construct is to be a kind of “Pidgin” incapable of expressing nuances of thought which are necessary to highly cultivated
Europeans…The interlanguage that I am advocating…is totally different from such languages through being expressive and
efficient, though being extremely simple in its grammatical structure.’ One of the few attempts to base an a priori language on
the structures of the Mediterranean Lingua Franca and other pidgins is Steiner’s Pasilingua (1885). Schuchardt, who did
intensive research on both pidgins and creoles and international auxiliary languages, also emphasises the importance of the
former for the latter and argues that instead of deliberately simplified forms of Latin (such as Nov-Latin or Mondo lingue) one
could equally well (1928:382) ‘introduce a creolised version of a Romance language as an international language.’
The originator of the idea of an international auxiliary language is Descartes who, in a letter to Père Mersenne of 1629,
outlines proposals for both simplified existing a posteriori languages and a priori ones, a proposal which met with considerable
acceptance in intellectual circles, as the eclipse of Latin and the rise of national languages had made intellectual discourse

between scholars from different European nations increasingly difficult. During the seventeenth and first half of the
eighteenth century a number of a priori languages were constructed (see above) as well as mixed systems such as those of the
type proposed by Comenius and Leibniz, the latter (1646–1716) combining a Latin-derived lexicon with universal
philosophical semantax. The earliest purely a posteriori language is discussed in the Encyclopedic of Diderot and d’Alembert
in an article by M.Faiguet dated 1765. Whilst only a sketch is provided, a number of principles of simplification are clearly
stated which are appealed to over and over again in later examples of a posteriori languages. Thus, inflexional morphology is
vastly reduced, grammatical accidents such as gender no longer coded and articles disposed of. What little morphology
remained was entirely regular.
A complete proposal for an a posteriori language based on Latin is that by Carpophorophilus (1734): the Latin lexicon is
reduced in size through the elimination of synonyms, the declension is replaced by articles, there is a single plural affix -im
borrowed from Hebrew and the verbal inflection is regularised.
Numerous similar proposals for a posteriori languages follow in the nineteenth century. Most of them were the work of
individuals working in isolation and thus never got implemented. The majority of them were based on a few languages spoken
in Western Europe, particularly Latin and other Romance languages and, as far as we can see, all of them were conceived as
closed systems rather than languages capable of change and adaptation. The first artificial language actually spoken by a
significant number of human beings was Volapük, an invention of the German priest Schleyer (1880). As already indicated,
Volapük combined elements of both a posteriori and a priori languages. Thus, part of its lexicon was derived from other
languages (mainly English, German and Latin) by regular as well as ad hoc processes (e.g. vola from English ‘world’ and pük
from ‘speak’), whilst other items were invented, particularly the bound stems of words such as -el ‘inhabitant of’, as in
Parisel ‘Parisian’ or -af ‘animal’ as in suplaf ‘spider’.
As Volapük was designed eventually to replace all other languages, simplification was restricted to the regularisation of
morphological patterns but not the shedding of grammatical categories such as tense, gender or voice. Rather, it was a highly
synthetic language, demanding considerable encoding and decoding skills from its users. The following forms chosen from
the verbal paradigm of löfön ‘to love’ illustrate this:
löfob ‘I love’
älöfob ‘I loved’
464 LANGUAGE ENGINEERING
ulöfob ‘I shall have loved’
löfom ‘he loves’
löfof ‘she loves’

löfob-s ‘they (fem) love’
pe-i-löf-o-f ‘she was always loved’
löfo-b-seok ‘we love one another’
no-li-e-löfo-s-la ‘will you not have loved’
e-löfo-m-la ‘that he has loved’
and about 505,430 possible verb forms.
In spite of such difficulties, Volapük rapidly acquired an international following beyond the wildest dreams of its inventor.
An estimated million speakers used this language in 1888 and three international Volapük congresses took place in rapid
succession in 1884, 1887 and 1889, the last one with representatives from 13 countries. It is reported that the decision to hold
all proceedings in Volapük revealed its deficiencies to the delegates. More important, it appears, was the ideological conflict
between Schleyer who wanted to see his invention capable of expressing the full range of semantic distinctions found in other
human languages and those who wished to reduce Volapük to a more modest status of an international auxiliary language.
Some of his previous followers (such as the creators of Bopal, Spelin, Dil, Balta, Veltparl and La langue bleue) proposed
simplified adaptations, others deserted to the Esperanto movement, which was then gaining impetus.
Esperanto, invented by the Polish oculist Ludwig Zamenhof (1859–1917) is an a posteriori language based primarily on
Romance and Germanic (Teutonic) lexicon and grammar. A first version published in 1887 and a second version,
incorporating many outsiders’ suggestions as well as changes by Zamenhof himself, was put forward in 1894.
The spread of Esperanto was slower but more steady than that of Volapük. Its success is due to the more felicitous
grammatical structure of the language (considerably less synthetic than Volapük), its flexibility, and possibly political
persecution in both Czarist Russia and Nazi Germany which gave the language the status of a progressive liberating anti-
language. The equilibrium between an international lexicon and a small set of (about 50) agglutinating inflectional and
derivational formatives made it a very easy language to master, particularly for speakers of SAE languages. Some of its
structural properties, however, are so typically European (six pronoun system, the distinction between nominative and
accusative forms of the noun, the agreement of adjectives with the nouns they modify) or abnatural (the derivation of all
feminine forms from a masculine base by affixation or that of polar opposites by means of a prefix mal-) that learners from
other cultural backgrounds are at a great disadvantage. Another criticism is that a number of widely known international roots
(e.g. possible Latin-derived forms such as anno ‘year’ or skola ‘school’ are rejected in favour of less well-known roots such
as jar (taken from German) or lernejo (German root with Esperanto affix).
As in the case of other a posteriori languages related to SAE languages, the yardstick for what constitutes a concept and
consequently a lexical item is closely modelled on the inventor’s own knowledge of languages, in particular German. Thus,

there is only one verb translating ‘to be’, namely, sein but two verbs for ‘to look after’, flegi from German pflegen for ‘look
after the sick’ and varti from German warten for ‘to look after children’. Speakers of languages other than SAE ones would
find such culture-specific lexicalisation a considerable difficulty.
Selecting a single semantic domain, colour terminology, and comparing Esperanto with other artificial languages will
illustrate this point. The results of numerous studies on colour systems in human languages (summarised, for instance in
Berlin and Kay 1969) show (a) that human languages can differ widely in the number of colour distinctions they lexicalise, (b)
that only a small subset of all mathematically probable combinations of colour terms are permitted in human languages and (c)
that certain colour distinctions are considerably more basic than others. For constructing a universal auxiliary language this
means that one must work from the premise that the only colours one can assume to be readily intelligible to everybody are
black, white and possibly red. If one considers concrete proposals, however, one finds a very different picture:
Esperanto (according to Wells 1969) appears to have 12 primary colour terms, including mauve, purple, pink, grey and
brown, the latter reflecting Western Germanic predilection for naming dark reddish colours. Unsurprisingly, brown is also one
of the colours in Schleyer’s Volapük (1885) together with eleven other primary colours including purple, pink, grey and
orange. Brown, pink and purple are absent in Henderson’s Lingua (1888) where only six basic colours (black, white, red,
green, yellow and grey) are found.
The colour brown is absent in Frater (Thai 1957) in spite of the fact that it is a basic colour term in the mother tongue
(Vietnamese) of its inventor. The inclusion of another basic colour term in this language, klor ‘greenish-yellow’, may be an
instance of hypercorrection or overcompensation for the absence of a blue-green or yellow-gold distinction in Vietnamese.
From such considerations it follows that to call an international auxiliary language La Langue Bleue is not quite felicitous,
leaving alone the problem that the sky, which is supposed to have been the limit for this language, tends not to be blue in
some tropical parts of the world but has a glaring whitish tone.
However, lexical ethnocentrism was not one of the reasons for the developing rift in the Esperanto movement; rather, as in
the case of Volapük, the later fate of Esperanto is characterised by a split between those in favour of the original or a mildly
AN ENCYCLOPAEDIA OF LANGUAGE 465
revised language and those advocating radical revisions. The split was triggered by the foundation of the ‘Delegation on the
adoption of an International Auxiliary Language’ in 1900, consisting of a group of competent scholars including the linguist
Jespersen. After initial discussion of a larger number of languages (discussed in detail by Jespersen 1921) the choice was
narrowed down to Esperanto and Idiom Neutral, a language designed by a group that had left the Volapük movement in 1903.
At a later stage Ido, an artificial language derived from Esperanto, was submitted anonymously and received very favourably
by the Delegation. The final decision was to recommend Esperanto with a number of reforms along the lines of Ido, a decision

unacceptable to many Esperantists. Ido, in turn, received several improvements and changes and a language called
Esperantido was proposed by René de Saussure. None of these languages derived from Esperanto could arrest the further
spread of the original version which continues to remain the most widely spoken artificial language, and is seen by some as
the future language of the Common Market.
The First World War affected the artificial language movement in several ways. First, existing languages, in particular
Esperanto, were used in war propaganda by both German and French groups, and second, it promoted the development of a
posteriori languages derived from single national languages. The need for a language which could bring about international
co-operation was felt even more strongly after the war. By then, a number of principles and practices of artificial language-
making had crystallised, including:
(a) that aposteriori languages are preferable to apriori ones
(b) that naturalness (as yet ill-defined) should be promoted as much as possible
(c) that the language should be for the whole world rather than the educated classes of Western Europe
(d) that the solution should be the result of teamwork rather than an individual effort
(e) that the implementation of the solution should be in the hands of an international authority such as the League of Nations
The first two points are illustrated by an increasing tendency to favour solutions found in actual human languages over
invented solutions, even at the cost of reduced regularity. Point (c) is not widely applied, however, as the majority of linguists
concerned with the improvement or creation of artificial systems continue to be speakers of SAE languages with a very
limited experience of languages of other families. Conspicuous by its absence is an understanding of the sociolinguistic
mechanisms of cross-linguistic communication. This criticism has been made of virtually all the languages that succeeded
Esperanto and Ido, such as Peano’s Latino sine flexione (1908) Jespersen’s Novial (1928) and Ogden and Richard’s Basic
English (1931). Among the few exceptions are Hogben’s Interglossa of 1943 which combines Greek vocabulary with Chinese
syntax, and Thai’s Lingua Sistemfrater (Frater) of 1957. Of the above languages Basic English needs to be singled out as a
special type since in this reduced form of English, we find that the spelling, pronunciation and grammar of the English
language have been left intact, the principal change being a carefully limited lexical inventory of 850 items. This language
thus could fulfil the dual functions of international auxiliary language and transitional stage in the acquisition of English as a
second language.
The failure of most proposals for auxiliary languages is related to the limited insights and authority of individuals. Thus,
decisions of importance tended to be made in international bodies and one of the major present-day proposals, Interlingua, is
the work of a team of scholars (including Jespersen, Sapir, Martinet and Swadesh) belonging to the International Auxiliary
Language Association.

The problem of implementation is the most difficult in the spread of linguistic innovations of any type. What is needed are
i) authority ii) reasons why innovations should be accepted and iii) considerable financial resources. Esperanto in the days
before the outbreak of the First World War had all of these, though the new totalitarianism that followed made its further
spread and acceptance difficult. Support from organisations such as the League of Nations and the United Nations for
international languages has been very cautious over the years, even more local proposals such as those for a common
European language (Eurolengo, Jones 1973) meeting with very little support.
Recent work in sociolinguistics suggests that the acceptance of a language is the consequence of groups of people
identifying with one another and with a common means of communication. Another reason why people are willing to learn
languages for intercommunication, in particular pidgins or similar lingue franche, is that they expect economic rewards. As
long as members of the international community can derive more profit (in various senses of the word) from learning
languages such as English or French, the chances of Esperanto, Interlingua or similar systems remain very slender.
6.
NAÏVE LINGUISTIC ENGINEERING
A form of linguistic engineering is only now being recognised in the languages of small and often illiterate communities: the
effect of influential individuals on the speech of the whole community. We can call such linguistic engineering ‘naïve’ in the
sense that it does not proceed from any overall plan, nor is it necessarily aimed at any improvement of the language; mere
466 LANGUAGE ENGINEERING
change is usually sufficient. The principal motives would seem rather to be the need for linguistically differentiating a group
from all other groups. The area is little researched; some of what information there is (Laycock 1982a, b) comes from the area
of New Guinea, where language groups have the requisite smallness. The maximum size of a community in which an
individual can impose linguistic whims, without the assistance of any form of media, would seem to be about 8,000, since this
is the largest size that permits of a network in which each speaker is known to every other speaker, or is at least in a
specifiable close-kin relation with another known speaker. In such a community the linguistic whims of a chief, or ‘big man’,
or other forceful individual, can ‘blanket’ an entire community; in larger communities, such innovative language changes
spread only in waves, and do not reach the peripheries.
A classical example of such humanly-induced linguistic change would seem to be the metathesised forms of a language like
Ririo, in the Solomon Islands (Laycock 1982a), which corresponds to unmetathesised forms in the neighbouring, and closely-
related, language of Babatana. The fact that such a difference is not caused by regular linguistic change is seen in the
circumstance that Ririo speakers create new metathesized forms from loanwords that have come into the language only after
the metathesis must have already been established—for example, Ririo kias, Babatana kesa, borrowed from the English word

‘cash’.
An even more clear-cut case is provided by the language of Buin, spoken by some 17,000 persons in the North Solomons
province of Papua New Guinea. Less than a tenth of the population speak a very distinct dialect called Uisai. The language
has several hundreds of grammatical forms (demonstratives, deictics, pronouns) showing a gender distinction between
masculine and feminine, through three numbers (singular, dual, plural). In the Uisai dialect, all such forms have reversed
polarity—that is, all the forms indicating ‘masculine’ in the other dialects indicate ‘feminine’ in Uisai, and vice versa. In such
a case, there is no room for normal mechanisms of slow linguistic change, since the change involves a ‘flip flop’ of two
opposing categories, masculine and feminine. Such a change is analogous to the change in Sweden from driving on the left to
driving on the right; the change had to be accomplished instantly. The most plausible explanation for such a linguistic change
—although this is not subject to direct investigation at the moment—is that some influential person of the Uisai community
once announced something like ‘As from tomorrow we are not going to speak like other Buin speakers; we are going to say
‘he’ for ‘she’, and so on’. Once adopted, the change would be difficult at first for adult speakers, but would soon become
natural to younger generations; and within one generation there would be a new group of speakers to whom this was the
‘natural’ form of speech.
As both the Ririo case and the Buin case involve a single interchange of two elements—reversing final consonant and
vowel in the Ririo case, masculine and feminine in the Buin case—it seems reasonable to postulate that one of the major
mechanisms in linguistic change, whether conscious or unconscious, may be expressed in terms of interchange of binary
opposites. Certainly there is considerable evidence from speech errors, and speech pathology, of interchange of elements, such
as initial consonant clusters in spoonerisms (town drain for down train), or the phenomenon of ‘opposite speech’ (black for
white) among schizophrenics. Secret languages involving ‘opposite speech’ are frequently reported, one classic instance being
that of the Walbiri (Hale 1971). It is not surprising that naïve linguistic change should exploit a mechanism that appears so
close to the surface of our normal linguistic perceptions.
Naïve linguistic engineering is still under investigation, but further research e.g. on taboo and secret languages, may be
expected to provide insights into mechanisms of language change, and strategies for language elevation, that will be useful in
the study of more sophisticated forms of language engineering.
7.
LANGUAGE ELEVATION
The term ‘language elevation’ refers to planning activities that differ in a number of ways from the ones discussed so far
including:
(i) Planning is aimed at actual human languages that either are or, in cases such as Hebrew in present-day Israel, once served

as, the language/ vernacular of a genuine speech community,
(ii) Language engineering is usually kept at a very modest level, such a posteriori changes as are introduced not greatly
affecting intelligibility and continuity,
(iii) Planning is typically exercised for some functions, domains and media only. The spoken vernacular is not affected to a
significant extent,
(iv) Language elevation typically involves the creation of national languages rather than languages of intercultural or
international communication.
Language elevation is manifested in two domains, (a) the social domain, where status planning is carried out to enhance the
prestige of the vernacular selected for elevation and (b) the structural domain, in the form of corpus planning. Corpus
AN ENCYCLOPAEDIA OF LANGUAGE 467
planning can be further subdivided into graphisation (development of suitable writing systems), standardisation
(regularisation of grammatical patterns and elimination of lectal variability) and modernisation (adapting the language to the
needs of a modernised society). Acts of language elevation in the sense used here have their beginnings with technological
changes such as printing and political ones such as the development of nation states. The relationship between language and
the medium through which it is expressed is roughly as follows:
(a) non-literate societies can tolerate small dialects and considerable dialectal variation,
(b) literate societies with printing tend to favour standardised languages with little dialect variation,
(c) electronic media such as radio and television favour larger languages with little internal variation,
(d) computers call for one large language or a very few such.
Furthermore, the existence of large states does not promote single languages in the absence of strong central governments.
Thus, for the empires of antiquity, the existence of many languages within their boundaries made a divide et impera policy
possible. The emergence of standard written languages in the wake of the invention of printing is illustrated by the example of
German where, from the late sixteenth century onwards, a number of language societies promoted first the status of German
vis-à-vis Latin and subsequently the linguistic purity of the German language itself. The recognition of a superregional
German language was achieved with the publication of Schottel’s grammar of 1641. However, the general acceptance and
spread of spoken forms of German related to its ‘Schriftsprache’ did not occur until the nineteenth century, with the
introduction of compulsory schooling and the creation of a unified nation. The developments in Italy exhibit striking
similarities, as do developments in some of the newer nation states of the Third World such as Malaysia or Indonesia.
Numerous longitudinal studies are found in Le Page (1986 ed.), one of the few documents where language elevation is
discussed in a diachronic framework. The principal uniting factor is a picture of ad hoc decisions, inefficient and costly

solutions and lack of methodology. In a large number of instances, language elevation has continued to be carried out in a
very haphazard fashion following the whims of politicians and pressure groups rather than the advice of such language
planning experts as have established themselves in the recent past. The following sketch of an ideal case of language elevation
is hardly a mirror of present-day reality. Nevertheless, it would seem useful to bring together a number of important results
from recent work in this field.
The process of language elevation must be seen as consisting of the following four main stages:
(i) problem identification
(ii) development of concrete proposals
(iii) evaluation of proposals
(iv) implementation
Some differences in terminology and emphasis are found. Haugen (1922), for instance, distinguishes the stages of i) norm
selection, ii) codification iii) implementation and iv) elaboration. The stage of problem identification involves a diagnosis of
what is wrong both with the nationwide patterns of communication and with the structural responses of the vernacular
appointed to serve as a national language. The criteria appealed to when evaluating the situation include:
(i) referential adequacy, i.e. ‘the capacity of the language to meet the needs of its users as an instrument of referential
meaning’ (Haugen 1966:62);
(ii) systematic adequacy, i.e. a language should be structured in such a way that its rules are maximally general and natural;
(iii) acceptability, i.e. a form must be adopted or adoptable by the majority of whatever society or sub-society is involved.
Subordinate to these considerations are others, such as euphony, brevity, and symmetry between expression and content,
which are listed and discussed by Tauli (1968:38 ff). Of these, the criterion of acceptability most directly relates to the choice
of norm. This choice involves the appointment of one or more vernaculars of a multilingual nation as national languages, and
that of a suitable dialect in a linguistically more homogeneous one. In both instances, questions of political power and control
are dominant. In principle, the code selectors can adopt one of two strategies: adopt the code associated with the most
powerful or most numerous group, or else elevate a vernacular spoken by a small powerless group. Selection of the powerful
code is illustrated by the examples such as French (based on the variety spoken around Paris) and Turkish. In both these cases
the power was extended, over time, first over non-standard dialects and subsequently other languages spoken within the
nation state. Examples of small or non-powerful languages being selected as the basis of a national language are that of
Pilipino (or Filipino), the national language based on the central Luzon language Tagalog, and Bislama of Vanuata, a pidgin
English which was elevated to a national language after independence. In both instances the choice appears to have found
very widespread acceptance. Pidginised forms of vernaculars have also been given official status in the case of Swahili

468 LANGUAGE ENGINEERING
(Tanzania) and Sango (Central African Republic). (For explanation of the terms ‘lingua franca’, ‘pidgin’ and ‘creole’ see the
note under the References to Chapter 26, below.)
Code selection can involve more than one language. Thus, a number of the new Pacific nations have designated a number of
languages as ‘official’, typically the former colonial language and one or more regional languages, often assigning to them
differential status such as ‘official language’, ‘national language’, or ‘parliamentary language’. Elevating the status of some
languages or varieties tends to promote the status decline of other languages. This is manifested by (a) former independent
systems becoming subsystems of the standard e.g. Bavarian becoming a dialect of German, (b) the functional restriction to
private regional matters and (c) after a shorter or longer period of inegalitarian bilingualism, eventual language death. The
examples of Maori and Hawaiian illustrate that even large majority languages cannot survive in a political climate that offers
no adequate status.
The determination of the referential adequacy of languages is made difficult by the observation that languages do not refer
to a fixed set of pre-established entities in the real world but rather impose a referential grid on so-called reality. It is therefore
not clear for instance how many lexical entries a fully adequate language must have. Moreover, adequacy can only be
determined against the background of the functional use speakers make of a language particularly in bi- and multilingual
societies. Generally speaking, the adaptation of referential resources to communicative needs is determined by the degree of
technological and social change in a given community. The gap between what a language needs to refer to and what it can
actually refer to is greatest in times of rapid change or upheaval, such as the westernisation and modernisation of societies. The
importance of lexical corpus planning in such a context has been emphasised by a number of researchers (e.g. Tauli 1968 and
Wurm, Mühlhäusler and Laycock 1977).
Systematic adequacy refers to the extent to which languages follow regular rules, or, put differently, the relative importance
of grammar and lexicon. Old languages, particularly those which have borrowed from other systems or have been used in
esoteric functions, tend to be full of exceptions and lexicalisations. Whilst the creation of greater symmetric adequacy is a
universal motif in language engineering, there are certain external constraints, in particular the assessment of the relative
merits of regular indigenous word formation against borrowing of widely accepted international lexemes (see Rosario 1968).
An internal constraint, already raised in the discussion of a posteriori languages, is the conflict between linguistic regularity
and linguistic naturalness. Generally speaking, those in the business of language elevation have paid insufficient attention to
the second aspect.
The development of concrete proposals in the past tended to be the work of single or small groups of self-appointed
language reformers, such as Aavik in Estonia or Aasen in Norway. In countries of the Third World, missionaries often carried

out language elevation and engineering almost single-handedly. One of the best-known mission lingue franche of New
Guinea, the Papuan Language Kâte derives many of its present-day structures from the grammatical and lexical efforts of the
German missionary Pilhofer; the standard version of Central African Sango is a similar case. Committees of teachers,
linguists and politicians were involved in the elevation of Turkish under Atatürk in 1930 and Greek Katharevousa in the
nineteenth century. Teachers and politicians were also the originators of an elevated Bahasa Indonesia.
The area in which the majority of work has been carried out is that of the lexicon, in particular the creation of terminologies
and the regularisation of word-formation processes. The simultaneous action of both processes can be seen in the following
examples from Pilipino (discussed by Rosario 1968):
English Pilipino Etymology
numeral pamilang pang- (instrument prefix)+bilang (number)
integer buumbilang buo (whole)+bilang (number)
fraction bahagimbilang bahagi (part)+bilang (number)
numerator panakda pang- (instrument prefix)+takda (schedule)
denominator pamahagi pang- (instrument prefix)+bahagi (part)
Another example is afforded by proposals to increase the referential adequacy of Tok Pisin in the area of descriptive
adjectives (discussed by Wurm, Mühlhäusler and Laycock 1977). Instead of borrowing adjectives from external languages
(e.g. English derived blain ‘blind’ or hepi ‘happy’) it is suggested that an existing word formation mechanism of the language
is extended to cover new instances. Thus, the fact that syntactic constructions such as ai i pas ‘the eye is obstructed’ already
have corresponding lexical compounds (aipas ‘blind’) is exploited in new forms such as:
belgut ‘belly good’, happy
lekbruk ‘broken legged’
belpas ‘constipated’
AN ENCYCLOPAEDIA OF LANGUAGE 469