Syntactic Typology: Studies in the Phenomenology of Language

Winfred P. Lehmann

1. The Great Underlying
Ground-Plans

Winfred P. Lehmann

1.1. The Bases of Linguistic Typology

Languages, in spite of their outward differences, are formed by identical principles. This conclusion follows from the ability of humans to master any natural language, so that a child learns as native language the language or languages spoken to it. The ability persists for children and adults; they may acquire additional languages quite different in external features from those native to them. And if the new language is to be used only for simple communication they may produce an amalgam from languages of different outward patterning — a so-called pidgin. The outward differences then are shown to be surface characteristics which may be modified or eliminated while underlying principles direct the newly formed language.

Assumption of an underlying similarity among languages is also supported by the possibility of translating materials of great complexity, whether literary, scientific, or religious, from one language into any other. In translation as in the formation of a pidgin the surface differences are resolved but a common thread is retained. This underlying structure makes possible the highly abbreviated languages used by logicians and mathematicians, whatever their native language. Movement from one language to another, whether innovative like the language of poetry, or controlled, like that of logic, suggests that all human languages are based on common inward patterns and principles. This book seeks to identify those common patterns and principles, and to illustrate how they are manifested in languages of outwardly different structure.

Students of language have long concerned themselves with attempts to determine the common features of languages. Early scholars manifesting such concern, whether philosophers like Plato (427?-347 B.C.) and Aristotle (384-322 B.C.) or linguists like Panini (fourth or fifth century B.C.) and Sibawayhi (eighth century A.D.), were hampered by lack of information about a broad range of languages. In the last several centuries, however, steps were taken by philosophers, notably Leibnitz (1646-1716), and by linguists, such as Adelung (1732-1806), to overcome this shortcoming by assembling materials from many languages with the aim of determining their essential characteristics. These steps led to the activity which is the topic of this book: the typology of languages.

Typology in the broader sense is concerned with the study and classification of any selected human activities and products. Rather than concentrate on language, a typologist might deal with other social phenomena, such as games or tools — for example, digging implements. The importance of such study depends on the characteristics selected as central and the comprehensiveness and quality of the data. Thus the characterization of a language as guttural is based on external phenomena; moreover, since these have been inaccurately observed, the characterization is of little value. Characteristics used in earlier works, such as that of Adelung illustrated below, led to inadequate conclusions because the available data were not yet adequate. Subsequent linguistic study, notably in the past several decades, has achieved from comprehensive materials in a broad array of languages an understanding of language which permits a well-founded typology of language.

Such a typology is far more than a taxonomy. Taxonomic analysis is constructed on selected external characteristics. The influential taxonomy of plants produced by Linnaeus (1707-1778) is based on characteristic physical features such as stamens. A taxonomy of languages might be constructed on classification of characteristic parts of speech, or shapes of words, or kinds of sounds. By contrast with these external characteristics linguistic typology is based on the analysis of patterns and principles which have been identified as central in language, such as the structure of the simple sentence and its constituents, and processes like government, modification, and subordination. Successful typology then requires an accurate understanding of language and its elements.

Human language may be defined as a system of communication conveying meaning by means of speech sounds. Its mechanisms are treated in three distinct though related components: semantic, syntactic, and phonological. Since the semantic component is constructed at least in part in accordance with the outside world, and the phonological in accordance with parameters of speaking and hearing, the syntactic component is the most distinctive of human language. It is also the most significant for linguistic typology.

Typological analysis accordingly takes syntax as the central component. The analysis is based on the structure of the sentence, and on that of its constituents. For example, English and a large group of languages are referred to as SVO (Subject-Verb-Object) languages because the typical structure of the simple straightforward sentence involves these essential elements in SVO order, as in:

The syntactic pattern is taken as more central than morphological characteristics because it regulates and makes use of varying forms; saw, the past tense form of see, is less fundamental for construction of this sentence than is the use of a verb between its subject and object. The syntactic pattern is also more significant than the phonological, regulating phonological processes when conditions are appropriate; though the chief accent falls on the object (gosling) in a sentence like this, if one wished to identify and contrast a particular animal (a tiny gosling), the adjective tiny would receive the chief accent and gosling would have a different pronunciation from that in sentence (1). Moreover, fundamental meanings are expressed by the syntax; since in SVO languages the agent of the action is placed before the verb, the meaning of the sentence is totally different if the syntactic order is modified to:

These and many other characteristics that might be cited illustrate that the syntactic component is central in language, and therefore might be expected to provide the basic criteria for the typology of language.

Among syntactic constructions, that of the verb with regard to its object is most fundamental. Since two orders of V and O are possible, there are two types of language, VO like English, and OV like Turkish. In Turkish, sentence (1) would be expressed as follows:

The distinction between these two basic types of language is manifested in many constructions which will be identified and illustrated in the course of this book. Observation of language patterning and also investigations into control of language by the brain support the assumption that the verb is central in human language and that it in combination with an object is a basic construction.

The central position of the verb may be demonstrated in many languages, for the simplest kind of sentence may be made up solely of a verb. Such sentences are commonly found in expressions for natural phenomena, such as rain and thunder, or for feelings, such as anxiety and fear. In Classical Latin, Greek, and Sanskrit, for example, the English statement of example (4) is expressed with a simple verb form.

When such sentences consist of more than one word, the additional word is not the subject. Thus in Japanese the statement is made with a so-called subjectless sentence (Kuno 1973: 33).

Comparable verb sentences expressing feelings are extended by an object, not a subject, in Latin and German.

These examples may illustrate that the most basic sentence type consists of a verb, or a verb plus object — not of a noun, or a verb plus subject.

Sentences consisting solely of a noun or other element are abbreviatory. Thus the utterance:

in answer to the question:

is an abbreviation of the more complete sentence:

And when used in address a single noun like:

is comparable to the object of a construction like: 'I want ...' or 'I want the attention of ...'

The verb is thus the basic constituent of the sentence. Constructions consisting of a verb plus an object are the primary constructions made up of more than one constituent. Every language is accordingly classified as either VO, like English and the languages of Europe and the Semitic languages, or OV, like Japanese, Turkish, and the Dravidian languages of India.

The primacy of the verb in human language is further exemplified in the control of language by the brain. In the process of maturing, one of the hemispheres of the brain becomes dominant in the control of language; for 98 percent of human speakers this hemisphere is the left. Such lateralization has been investigated in many ways, including simple binaural experimentation. Probably the most intriguing studies have been undertaken with speakers having bisected brains (Gazzaniga 1970). The studies have indicated that while the right, nondominant hemisphere can manage utterances of nouns, especially concrete nouns, only the left, dominant hemisphere can manage the utterance of verbs. Since human language is intimately connected with specially developed sections of the left hemisphere which have the unique capability of controlling verbs as well as the information conveyed through nouns, we have nonlinguistic evidence to support the linguistically based conclusion that the verb is the most characteristic segment of human language. Linguistic typology then must concern itself centrally with the verb and its constructions.

Other features and constructions of language must be examined in their relation to the fundamental syntactic patterns. An array of features presented simply because they can be discriminated contributes little to our understanding of language, as may be determined by scanning Adelung's characterization of Turkish given below. From improved understanding of the basic characteristics of language we can now identify its most basic patterns as well as the processes and devices involved in them.

1.2. Characteristic Processes and Devices of Language

In using language we are compelled to deal with it as a linear phenomenon both in speaking and in hearing. Disregarding the more complex processes of these two activities, we may observe here that one process in the use of language involves the arranging of elements in sequence. This is one of two fundamental processes: selecting recognized elements (words) from a large set and arranging them in an accepted order. These requirements govern the fundamental structure of language, as examination of sentence (1) illustrates. This sentence consists of four words: children, gosling, the, saw. These can be arranged in one of two orders to make a normal, acceptable sentence in English, the order of sentence (1) or that of sentence (2). The device known as arrangement or order is the most important process of language. It is utilized in the phonological and semantic as well as the syntactic component; make is a different entity from came, and income is a different entity from come in. Arrangement owes its significance to the linear structure of language. Its linearity requires that entities must be arranged in order.

While arrangement is most significant, it requires in language sets of entities subject to various patterning. Such sets have long been classified and studied as parts of speech. Among these, verbs and nouns make up the largest and most important sets. They also play the most important role in the basic syntactic pattern, for nouns typically fill the role of object, in this way amplifying the central element — the verb. The process of filling the positions in any given sentence is known as selection. In producing a sentence such as The children saw the gosling, saw is selected from the possible set of English verbs to fill the V position; children and gosling are selected from the set of nouns to fill the S and O positions. Selection and arrangement in this way result from the possibilities available in a linear communication system.

In a linear sequence of entities two processes are most prominent: arrangement and selection. Yet because of the possibilities inherent in the mechanisms used for speech, two further processes are found in the production of speech: modification of the selected elements, chiefly their beginning or their ending though also often the entire element, and intonation, that is, melodic variation superposed on the linear unit. Both processes are associated with concentrations of energy in speech, that is, with distribution of stress or pitch patterns. For example, when the strongest stress is placed on the pronoun in the following sentence:

the pronoun has a full vowel, in contrast with its pronunciation under the following stress placement:

The variation of us results from the process known as modification.

Modification is associated with differences in meaning and is accordingly significant — signaling (14) as an exclusive first person plural (= we but not you are to go) and (15) as an inclusive (= all of us are to go). Varieties of modification are closely associated with characteristic types of languages. VSO languages foster modification of initial elements, OV languages of final elements. But the uses of modification in language are less central than arrangement and selection, and accordingly less significant in typology.

Intonation is also less significant, though for a different reason. Some characteristic patterns, such as final rise of pitch to indicate uncertainty, are widespread, virtually to the point of being universal. If, for example, sentence (15) is uttered with final rising pitch:

the utterance is an expression of surprise or challenge (= you really mean we should go?). Intonation patterns are accordingly in great part regulated by general principles, rather than by the specific principles which lead to a characterization of language.

As the processes leading to linguistic sequences are applied, two forces are at work: control and delimitation. Control involves the domination of one element by another. It leads to hierarchization. It has been most widely investigated with regard to verbs controlling nouns, where the force is traditionally known as government. In the central segment of sentences a governed element is often identified by a change in form, e.g.:

The change in form of the substantive rather than of the verb reflects the dominance of the verb in this construction.

Another word class involved in government is that of adpositions, prepositions in VO languages and postpositions in OV, e.g.:

The patterns of control in verbs and adpositions are highly transparent.

More subtle control is found in causative verbs, or in verbs accompanied by a complement; yet in such patterns the control is in accordance with the arrangement of verb and object. In the following English sentences the controlling verb precedes the controlled.

The extent of control in such sequences varies; in sentences (l8) and (19) the actions of "resigning" and "going" were carried out, while in (20) it is not certain that "negotiating" was carried out. Control or government plays an extensive role in language, with subtleties that still demand much study.

Delimitation may seem similar to control, yet it operates differently; the delimiter modifies rather than controls and is thus subordinate to the element it delimits; when an extended modifier, essentially an entire clause, is incorporated in a construction, the process is referred to as embedding. While a controlling element often brings about change of another element, as illustrated above, a delimiting element is commonly itself changed to "agree" with the element it modifies, in this way reflecting its secondary role. The process is known as agreement or concord.

A characteristic pattern for modifying nouns is that known as relativization. In English, relative clauses provide examples of the changes found in embedding, using who for animate, generally human, nouns in contrast with which for inanimates:

Modifiers however are commonly identified simply by their arrangement rather than by the process of selection. Thus adjectives in English are placed before nouns, with no inflection to indicate congruence. Similarly, in Japanese, relative constructions are identified by their position before nouns.

While relative constructions are the characteristic modifiers of nouns, relationships may also be expressed which are less direct, as illustrated in the following example.

The last clause in (24) and (25) is not viewed as a relative construction but rather as a subordinate clause modifying a verb, even in (24) where it is adjacent to a noun — hence it is classified as an adverbial clause. Because conjunctions commonly take over the role of relating subordinate clauses to matrix clauses, as in example (25) as opposed to (23), grammars distinguish sharply between relative and adverbial clauses.

Adverbial clauses may indicate "agreement" variously with some element involved in the embedding. Special verb forms, such as subjunctives, may be used, pronouns may be modified, and so on, as in example (27).

If a clause is the object of a verb, it is referred to as a complement. Like adverbial clauses, complements are comparable to relative clauses, as the following examples illustrate.

Complements too reflect in various ways the relationship with the verbs accompanying them.

In the expansion of the central constructions of language, either OV or VO, government and agreement are primary forces, as the constructions discussed above and the selected patterns presented in the following section illustrate.

1.3. Selected Patterns in VO and OV Languages

The typological sketches of selected languages given in subsequent chapters will indicate how comprehensive are the constructions regulated by the basic structure of a language. This section includes a brief number of those constructions which are given here to illustrate the processes discussed above, and to provide concrete examples for some of the patterns listed in section 1.4. For these illustrations Modern Irish and Modern Sinhalese are used. Both are Indo-European languages. Yet in the course of their history they have become strikingly different in structure. Irish is a highly consistent VSO language, Sinhalese a highly consistent OV language.

In examining languages for characteristic typological patterns we distinguish these more clearly for the VO type in the VSO subtype than in the SVO subtype, for in this the subject must be taken into consideration when verbal modifiers are introduced. SVO languages have characteristic constructions like modal auxiliaries, due in part to the placement of the subject before the verb, as will be observed in Chapter 4. In VSO languages, on the other hand, this problem is not met, for in them as in OV languages verbal modifiers can be placed between the V and # (sentence boundary) with no interference from other constituents.

As indicated in section 1.4, typological characteristics will be given in accordance with nine patterns, each with subclasses:

• [IV. Sentence adverbials] [Note: *]
• V. Compound and complex sentences
• [VI. Marking] [Note: *]
• [VII. Grammatical processes] [Note: *]
• [VIII. Morphological characteristics] [Note: *]
• [IX. Phonological characteristics] [Note: *]

[Note: *Omitted for reasons of space.]

From this list it is obvious that complete accounting for these patterns in any language would require an extensive grammar. It is one of the aims of this book to provide guidelines for improved explanatory grammars; yet the examples introduced with this aim must be severely limited in the interest of pedagogical effectiveness. Accordingly, primarily the first three patterns will be illustrated here.

The process determining the arrangement of the constructions under I is government. The patterns exemplified are parallel. Just as a verb governs its object, so does an adposition. It is accordingly understandable that VO languages include prepositions while OV languages include postpositions.

While the constructions with a standard do not strictly speaking reflect government, the relationship between a variable and its standard is comparable to that between a verb and its object. Thus the comparison of inequality involves a situation in which a variable is related to a selected object or standard. In such a comparison the standard may be an object like a cat or a house and so on. In other constructions involving a standard, such as names, the same relationship is taken into consideration: the standard is the family name, with possible variables selected from titles or given names. In a VO language, the variable element precedes the standard, and in an OV language the reverse is true. Curiously, this sequence also applies to addresses, as in Japanese before addresses were influenced by international postal conventions. The name of the largest entity, the state, was placed first, followed by designations for progressively smaller units and finally the surname of the addressee. The order is also observed in other "additive" patterns such as numerals compounded on the basis of adding, a pattern commonly used to construct "teen" numerals. In these the lower digit is given with reference to a specific standard, for example, 3 with regard to 10 = 13.

The patterns regulated by government are relatively few, yet they make up the principal constructions of the simple clause.

When the constituents of the simple clause are modified, the modifications are carried out in accordance with a fundamental principle for both nominal and verbal elements (Lehmann 1973a). By this principle, which is given below, the central sequence, whether VO or OV, must not be interrupted. Any modifier is then placed between the modified constituent and the sentence boundary.

The symbols used here will be accounted for more fully in 1.8 below. In the grammar on which the principle is based, N^Obj. is used for an object, Q or Qualifier for a verbal modifier. In accordance with the principle, nominal modifiers precede nouns in OV languages and follow them in VO languages; verbal modifiers follow verbs in OV languages and precede them in VO languages. The constructions involved include many of the patterns in language.

The basic construction for introducing descriptive modifiers with nominal elements is the relative clause, as noted above. In this construction, two sentences with an equivalent noun or noun phrase are combined, as syntacticians have stated for more than a century. Thus the two sentences given as (30a) and (30b) may be combined if the noun phrase the dog has the same reference.

In accordance with the principle the modifying sentence is embedded after its antecedent in VO languages, before it in OV, as illustrated in Irish and Sinhalese (II.5 above). The embedded sentence then is generally reduced, especially through "deletion" of one of the equivalent noun phrases; the results vary in detail from language to language. If the verb in the embedded sentence is the copula (BE), it as well as the equivalent noun phrase is deleted, leaving a descriptive genitive, as in (II.6), or a descriptive adjective, as in (II.7). Other nominal modifiers are not illustrated here, nor are multiple nominal modifiers, such as adjectives plus relative clauses with one noun.

Verbal modifiers affect the entire clause, as might be expected from the role of the verb. Here only two frequent verb-modifying constructions are exemplified, interrogatives and negatives. As in Sinhalese and Irish, these modifiers may be expressed through particles. These particles typically stand between the verb and sentence boundary, as in the three sentences of (III.11) and (III.12) above. In SVO languages, however, other devices are commonly used, such as auxiliaries in English or shifts of word order in German.

Further, the verbal modifier may be expressed through intonation, as in the echo questions of English:

Intonation is the dominant means of expression for declarative statements (III.10), though some languages also use particles for it; the particles concerned are then closest to the sentence boundary. And, as the combined modifiers in the sentences translating Didn't John see the dog? illustrate, the expressions for interrogative and negative are generally placed in this order with regard to the sentence boundary. Such placement of surface markers reflecting underlying hierarchy of linguistic elements is remarkable. It is also troublesome, for it may lead us to overstate how surface patterns mirror underlying syntactic structures; they may do so, especially in OV languages, which have strings of elements marking the verbal modifiers generally arranged in the order indicated in the section from (III.10) to (III.19).

Of compound and complex patterns only complement structures (V.25) are exemplified here, largely to illustrate the comparability of such patterns with single elements. As noted above, complements can be compared with objects. They observe the order of objects with regard to their verb, as the examples indicate. But, like all extended constructions, they vary in accordance with the type of language, as will be illustrated in the chapters dealing with individual languages.

The four groups noted here — I. simple clauses; II. nominal modifiers; III. verbal modifiers; V. compound and complex sentences — are those best understood in typological patterning. The grammatical processes listed in VII are becoming better understood, as a result of recent study. On the other hand, sentence adverbials (IV) have been largely untouched.

The process of marking (VI) has also been dealt with only peripherally, with references to study of it in phonological and stylistic investigations, especially as carried out by Roman Jakobson. Marking involves various devices from language to language, such as shifts in the basic clause order, sentence adverbials, the processes included in VII, and special intonation patterns., as in the marked variants of the following sentence.

Marking then can be carried out in various ways, so that it could be treated at length. Here correlation of various marking devices with language types and subtypes will be of primary interest.

Grammatical processes also vary in prominence in the various language types and subtypes. Clefting, for example, is particularly prominent in SVO languages. Passivization is prominent in SVO languages, but not at all in OV languages; it is essentially a tool for achieving topicalization for the object, and such a tool is unnecessary in OV structures. Moreover, OV languages are poor in pronouns. As with marking, variation in the use of grammatical processes among the different types will be of major interest here.

Relationships between types and the characteristics included in sections VIII and IX are little known. Yet it has been clear for some time that VSO languages are characterized by prefixation and OV languages by suffixation. Moreover, VSO languages are notable for initial morphophonemic modifications, while OV languages undergo final modification. In general, the structure of clauses has diminishing effect as the device in question is less central. Thus it is difficult to suggest any correlation between the number and kinds of vowels and a given typological structure. Less concrete phonological features, like syllabification or use of supra-segmentals, may however show such correlations. Thus OV languages tend to have simple syllables, of C(C)V structure. Moreover, the effect of phonological processes is "progressive" in OV languages, "anticipatory" in VO. OV languages tend to have vowel harmony and progressive assimilation; VO languages tend to have umlaut and anticipatory assimilation. Yet for such processes to be productive a language must have the requisite structure of words and syllables. Correlations between phonological patterning and language types must then be examined with delicacy.

Such correlations, as well as other characteristic typological patterns and processes, will be better known and described as further languages are investigated for their characteristics. The following section presents in schematic form patterns which are dealt with in the subsequent chapters here, and might well be the basis for examination of any language that is being studied.

1.4. Patterns and Processes

In typological investigations certain patterns and processes have been identified as characteristic for possible language types. These are listed here. Some may well be expanded, as I and (I.4) are here. Others may need amplification, such as III.A and III.B, expressions for modality, aspect, and tense. Most have been extensively investigated, though as noted above others are poorly known, such as IV, sentence adverbials. Undertakings to illuminate murky areas, such as causative expressions and topicalizations, have begun to illuminate more of the patterns and processes (Shibatani, ed. 1976b; Li, ed. 1976). This book has been designed in part to arouse further such study.

The patterns and processes are as follows.

• VIII.C. Morphophonemic processes; sandhi
• IX. Phonological characteristics
• IX.A. Syllabification
• IX.B. Suprasegmentals
• IX.C. Segmentals

A full account of each of these topics for any language would lead to a lengthy statement, virtually as extensive as a grammar. Rather than such an extensive account, the characteristic patterns of syntax and the processes they undergo are aimed at here.

1.5. Earlier Typological Studies

It is instructive to compare the criteria used in earlier typological studies. Two major undertakings have dealt with the characteristics of individual languages, Adelung's Mithridates (1806-1817) and Müller's Grundriss (1876-1888). Adelung's is far more extensive; Müller's undertaking was limited by his relatively early death, with no collaborator to continue his work. Subsequently published studies have been more limited in scope (Finck 1909), or directed at specific characteristics (Schmidt 1926) or at external accounts of languages (Meillet and Cohen 1952; Voegelin and Voegelin 1973). Even Müller, in spite of his assertion of progress beyond Adelung, is more restricted than the Mithridates in concentrating largely on the morphology of selected groups of languages. Accordingly an excerpt will be given only from Mithridates, to illustrate earlier typological study. The excerpt will consist largely of Adelung's characterization of Turkish, in translation; Turkish citations are given in the spelling of today, and printing errors have been corrected, but the text is otherwise maintained.

Adelung's purpose is clearly stated in introductory essays. The following sentence from one of his prefaces summarizes his aims: "The most important thing for me was to penetrate into the inner and outer structure of each language, because only in this way can the distinctiveness of each and its difference from all others be recognized" (1806: Vorrede XIV). As will he noted from his account of Turkish, Adelung pursues these aims, giving morphological elements or outer structure as in (7), as well as internal patterning as in (9).

The characteristics are arranged in accordance with the usual order of grammars, with statements on the phonology given first, followed by statements on morphology and syntax. There is no attempt to state the characteristics of Turkish in reference to a general framework for languages, except in (1) and (2), which use the languages of Europe as norm, especially the Germanic. Such concentration on one's own language and ethnocentricity are unfortunate, but not unknown even among linguists today. In spite of their shortcomings, Adelung and his successors are to be commended for identifying characteristic patterns such as verb-final position (17), the presence of postpositions (16) and the type of comparison (9). Since Adelung's insights and shortcomings are readily apparent, no further commentary will be given here.

Mithridates I.462-466:
Character of the Turkish Language

• (1) In this language not only is the old Tatar basic material mixed with Arabic and Persian words, but there is also much that is Germanic in it, which permits one to conclude that there was a close relationship of both peoples, presumably in their first habitations in Central Asia.
• (2) Otherwise it has a very distinctive character, and only in its substantives and adjectives some similarity with European languages.

• (3) It has also the letters of the Persians, besides their sound, and in addition still, a nasal, accordingly in all, 33 letters.
• (4) The inflectional and derivational sounds at the ends of words are very manifold.
• (5) The accent most frequently falls on the last syllable, even in polysyllabic words.
• (6) Substantives have no gender, except by meaning. In order to avoid ambiguity the natural genders are distinguished by special modifying words: karindaş means 'brother' and 'sister'; in order to distinguish them, er karindaş 'brother', kiz karindaş 'sister'. Augmentatives, diminutives, nouns of action, of place, of time, etc., have their own derivational syllables.
• (7) The Turk has no article, but on the other hand a declension with six very definite cases, including the vocative, in final syllables, which are added to the plural after it is formed, as well as to the singular. The plural is always formed with the syllable ler. A dual the Turk doesn't have. E.g., er 'the man'.

  Singular Nominative er   Genitive erin   Dative ere   Accusative eri   Vocative Ya er   Ablative erden

  Plural Nom. erler   Gen. erlerin   Dat. erlere   Acc. erleri   Voc. Ya erler   Abl. erlerden

• (8) Adjectives designate neither gender, nor number, nor case, but are added to substantives like adverbs: güzel er 'a handsome man', güzel erler 'handsome men', güzel erlerden 'from the handsome men'.
• (9) The comparative is either expressed through a preceding ablative: erlerden güzel 'handsomer than the men'; or through the adverbs 'much, very' etc. or also through a suffixed syllable, güzelrek 'more handsome'. The superlative is paraphrased by a further word.
• (10) The pronouns do not designate gender, but are declined in all six cases. Possessives have the special characteristic that they at the same time place the genitive of the personal pronoun in front of the substantive, as if one would say in Latin tui frater tuus, tui fratrem tuum, tui fratres tuos ('your brother your').
• (11) The verb is the most difficult segment of the Turkish language, because of the quantity of moods and tenses, and of their special inflections. The Turk has an indicative, subjunctive, and optative. Of tenses he has a present, two imperfects, two perfects, a pluperfect, and four futures, the latter with regard not to time relationships but to inflection. Furthermore, an imperative, infinitive, participles, and gerunds. The conjugation can be explained for the most part from the combinations of the the verb BE with the participle. All infinitives end in mak or mek. One might believe they were actually phrases compounded with the word machen 'make': make (a) cut rather than cut, make love rather than love, especially since this make disappears completely again in the remaining conjugation.
• (12) The passive is formed by intercalating an l between the mak or mek of the infinitive, and in all moods and tenses between the main word and the personal ending: sevmek 'love', sevilmek 'be loved'; sever 'he loves', seviler 'he is loved'.
• (13) In the use of participles the Turk is like the European, especially in the fact that he uses them so often for the formation of his conjugation.
• (14) Negation is brought about in the verb through insertion of m or ma: olmak 'be', olmamak 'not be'; sevmek 'love', sevmemek 'not love', sevdim 'I have loved', sevmedim 'I have not loved'; and thus in all moods, tenses, and persons. Here too there is a phrase-analog, which can be explained by love not make, especially since ma is also a negative particle in Arabic.
• (15) On nouns, negation is expressed through the syllable siz or suz: korku 'fear', korkusuz 'fearless, unscared'.
• (16) Prepositions are postpositions in this language.
• (17) Syntax is far more complex than in Semitic or in the Persian language. Since the Turk has specific cases in declension, he permits inversions, where frequently the governing word, as in Latin, stands quite at the end of a relatively long sentence.
• (18) The Turkish language is just as rich in compounds as the Persian, namely of those that arise through simple compounding, without inflection: daǧ-burunu 'mountain nose', i.e., 'foothills'; çayır-kuşu 'field-bird', i.e., 'lark'; arpa-suyu 'barley-water', i.e., 'beer'; buz-at 'ice-horse' (buz 'ice'), i.e., 'a gray horse', frengi-çeşm 'Frank-eye (European-eye)', i.e., 'a pair of glasses'. Such compounds the Arab cannot make. In order to give a name for steel, the invention of which he ascribes to the Europeans, the Arab says hadid afrendschi 'Frankish iron'. Accordingly he uses the adjective because he cannot, like the Turk, say 'Frank-iron'; he would then have to say in the genitive plural 'iron of the Franks'; that however is no compound.

Adelung's characterization of Turkish may be taken as representative of his typological approach. The Turkish statement is shorter than some, longer than others, especially those for indigenous languages of the Americas and Australia; these may have been the languages which Müller considered more fully treated in his Grundriss. While it may seem highly compact, it should also be noted that Adelung included five hundred languages in his undertaking. Such an aim would be totally unrealistic today, for we do not have adequate information on five hundred of the world's languages to state their patterns and processes in accordance with the outline given in section 1.4. This shortcoming in linguistic study can be remedied if students of individual languages describe them adequately so that descriptions like those presented in Chapters 2-5 are available; such descriptions could be far more extensive than is possible here.

1.6. Implications of Typological Study for the Description of Languages

Typological analysis is of fundamental importance in linguistics because of the framework it provides for the description, explanation, and understanding of individual languages. These aims represent the goals of linguistics. They have been pursued in various ways. Traditional grammars set out to present languages in accordance with a time-honored approach based on previous work; in Western linguistics this work goes back to the Classical grammarians. The approach is constantly updated, in part with reference to other traditions, notably the Indian. Jespersen presented the views of his generation in two influential books: Language of 1922 and The Philosophy of Grammar of 1924. Bloomfield in his book Language, which more than Jespersen's is in the German tradition, "tried ... to present the accepted views" of his day (1933: viii). Like their predecessors in Western linguistic theory, both draw only on SVO languages, thus presenting a seriously flawed treatment of linguistic theory. One need only examine the treatment of the comparative by either of these influential scholars to note their unfortunately restricted views (Jespersen 1924: 244-253).

Some linguists have broken more sharply with traditional grammar, developing linguistic theory in accordance with other disciplines, notably logic. Hjelmslev and Chomsky are examples. Yet they too have based their theory on examination of the languages of Europe, notably English for Chomsky and his followers. Accordingly a rule as fundamental as that specifying the structure of the sentence is given as ordered by Chomsky and as though language is fundamentally SVO in structure: S → NP Aux VP. Attention to other disciplines has not therefore led such linguists to a more fundamental understanding of language. In view of the concentration on just one subtype of language, these linguists also "are deficient in that they leave unexpressed many of the basic regularities of the language with which they are concerned" (Chomsky 19-65: 5). Linguistic theory will be adequate only when it is based on study of all types of languages. Since typology sets out to determine and explicate these, it is fundamental in the projection of theory as well as in the description of languages.

If, as is generally held, language is constructed in accordance with certain fundamental rules, these should be determined by typological study. When fundamental rules are proposed and the patterns governed by them, the resultant view of language may however seem too rigid. Thus the patterns listed in section 1.4, as well as the additional morphological and phonological patterns not specifically stated there, may appear to reflect a view of language as a static product of an automaton, not the creative instrument of human speakers. Yet the patterns are governed by only two axioms, and accordingly are not as manifold as a lengthy list may suggest. The first is that government tends to operate similarly on all structures in which it is involved in a given language. The second is that modification also takes place in accordance with a general principle. In view of the rapidity of language acquisition by children and in accordance with observed linguistic phenomena, the assumption of two such axioms is highly plausible. Moreover, a language may accord with the ideal scheme, or nearly so. The languages presented in Chapters 2 and 3 closely approximate ideal types.

Yet there are inconsistencies. A notable inconsistency in Japanese has to do with the expression for reflexivization in an OV language. In accordance with the discussion in section 4.3, it should be expressed through suffixes, as it indeed is in many OV languages. Japanese however uses a substantive, zibun 'self', also labeled a reflexive pronoun. The constructions in which zibun is used are highly intricate, creating many problems for grammatical description (see Kuno 1973: 291-323). On the basis of these problems one might propose that zibun is irregular, even foreign to the structure of the language. We know the history of Japanese adequately to identify zibun as a borrowing from Chinese. This knowledge makes us receptive to acceptance of its irregular status and leads us to seek an explanation for its introduction as well as to propose the expression for it in the earlier stage of Japanese (Lehmann 1974b). The Japanese pattern of reflexivization also illustrates the importance of examining languages with reference to a typological framework. With it we identify problems in the structure and history of languages, and attempt to provide solutions.

Another such problem is the use of VO type relative and subordinate constructions in Turkish, as in:

In contrast with the expected preposing of embedded clauses in an OV language, these constructions are felt to be foreign; today they are infrequently used, in accordance with puristic tendencies after the Atatürk revolution.

While such examples need little clarification, since Japanese and Turkish are well documented for the last millennium, structures in languages with less data may be recognized and accounted for if these languages are examined in accordance with the typological framework now available. With it we can account for the aberrant comparative patterns like Latin tē maior 'from-you bigger = bigger than you' in the early periods of the Indo-European languages. The comparative is only one of the structures to be treated in this way. Another is the use of adpositions. Even the eminent syntactician Jacob Wackernagel provides only an awkward account of postpositions in the Classical languages, referring to them as "nongenuine prepositions" (1928: 157). An example is Greek dé, in the Phrase Oúlumpon dé 'to Olympus', which is cognate with the English preposition to. Its cognate zu is also preposed in German, and accordingly Wackernagel notes that "the word has developed to a full preposition" (1928: 157). But he does not provide the explanation that such a development would be expected when a language changes from OV to VO structure.

Languages then may not be consistent, not even as consistent as Japanese vith its imported reflexive as one of its very few non-OV patterns. The reasons for inconsistencies are those noted for the examples above: (1) diffusion of characteristic patterns from one language to another, or borrowing, as it is known in linguistics; (2) internal change, whether from OV to VO structure, as in most of the Indo-European languages, or from VO to OV, as in Sinhalese, Chinese, and many languages of Africa. Until such changes are completely carried out, the languages in question may be labeled inconsistent.

Identifying languages as either consistent or inconsistent is not sufficient to account adequately for variant typological forms of language. In a period of a change from one type to another a language may show both VO and OV patterns, with neither predominant as they are in inconsistent as well as consistent languages. Such languages may be referred to as ambivalent.

One construction has been accounted for through the understanding that ambivalent languages must be recognized, the so-called absolutes. Absolutes are embedded sentences which have no formal relationship to the matrix clause, as in the following ablative absolute from Caesar's Gallic War I.6:

Absolutes can be accounted for by assuming that the language is so structured that it is unclear whether an embedded segment should precede or follow, and accordingly it is unattached. They were prominent in Classical Greek and Sanskrit as well as Latin, though not in earlier or later stages of these languages. Later stages of Greek and Latin came to be SVO, so that embedded elements followed their head — and accordingly absolutes were no longer in keeping with the structure of the language and were not used. Until languages were examined for their typological structure, the absolute construction was unaccounted for. The explanation given here at once clarifies it and its abandonment, and illuminates the kind of language in which absolutes are found.

The increased understanding of language resulting from typological investigations is especially important in accounting for languages with scanty data. If crucial constructions can be identified, the language may be clarified more than the amount of data would suggest, even though it was not consistent. Thus the language of the Mohenjodaro inscriptions is doubtless OV, in view of its suffixation, even though the texts have not yet been surely interpreted. And in spite of their confused and scanty texts one might propose that the Tasmanian dialects were VSO because adjectives and numerals were placed after nouns. Similarly, hypotheses may be proposed concerning the structure of the languages of Anatolia in the second millennium B.C. Hattic with its multiple prefixes was VSO. Even the presence of conflicting patterns may permit improved understanding of languages, especially as our grasp of the relationships between the various characteristics of language types becomes more secure. Hittite, for example, with its various OV characteristics inherited from Proto-Indo-European and with VO characteristics like its sentence-initial clitics, may have introduced the VO patterns from neighboring languages, much as Sumerian has. Further studies of these and other languages will illuminate both their interrelationships and the patterning of languages in contact with one another.

Typological study accordingly provides guidelines for identifying characteristic patterns in the study of any language. These patterns should be accounted for in all grammatical sketches, as they are in the following chapters. Moreover, various processes affecting them must be noted and elucidated. It will not do simply to present conflicting patterns in specific periods or individual authors, such as statistics on the number of clauses in which verbs precede their objects and the number in which they follow their objects or stand in even more complex relationships. Such statistics are valueless unless each order is interpreted. The normal pattern must be identified, and thereupon the patterns used for marking and other special processes, such as topicalization. When it is pertinent, the sociolinguistic situation should also be taken into consideration, as when speakers of two or more languages are in contact. It is highly likely, for example, that late Middle English authors were influenced by French, as in the arrangement of adjectives In short, the guidelines provided by typological study identify not only patterns which must be ascertained, but also variant forms and the reasons for these. Consistent languages like the Dravidian or those used as examples here are indeed found, but more commonly languages have unexpected constructions, like the preposed adjectives of modern English in contrast with the postposed position of French and the other Romance languages. If adequate background materials are available, such constructions may be accounted for, either by sociolinguistic or by historical explanations.

Explanations of many patterns of languages are possible on the basis of typological studies. Such explanations must be viewed as major tasks of descriptive linguistics.

1.7. Implications of Typological Study for the History of Languages

Through the identification of aberrant patterns typological analysis provides perspective for the historical study of language. English furnishes examples, both in its present form and earlier. As illustrated in section 1.3, adjectives should follow nouns in a VO language, as they do in French, Spanish, and the other Indo-European languages of southern Europe. From their aberrant position in English one may propose the hypothesis that English is a language which has been changing from OV structure to VO structure.

By itself the aberrant position of adjectives would not be decisive for support of this hypothesis, because adjectives precede nouns in relatively many SVO languages. But English had additional OV characteristics in earlier periods. In Old English, genitives also preceded nouns in 90 percent of their occurrences. In the millennium between Old English and the present the order of genitives has gradually been changed so that today 90 percent of them follow their nouns (Fries 1940: 199-208); unless animate nouns are involved, genitives are generally expressed today with of constructions. The direction of the syntactic change suggests that additional OV constructions might be expected in Old English. Such constructions are indeed attested, especially in the earliest poetic texts, such as the Beowulf.

Besides the OV order for adjectives and genitives the Beowulf includes government constructions of the OV pattern. According to an eminent editor, end-position of the verb predominates in the Beowulf (Klaeber, ed. 1950: xciv). Adpositions may follow the element they govern, such as mid in the line below.

OV type comparatives also occur, as in Beowulf 1850 with the standard be preceding the comparative sēlran:

The breadth of OV constructions in the oldest English texts then supports the inference based on the change in order of the genitive that English at an earlier period was strongly OV in patterning. This conclusion leads to improved understanding of the history of English, permits explanation of aberrant constructions in its early period, and illuminates the gradual development of patterns like the passive and modal auxiliaries.

In addition to these developments many others are clarified. Some have to do with morphological change. Suffixed inflections, which are characteristic of OV languages, become fewer and fewer in English. Instead preposed particles are introduced as grammatical markers. Besides of for genitive constructions and to for datives, adverbs have been introduced before the standard in comparative constructions; the variety of particles &#8#8212; as, be (by), nor, or, as well as the ultimately general than — suggests by itself that a new pattern was developing. Similar observations may be made for all the Western Indo-European languages; the patterns listed in section 1.4 are clarified in the course of their development in these languages by examining them in relation to a change from OV toward VO structure.

The opposite direction of change can be observed for Sinhalese, from VO to OV. Like Sanskrit it had developed toward an ambivalent language around the beginning of our era, with many SVO characteristics. Thereupon, heavily influenced by the neighboring Dravidian languages, it changed to its virtually consistent OV patterning of today. Characteristic patterns of the language of the present were illustrated above. Further additive numerals are given here, which in contrast with the forms in Old Sinhalese and Classical Sanskrit, show the OV order in the teens, with exceptions explained below.

It is noteworthy that the VO order of Old Sinhalese is preserved in some numerals of contemporary Sinhalese; like that for 11 presented here, those for 12 and 15 have not been modified. The numerals for 11 and 12 are also maintained in aberrant patterns in the Germanic languages (11, 12 = 'one/two left over/ as opposed to 13='3 + 10', etc.). The numerals for 11 and 12 presumably maintain aberrant patterns because numeral systems often include them with the set from 1 to 10, and hence would not associate them with the group to be restructured.

Whatever the reasons for retention of old patterns in some of the teens, irregular sets, like that of the teen numerals in contemporary Sinhalese, will in time be used by historical linguists to reconstruct earlier patterning in languages by means of methodological principles based on observations in well-documented languages, even when older materials are unavailable. Tentative attempts of this kind have been made in dealing with African and Amerindian languages attested only today (Givón 1975; Hyman 1975; Lehmann 1975b).

The use of a typological framework will then permit explanations for many syntactic, morphological, and phonological phenomena. The analytic procedures and the phenomena are parallel to those long observed in the study of phonology and morphology. A small number of irregular forms, for example, are generally relics of earlier patterns, such as the plurals of common English nouns like goose : geese, mouse : mice, and so on. Such irregular forms generally survive among the frequently used everyday words; it is assumed that in language acquisition they are mastered as individual entities rather than as members of a paradigm like that generally used for English plural formation — book : book + s. The English plurals with internal change are survivals of plural formations of a paradigm with high front vocalic suffix that were once more widespread but have gradually been replaced by s-plurals. In much the same way a number of common adjectives are preposed in French rather than postposed, for example bon 'good', grand 'big', petit 'small'. Their arrangement may be accounted for by viewing them as relics of the earlier OV order, which had for the most part been replaced even in Latin.

Investigations using a typological framework in historical study of well-documented languages will then provide guidelines for improved understanding of the development of languages. Among findings which will permit extrapolation for earlier periods, such as pre-Old English, or for languages with no earlier materials, such as those of Africa and the Americas, are conclusions on rate of retention, periods required for specific shifts, and interrelationships of changing patterns.

The shifts in English and French lead us to propose different rates of retention for specific structures. Among the nominal modifying patterns the rate of retention is greatest for descriptive adjectives; relative constructions apparently change position earliest, followed by the genitive and subsequently the adjective, and eventually limiting adjectives and adjectival numerals. Such hierarchies of maintenance must be verified from study of all languages.

Further, if the shift of the English genitive noted in Fries's study is taken as typical, a millennium may well be required to carry out changes in one of the nominal modifying constructions. Such studies must be pursued on changes in other patterns where data are available, as in Sinhalese. Another such area — Latin and the Romance languages — can yield figures on changes from a predominantly suffixing verbal system to one in which many forms are periphrastic. And though the situation in the early stages of Latin is not well documented, the shifts in such constructions as comparatives (from tē maior 'you-from bigger' to maior quam tū. 'bigger than you') and adpositions may provide data on the periods involved in carrying out these shifts as well as the interrelationships between shifts in the various constructions noted in section 1.4.

Interrelationships between changes of patterns have also been hypothesized on the basis of very little data to the effect that datives may be among the first patterns to be shifted in OV languages, when postposed as a kind of afterthought, as in example (37), from Vedic Sanskrit, in contrast with (36).

Such placement of the dative, leading toward disruption of the OV structure, signals a shift in basic word order; it would be followed in turn by a change in other government constructions as well as in nominal modifying constructions like relative clauses.

The sequences of shifts, as well as identification of pertinent patterns, must be confirmed by observations of further languages and their changes. In its concern with a wide number of languages, typological study is closely related to the study of universals, for the processes as well as the structures examined are identified by relationship to them.

1.8. Typological Study and the Study of Universals

The study of universals has been pursued vigorously in recent investigations of human behavior and society, not merely in linguistic study. The most notable event arousing interest in such investigation of language today was a conference held in 1961 which led to the subsequent publication of the papers, entitled Universals of Language, edited by Joseph H. Greenberg. The conference was arranged after circulation of a memorandum by Greenberg, C. E. Osgood, and J. J. Jenkins which included the following definition: "Language universals are by their very nature summary statements about characteristics or tendencies shared by all human speakers" (Greenberg 1966: xv). The definition clearly resulted from haste. For if only statements applicable to "all human speakers" were formulated, they would be severely limited; many of them might well be trivial, for example that all languages include consonants, words, and sentences. After the conference the statement was amplified as follows: "There was general agreement that it was necessary and completely legitimate to include as universals in addition to statements of the simple type 'all languages have a given feature x,' likewise implicational relations, universal frequency distributions, statistically better than chance correlations, and other logic types ..." (Greenberg 1966: xii). The study of universals was accordingly broadened to deal with more than "summary statements about characteristics or tendencies shared by all human speakers."

The "logic types" comprise matters of great interest to typological study. If one defines processes like arrangement and modification as "features," it is indeed of importance for typological study that "all languages have" them. Features of major concern, as noted above, are patterns motivated by others, including those with "statistically better than chance" correlations. In surveying languages for such features, typological study has yielded major contributions to general grammatical theory, for it has made clear the segment of the grammar which is indeed universal, and has disclosed guidelines for its patterning.

Among these guidelines is the observation that the basic patterns in the universal segment of grammar must be unordered. While one may assume that sentences will contain verbs and objects, their arrangement is determined not by universals but by typologically specific rules. One kind of grammar positing unordered underlying rules has been labeled case grammar. In such a grammar the sentence rule specifies a verb, but neither subject nor object. These are subsequently introduced from abstract nominal categories in accordance with the lexical item functioning as verb in a given sentence. The resultant grammar is in keeping with observations about language, and leads to identification of the patterns listed in section 1.4. An initial rule specifies both the propositional constituent of sentences and the modalities or qualifiers expressed in it. Since language permits arrays of two or more sentences, the initial rule in the grammar provides for such construction.

The terminology and notation utilized are of secondary importance, though symbols are best selected which avoid ambiguity. Since the abbreviations V = verb, S = subject, O = object are well installed in typological study, the unambiguuous symbol Σ is chosen to indicate sentence. And since modalities make up only a portion of the qualifying elements of the sentence, the symbol Q, suggested by Seuren, represents these. The universal rules in grammars of any language may then be formulated as follows.

• 5. K → {NP_Σkⁿ}
  (k=inflections, or function words, selected in accordance with the lexical entries for V)
• 6. NP → (Determiner) N (Σ) (optional)

These rules have various implications.

The first rule implies that all languages have compound sentences; these may consist of two or more simple sentences.

The second implies that all languages include the categories listed in rules 3 and 4. The K-categories are optional in any given sentence, as indicated by their inclusion in parentheses, but the verbal qualifiers are mandatorily expressed, on a binary plus-or-minus basis. This formulation results from the possibility of making sentences with no nouns, such as the Latin:

The sentence represented by this expression has no P-element but the verb, yet is designated by either + or - for the various Q-categories. Expressed as a factual statement it is +declarative; -interrogative, i.e., not a question; -negative; -middle, i.e., not formulated only with reference to the speaker; and - for the remaining categories except momentary. Such simple sentences have been found to be restricted to specific situations in speech, such as observations on the weather and on human feelings. The following utterance, though also relatively simple, may well be more typical:

While designated as - for three of the Q-categories, necessitative, voluntative, and perfective, it is + for all the others.

Since the qualifiers listed in rule 3, though subject to refinement, are universals, they must be examined for their type of expression in every language. Yet, while they are essential elements of every grammar, their order and their means of expression are unspecified in the universal rules. These are determined by the specific rules for every language; those rules lead to the wide diversity in the patterning of expression in language. Determination of the "implicational relations" and "correlations" among these features is a primary concern of typological study. The principle stated in section 1.3 was designed to represent the arrangement of the qualifiers, which in some measure determines their means of expression. As an example, this principle requires postverbal expression of a marker for the interrogative in OV languages, such as də in Sinhalese: ... däkka də? and preverbal in VSO languages, such as an in Irish: An bhfaca...? Expression for the negative is placed between the interrogative and the verb, as is also that for the other qualifiers.

Many languages remain to be explored for their expression of even these straightforward categories. But on the basis of those languages that are known, some conclusions can be formulated. The qualifiers ordered first, such as declarative and interrogative, are often expressed by intonation patterns. This observation may be illustrated with English, where the category for declarative is expressed by an intonation pattern with falling final pitch, often represented: 2 3 1 #, as in:

Moreover, the interrogative in echo questions is also expressed by an intonation pattern, represented 2 3 3 ||, as in:

Similarly, in Japanese, questions may be expressed by intonation alone, without the final particle ka. Yet the subsequent qualifiers, beginning with the negative, generally require instead of or in addition to intonation some kind of sequence of segmental material. This observation, as well as the relative placement of such segments, determined the order of the qualifier categories in rule 3

Those qualifiers which stand closest to the verb, on the other hand, notably the causative, are often merged with lexical elements. In early Proto-Indo-European the causative was expressed with an -n suffix; this was replaced in late Proto-Indo-European by an -eyo suffix. This suffix is represented with no reductions, though with vowel changes, in Sanskrit sādáyati 'causes to sit, sets' = Proto-Indo-European *sod-éyo-ti. By the time the Germanic dialects are attested, *sod-éyo-ti has been reduced to the form found in Gothic, satjiþ 'sets'. In Old English the form is seteþ, with the causative marker lost but leaving a trace in the changed vowel of the root. It has in this way merged with lexical elements, and a few relics represent different forms from the noncausative, such as set : sit, lay : lie; in general there are no surface differences, as in drop, open, etc. The category is either expressed lexically, as in kill, move, or with a phrase: cause/get/have + verb, as in:

Because of the causative's frequent merger with the verb, as well as that of the iterative and the perfective, and because of their placement with regard to it, these categories are arranged closest to the V in rule 3.

Such observations lead to questions concerning the relationship between surface expression and underlying categories. Clearly there is a close relationship in some aspects, especially in arrangement, as illustrated here. The relationship is plainest in (S)OV languages and in VSO languages. Yet the ordering given above may not be followed in some languages, and further categories may well need to be identified. In OV Quechua, for example, the following form accords only partially with the order given here:

The order of the category markers in this expression is as follows: verb +causative +reciprocal +reflexive +delimitative -momentary, followed by the third person plural suffix -nku. This form and others in Quechua provide problems, for rule 3 does not include a delimitative category; moreover, the momentary category is placed nearer the verb than is the middle, which has separate markers for reflexive and reciprocal in Quechua. The expression of categories, especially those other than the first three, cannot be expected to be ordered in the same way in every language. And the delimitative, which in Indo-European languages is expressed adverbially, may prove more generally to be a verbal category, as it is in Quechua. Such problems illustrate how meager our knowledge of language is. Our linguistic generalizations are based largely on study of the Indo-European languages, aided by consideration of those language families that have attracted some attention, such as the Afro-Asiatic, the Finno-Ugric and the Sino-Tibetan. Expectations of added insights from the study of other languages provides one of the chief reasons for intensive typological study, even if it now is carried out with principles that seem cogent on the basis of linguistic data in a small number of language families.

While the categories and universal rules must be regarded as provisional, the devices of language, especially those of syntax, seem well established; see section 1.2. These are arrangement or order, selection, modification or sandhi, and modulation or intonation. Of the four, as Sapir noted, arrangement is most fundamental. For this reason it affects the elements involved in the three other devices. Much concern with selection has been focused on determining parts of speech. While these have some validity, abstract underlying features are the primary elements of concern. The study of the other two devices, modification and intonation, has largely been confined in the past to analysis of them in individual languages. A well-designed typology demands attention to each of these devices. Used as a guide in the assembling of data as well as in their description and explanation, it will lead to more adequate grammars than those now available.

1.9. Contributions of Typology

This book throughout aims to indicate the contributions of typology to the understanding of language, chiefly through examples. Yet some preliminary evaluation will be included here, in part on the relations of a typological approach to theoretical views which claim to deal most adequately with languages.

Grammars set out to provide descriptions of languages. Like handbooks in any discipline, they are produced with varying degrees of rigor. In the past two decades much has been made of complete explicitness, to the extent that a computer might use a grammar to produce the grammatical sentences of a language and only those sentences. Such grammars have been labeled generative, resulting as Chomsky has pointed out in the tautological term "generative grammar," inasmuch as any grammar aims at accurate analysis and description. Besides being explicit, generative grammars propose to include an underlying or deep structure which reflects innate patterns controlled by the brain. The resultant grammars have been disappointing in two major respects: their parochiality and their restricted scope.

The deep structures which have been proposed are based on the analysis of SVO languages, for the most part English. Even the rules for the underlying, and presumably universal segment of the grammar, have been ordered to yield an SVO output. Other arrangements, like those of OV languages, are derived in the transformational section of the grammar. Yet this segment of the grammar is supposedly specific to individual languages and should not regulate patterns that are fundamental in a large proportion of the languages of the world. The purported deep structure in these grammars then is far removed from a universal, innate pattern.

Typological study by contrast takes into account all languages, and accordingly it leads to a deep structure which is indeed universal. Whether or not that proposed in section 1.8 is adequate, since barely a fifth of the world's languages have been considered in previous linguistic research, it has the merit of accounting for the common patterns of all known languages.

The six rules in this deep structure component have been surveyed above, as well as the probable inadequacies in the categories included. All underlying constituents of sentences can be accounted for with these rules, as a rapid review indicates. Rule 1 permits the generation of compound sentences, with conjunctions, sentence connectives, and the like; the constituents of any sentence are specified in rule 2. Verbal as well as sentence modifiers or qualifiers are generated through rule 3; adverbial as well as nominal elements through rule 4. Rule 5 provides for expression of simple noun phrases and sentence complementation. Rule 6 introduces nominal modifiers, of any pattern. Besides accounting for all basic syntactic structures, these rules yield unordered strings which can be arranged in accordance with any structure, including both VO and OV and their subtypes. The findings of typological study have led to the design of such underlying rules.

These rules are accompanied by a well-designed lexicon. The surface expressions in specific languages are then derived by means of particular rules, whatever their format and description. Transformational rules, regardless of the shortcomings of the grammars in which they have been applied, may well fulfill the requirements for explicating surface expressions. Transformational grammars have contributed especially to linguistics by sharpening the discovery procedures and formalisms used for expressing linguistic structure. Transformational grammar's rigorous procedures must be maintained as linguistics accounts for the patterns of languages. In these efforts all structures of a language must however be included. While the resulting grammar would be huge, a theory of language cannot be limited to treatment of favored constructions which may seem to be crucial for the development of a given theory. In response to the view that current theory permits only partial analyses, it may be pointed out that language has long been recognized as a "whole where everything is related — un ensemble où tout se tient"; any partial treatment will very likely then be inadequate for that segment it purports to account for, even incorrect. Such inadequacies can be avoided if crucial patterns are noted, such as those presented in section 1.4. All of these may be amplified, but if each pattern or process is identified in any language, the result will lead to proper analysis when larger grammars are undertaken.

Identification of the patterns also provides insight into the shifting structure of a language, and into possible influences of other languages. English, as noted above, is a relatively consistent SVO language, though the arrangement of adjectives before nouns suggests that its structure was at one time OV. French and Spanish, with few adjectives still placed normally before nouns, indicate that in them the shift to SVO has virtually been completed. German presents totally different problems, inasmuch as it includes two word orders in simple clauses: VO in independent clauses and OV in dependent. If we seek to dismiss either order and characterize German as VO, or as OV, we remove the possibility of understanding remarkable developments of German in the last four centuries. During this period after the reintroduction of OV order in dependent clauses, some postpositions were introduced, and also a preposed relative construction. By providing insights into such varying patterns and the resultant phenomena, recent findings of typology have added a new dimension to historical study of languages.

Chinese, like German, includes clashing patterns. Moreover, scrutiny of Chinese has led to broader conceptions of language structure than were recognized in the past, notably the identification of topic-prominent in contrast with subject-prominent languages. The recognition of this distinction by Li and Thompson has numerous implications. Among these it points up once again the secondary position of the "subject" in language, supporting the primary classification into VO and OV languages, and only secondary subclassification when the subject is included in classification for language types. Moreover, the recognition that an element with the role of topic may be found rather than a "subject" has directed attention to the functional approach to language, as exemplified in recent studies of Kuno and in the observations of Li and Thompson.

Among other characteristics of topic-prominent languages Li and Thompson have pointed out the avoidance or only marginal use of the passive construction in them. The widespread use of the passive in subject-prominent languages on the other hand they account for on the grounds that if in them a noun that is not an agent becomes a subject, the verb must "signal this 'non-normal' choice" (1976b: 467). In Chinese, for example, passivelike sentences are made with no special or signaling form of the verb, e.g.:

The second English version of (44) illustrates the requirement of a passive form in subject-prominent languages, with consequent functions that are well known and need not be discussed here. A topic-prominent language like Chinese, on the other hand, can simply present the verb, and an appropriate relationship will be assumed. By disclosing such traits, typological investigation has expanded the concerns and insights of linguistic analysis, leading to an increasingly adequate understanding of language.

Typological study accordingly has provided guidelines which assist linguists in their investigation of language, and in this way has led to increased recognition that the diverse "idioms" used in human communication are basically one, as Saussure expressed it.

The linguist works with facts and principles [in the study of Old French, for instance] similar to those that would be revealed in the description of an existing Bantu language, Attic Greek of 400 B.C. or present-day French, for that matter. These diverse descriptions would be based on similar relations; if each idiom is a closed system, all idioms embody certain fixed principles that the linguist meets again and again in passing from one to another ... The diversity of idioms hides a profound unity. (1959: 99)

The view that languages reflect an underlying unity was propounded by another great linguist, Edward Sapir, with a focus on language in change:

It would almost seem that linguistic features that are easily thinkable apart from each other, that seem to have no necessary connection in theory, have nevertheless a tendency to cluster or to follow together in the wake of some deep, controlling impulse to form that dominates their drift ... Some day, it may be, we shall be able to read from them the great underlying ground-plans. (1921: 152-153)

The studies of selected languages and problems in the following chapters have been carried out to present an account with illustrations of those ground-plans. Space is not adequate for complete grammatical sketches. The three following chapters present a selection of constructions revealing of typological structures. Chapters 5 through 7 then examine specific problems in relation to the observed typological characteristics. While no completely consistent language is known, the languages selected in Chapter 2 through 4 illustrate expected patterns in an OV language (Japanese), in a VSO language (Easter Island), and in an SVO language (English). Chapters 2 through 7, through presentation of "linguistic features [having] a tendency to cluster," illustrate "principles that the linguist meets again and again" and in this way provide insights into the phenomena of language.