Syntactic Typology: Studies in the Phenomenology of Language

Winfred P. Lehmann

8. Conclusion: Toward an Understanding of
the Profound Unity Underlying Languages

Winfred P. Lehmann

8.1. Principles Governing Language

Preceding chapters, especially Chapters 2, 3, 4, and 6, illustrated how specific principles govern language structure. Because these chapters deal primarily with consistent languages and with consistently regulated constructions the conclusions may suggest that any individual language is rigidly confined within a limited set of patterns. It is true that many characteristics of languages are so determined; yet change leads to inconsistencies in structure, as examples below illustrate. It will also become clear from these examples that change is limited to the patterns available for languages of specific structures, whether OV, or VO in the variants VSO, SVO, and VOS. A knowledge of principles governing language and their effects is then essential for analysis, description, and understanding of languages. Typological study is fundamental to linguistics, whether synchronic or diachronic, practical or theoretical.

As the treatment above has demonstrated, patterns in a specific language are controlled by linear structure in that language, and by the two forces molding its production: government and agreement.

Government regulates basic patterns of the nuclear clause, those numbered 1-4 (see section 1.4). Nominal and verbal modifiers, patterns 5-20, are introduced in keeping with these, so that they as well as compound sentences, 21-25, are determined by general principles. Through agreement, elements in these patterns may be interrelated, as well as elements involved in grammatical processes, especially 26-28. Basic patterns of any language are accordingly governed by simple but powerful principles, which affect morphological and phonological as well as syntactic characteristics.

But the regularity possible for a consistent language of any type encounters interference through a characteristic aim of communication, an aim often labeled pragmatic. In studies of pragmatic effects in SVO languages, the initial sentence constituent — the theme — has been shown generally to present old material, in this way providing textual continuity; the latter part of the sentence — the rheme — then is preferred for new material. The theme is often the logical and psychological as well as the grammatical subjects, as in:

But pragmatic aims may lead to arrangements in which elements other than such subjects are selected as theme, for example, adverbial elements. Specific grammatical processes, as well as marking, are involved in such selection, as in the following passage.

This passage could be presented in straightforward SVO sentences.

Apart from the dullness of such a presentation, its strict adherence to regular syntactic sequences obscures the emphases of the shifting situation. Immediacy of the "appeal" on the other hand is highlighted by the marking device of preposing the adverbial The moment, promoting it to the function of theme. The passive in the second clause, putting Alice in subject position, identifies her as the theme of that clause. Similarly the last clause makes she (= Alice) the theme, at the same time highlighting the difficulty caused by the simultaneous speaking of the executioner, the King, and the Queen.

SVO languages in this way include possibilities for expressions which contrast with straightforward sequences in their structure. So also do other language types. Special devices introduced in continuous texts are studied especially in discourse analysis. These devices, or pragmatic concerns, may lead to diverse patterning.

Diverse patterns are also found in languages as a result of linguistic change. When change occurs in language, it is handled so as to interfere minimally with communication. Accordingly, small segments of a language are affected, and the process of change is carried out over a long period of time. For example, voicing of t in some areas of American English is confined to those members of the t-phoneme which occur medially, as in Minnesóta, and which stand before an unstressed vowel — contrast detér — but not if the unstressed syllable ends in n — contrast bútton with bútter. The phonological change has been going on for at least a century and is continuing. Syntactic, morphological, and semantic changes may require even longer periods to be carried out. For example, regularization of English verbs, e.g., dive : dived vs. dive : dove, has been going on more than a millennium, as has the shift from OV to VO structures.

Facts of change are occasionally misunderstood when one's own speech is taken as norm. For any individual a specific change may well be immediate; change is adopted at a certain point, and the old pattern is abandoned. For a speech community, however, old patterns persist among many speakers, so that the change is adopted only slowly by the entire community, as illustrated with the examples cited here. The most comprehensive study of change in a community for an extended period of time has been carried out at Charmey, Switzerland. The study provides excellent examples of the position presented here concerning change (see Lehmann 1973b: 163-164).

Given this situation regarding change, no language is completely regular or symmetrical, whether in phonological structure or in the structure of any other component. Yet the lack of symmetry does not nullify the general principles discussed above. These instead are supported by observation of languages over long periods of time, when, as Sapir recognized, the changes may be viewed in accordance with underlying ground-plans, as a drift.

It is useful, however, to distinguish between languages which are in process of change and those which are relatively stable, like Japanese. Since typological conclusions are based on empirical data, generalizations presented above have been ascertained through observation of consistent languages.

When a language is at a stage in change that has affected only some patterns, so that some constructions are VO, others OV, it is said to be inconsistent. The characteristics of inconsistent languages can be best determined for languages with long documentation, such as Indo-European, Afro-Asiatic, and Sino-Tibetan. On the basis of such demonstration we can then analyze languages with only brief periods of documentation, such as the Amerindian languages and the languages of Australia and New Guinea. Positing of general principles for inconsistent languages is now an important goal of typological study, and indeed of general linguistic study.

It is also instructive to examine means of expression that are attested as opposed to theoretically possible means of expression. For expressions widely found in human language as opposed to potential though undocumented patterns lead to important observations for the understanding of language. Widely recorded and undoubtedly universal patterns, like the interrogative and negative, might well be expressed by means of a variety of devices, phonological, syntactic, or semantic. The phonological device of nasalization (indicated by ˜) could, for example, be employed in questions as opposed to statements.

Many other such phonological devices — palatalization, pharyngealization, and so on — could be applied to distinguish questions from statements or to express other verbal quantifiers.

Semantic devices also could be utilized, as by use of labels to characterize sentence types, for example:

While this suggestion too may seem odd, Old Irish actually includes such a mechanism, though it may be orthographic; the prefatory word ceist [keⁱšt] 'question' is found before some interrogative clauses.

More such instances of semantic or phonological devices may be uncovered in time; but typically the expressions for questions, and for other verbal qualifiers, are syntactic, whether of arrangement, selection, or intonation. Expression of such modifications of meaning through syntactic devices is a further indication of the central position of syntax in language, and consequently also its central position in typological study. Moreover, even when modifications are brought about by pragmatic and functional aims, or as a result of language change, the principles regulating syntactic structure are remarkably simple and general among languages of different structures. Such properties of language account for the rapid acquisition of language by children, for pidginization, and for the readiness with which languages are translated.

8.2. Inconsistency in Languages

While syntactic structures of languages are governed by pervasive principles, few languages are even approximately consistent. Japanese itself, selected here because of its relative consistency, underwent strong VO influence from Chinese, which is evident in many of its learned compounds and in its reflexive. Its history illustrates how inconsistent patterns can be introduced through diffusion or borrowing. Inconsistencies are also found because some patterns change more readily than others. For fuller understanding of linguistic structures specific inconsistencies must be studied, as well as their reasons and their effects in a given language.

Detailed illustrations will not be provided here, in part because any such study involves examination of the interrelations between many constructions in language over a long period of time. A brief illustration may however be given in the Turkish ki construction. Borrowed from Persian within the last millennium, it affords in Turkish a VO relative clause pattern. The tie between ki clauses and their matrix is loose, permitting a large number of interrelationships, as any grammar of Turkish will illustrate. Since ki clauses are used especially in technical and literary texts, most sentences including them are lengthy; two simple examples are given here, the first a standard relative clause, the second causal in its effect (from Jansky 1954: 233-234):

As these examples illustrate, use of postponed (VO) clauses with ki provides a flexible alternative in Turkish to the preposed relative clauses required in an OV language.

Yet ki clauses have fallen into disuse, especially since attention was turned to "purifying" the language following the Atatürk revolution. Both their current rejection, and their introduction from Persian would repay careful study. One may speculate, for example, that they were readily adopted because of the earlier presence in Turkish of a native particle ki indicating attributive relationship. This ki is found already in Old Turkish, as in temporal and locative expressions: söki 'former', yirdäki 'who is found on earth' (von Gabain 1974: 64-65). It is also maintained in Modern Turkish:

Just as we assume that the Persian ki was favorably regarded by Turkish speakers because of their somewhat similar native pattern, we might also propose that foreign patterns are more readily adopted when comparable constructions are already found in the language. These may well be chance similarities, like ki in Turkish; or they could be found in residues due to incomplete change, or in patterns due to marking. Investigation of further inconsistencies in specific languages, their possible reasons, and their effects on the language will contribute greatly to our understanding of change in language. The course of their possible rejection is also illuminating, not only the cause, such as the aim to purify a language and its effects, but also the consequences for alternate patterns that fill the former role of the rejected pattern.

Inconsistency may also lead to special processes in language. An example is that vowel modification known as umlaut in the early Germanic dialects. Umlaut was noted in Chapter 1 as a potential process in VO languages, for it is a kind of progressive assimilation. Vowels in earlier syllables of words, characteristically in the accented syllable, are modified in accordance with final syllables. Thus -gastiR, attested in the Gallehus inscription ca. 350 A.D., illustrates an earlier stage of the form found in Old English giest, NE guest, where e is an umlauted form of a. In Germanic, back vowels of stem syllables were modified by high front elements [i j], and front vowels by high back elements [u w]. Umlaut was carried through with great regularity, especially in northern Germanic dialects, such as Old Norse and Old English.

When a sound change is carried out so consistently, similarity of phonological environment is generally a factor. The consistency of umlaut can be ascribed to the preponderance of open syllables in initial portions of Germanic words, as in the complete Gallehus name: Hle-wa-ga-stiR. Such syllabic structure is expected in OV languages, as exemplified in the Japanese examples of Chapter 2.

These observations lead to the conclusion that umlaut might be expected to occur in an inconsistent language which is moving from OV to VO structure. Germanic umlaut illustrates some of the subtleties involved in changes from one structure to another. Other such phenomena will be noted below, though many further investigations of specific languages must be undertaken, to better explicate processes like umlaut as well as to understand specific languages at any stage in their development. It should be observed these proposals do not imply that shift in type causes phenomena like umlaut; on the other hand, these shifts are instrumental in setting the stage for such phenomena. Investigations of well-attested changes like Germanic umlaut provide further evidence on processes of change in language, on conditions under which they occur and their possible causes. Such information can be used to account for patterns of languages known only from the present.

Nez Perce provides an intriguing example. In it the vowels in words are regulated morphologically, neither progressively as in umlaut nor regressively, as in vowel harmony; because the vowels "harmonize," however, the term vowel harmony is used of Nez Perce. The mechanism is morphological rather than phonological, as in Germanic umlaut or in Turkish vowel harmony; for example, a specific morphological marker accounts for the specific vowels of /caqá‧t'ayn/ 'for a raspberry' opposed to /cé‧qet/ 'raspberry' (Aoki 1966: 760-761). A similar situation is found in Chukchee (Comrie, personal communication). These languages reflect results of inconsistencies, whether through internal changes or through borrowing. This assumption, or any other regarding the Nez Perce and Chukchee phenomena, is highly tentative because of our ignorance about their earlier history. Investigation of languages with known histories may in time furnish information through which we can account for phenomena in languages like Nez Perce having texts only from the recent past and present. Besides characterizing consistent languages, as was done in Chapters 2, 3 and 4, further studies should set out to provide characterizations of languages with various kinds of inconsistencies.

8.3. Change in Typological Patterning

Some languages attested for considerable time permit observation of effects of change in typological structure on individual patterns. Sinhalese, for example, is relatively well documented, at least in that its basic patterning is disclosed around the beginning of our era as well as today. Two millennia ago Sinhalese included many VO characteristics. In this respect it is comparable to Classical Sanskrit and Middle Indic dialects like Pali. Subsequently Sinhalese has become a highly consistent OV language, without doubt through Dravidian influence.

In the course of this change to OV structure all patterns noted above have been modified. Even such relatively conservative patterns as the teen numerals have for the most part adopted Dravidian order (Ratanajoti 1975: 51, 80). The numerals 11, 12, 15, which have not, are of further interest, as noted below; 13, 14, 18 are given to illustrate the regular Sinhalese pattern — others may be noted in Ratanajoti's lists.

The New Sinhalese patterns in 11 and 12, reminiscent of the forms not changed to VO order in English and German — eleven, Gothic ainlif 'one left'; twelve, Gothic twalif 'two left' — may have escaped change because the basic numerals are often viewed as a set of twelve rather than ten. The numeral for 15 in New Sinhalese on the other hand would have countered a proscribed phonological pattern if changed to OV order; a phonological constraint thus countered its syntactic change (Ratajanoti 1976). Yet the extensive changes between Old Sinhalese and New Sinhalese, even to the relatively inconsequential teen numerals, illustrates the cohesiveness of the patterning of a language.

This cohesiveness may be demonstrated by examining change in the languages of a family, such as the Indo-European, over several millennia. Indo-European is the family best known to us at present, because texts in the various languages have been most thoroughly studied, and because these texts extend through a period of almost four thousand years; eventually the Afro-Asiatic and the Sino-Tibetan families may be known as well, to the great advancement of linguistics. Even this extent of time does not take us back to the parent language, Proto-Indo-European, which, occupying the same relation to the attested Indo-European languages as does Latin to the Romance languages, must be posited no later than 3000 B.C. Although the languages of the Indo-European subgroups have been intensively studied for a century and a half, with the aim of reconstructing the parent language, many problems remain. Some of these have now been clarified with the help of insights derived from typological studies. The following are examples.

The early dialects differ in their relative clause markers, some like Sanskrit and Greek deriving them from *yo-, others like Hittite and Latin from *k^wi-. Moreover, the earliest texts, especially in Hittite and Vedic Sanskrit, have a majority of relative constructions preceding their nouns, as in the following example from the Hittite Laws (Justus 1976: 234):

Indo-Europeanists noted these facts, but were unable to account for them; they selected one of the alternates, typically *yo-, and posited it as the Proto-Indo-European relative marker.

The early dialects also have postpositions rather than prepositions, a situation which also was noted but not accounted for.

Moreover, no reflexive pronoun could be reconstructed for Proto-Indo-European. Even languages as recently attested as the Germanic differ in these, as English myself, herself, and so on compared with German mich, sich illustrate. On the other hand, the early dialects contained a "reflexive verb," the so-called middle. This gradually was lost, while constructions including a reflexive pronoun replaced it, so that Greek middle expressions like the following disappeared at some time after the beginning of our era.

Both the loss of the middle and the development of reflexive pronouns are accounted for by comparing OV languages. In these, verb forms indicating reflexivization with a suffix like that of the Indo-European middle may be expected, and also the absence of reflexive pronouns. A consideration of the devices to indicate reflexivization in the Indo-European languages, as well as the relative construction, postpositions, and many other characteristics then leads to the conclusion that the early Indo-European languages were developing from OV structure to VO. Proto-Indo-European itself then may well have been a consistent OV language.

These conclusions clarify many of the changing constructions. If Proto-Indo-European had been OV, a relative pronoun would not have been necessary, for relativization could have been indicated by preposing of relative clauses as in Japanese. But when the dialects were changing to VO structure, such a pronoun was essential, it was developed from a topicalizer in Hittite, as Justus has convincingly shown (1976: 213-245). Moreover, the postpositions of the early dialects would be explained, as would many other constructions. Among these are survivals of OV comparatives, as in early Old English poetry, such as the following line from a riddle (Williamson 1977: 92, Riddle 38.18).

Enumeration of all the OV characteristics in the early dialects would require too much space in a general account, yet the typical suffixation and absence of prefixation, the large system of cases, the preposing of genitives and adjective modifiers might still be mentioned as well as the common OV patterning of clauses in our earliest materials (for more discussion see Lehmann 1974a). All of these constructions changed, as the structure of the languages was modified from OV to VO. The process and the time involved in the change vary among the constructions. Problems having to do with the interrelationships of such changes and the extent of time involved in carrying them out require thorough study like that devoted to the change of the English genitive by Fries (1940; see also Traugott 1972). Eventually consistent VO languages have emerged, such as French, Spanish, and English.

The recognition of a drift toward consistency illuminates many further constructions, such as complements in their gradual development in the Germanic languages (Lehmann 1976b; see also Ureland 1973). Early Germanic languages show a variety of complementation patterns, including ACI (accusative with infinitive): I expect him to go, and that-clauses: I expect that he will go. There are also relics of preposed complementation types to be expected in OV structure. The patterning of Germanic dialects today developed only gradually; for example, verbs of perception regularly used that-constructions in the early period, and only later ACI as in I never knew him to be late. The history of complementation in other Indo-European dialects as well is clarified when one recognizes that the patterns in each of them were replacements for OV patterns, similar to those in Japanese. Even the that-complementizer of Modern English still reflects its origin as a nominalizer, as in:

Compare this with:

Similar observations can be made for early Greek; the typical Classical Greek complementizer hóti originally — like NE that — was a demonstrative pronoun which developed into a conjunction.

Study of the syntactic changes in the Indo-European languages then indicates that when languages change, patterns of the new type are introduced and gradually become established, while the language becomes increasingly consistent. The direction of change may encounter specific interferences, as in the Indian linguistic area, illustrated above. In the Indo-European family as a whole, especially in the Western languages, the change from OV to VO represents an internal drift. In the languages of India during the past two millennia, on the other hand, syntactic change represents the results of external influence. Yet in both groups consistency eventually emerges. Syntactic change, whether in accordance with internally directed drift or external influences, can be accounted for through use of the typological framework presented here.

8.4. German, an Example of an Inconsistent Language

A remarkable example of conflicting patterns is provided by German. It illustrates how social forces may interfere with development toward consistency. The simple clause pattern in German requires second position for the verb, so that clauses generally follow SVO order, though OVS is readily possible. By contrast, subordinate clauses require verb-final position, so they are SOV. It is illuminating to note the time and reason for introduction of the conflicting SOV pattern as well as its effects.

The OV pattern of subordinate clauses was adopted as the regular construction by learned writers around the beginning of the sixteenth century on the basis of Latin, though it was not wholly absent in earlier forms of German. Gradually the twofold patterning of VO order in independent clauses, OV in dependent, was installed, so that it is now regular in the standard written language. But it is not required in all dialects, nor in all spoken forms of the language.

Since adoption of OV subordinate clause order, various additional OV patterns have been introduced into standard German. Among these are postpositions. An example is wegen 'because', an old dative plural of Weg 'way', arising from expressions like von ... Wegen 'of ... ways' and first attested as a postposition in the seventeenth century. Another is entlang, comparable to New English along, first attested as a postposition in 1741.

A further OV characteristic is the use of preposed relative constructions, notorious in scientific German writing, as in the following sentence from Weizsäcker:

Like postpositions, this construction is widely found only after introduction of OV order in subordinate clauses.

Other OV characteristics of Modern German can also, if less surely, be associated with the newly observed order of subordinate clauses. One is the postverbal negative.

In Middle High German a preposed verbal element was used to indicate negation, in the VO pattern:

Another is use of interrogatives medially:

Placing of interrogative in such positions is reminiscent of Chinese, which also has OV characteristics, and of OV languages like Japanese. OV patterning may also be evident in constraints against ACI complements; one cannot translate literally into German the example cited above: I never knew him to be late.

In addition to studies of such patterns, investigations should be undertaken to examine patterning of standard written German in contrast with the spoken language and the dialects. Some studies have determined that spoken language has more notable VO structure than written language, as in the treatment of postpositions of the written language as prepositions. Further, dialects preserve the Middle High German negative with a twofold marker, as in the medieval Song of Roland, line 8918:

But investigations are still far too limited, generally dealing with only one area or one dialect, so that much remains to be learned about the impact of the OV change on standard written German.

These studies have broad theoretical implications. Among the points to explore is existence of dual patterns in a language. Many scholars examining German have declared it to be OV in underlying structure, while others see it as VO, with views varying in accordance with the criteria chosen as decisive. If, for example, one assumes OV structure, the independent clause pattern can be simply generated with a transformational model. Yet a criterion frequently used to classify languages, gapping of clauses containing objects and verbs, is carried out in German as in English, leading to the conclusion that German is VO:

A decision must be reached by identifying the primary constructions in the language, such as the principal clause, especially if its type of government is prominently found in adpositions and comparatives.

A further problem suggested by the German constructions has to do with the importance of arrangement as a syntactic device, notably with reference to syntactic change. To judge by syntactic modifications following introduction of OV order in subordinate clauses, arrangement is the central syntactic device, in some sense determining all others.

German is also instructive for investigating relationships between surface structures and underlying forms. Its recent history supports the suggestion that surface arrangements, especially of clauses, closely reflects underlying form, even though patterns of selection may not.

All such questions need further investigation, and comparison with patterning in other languages. It would be highly interesting, for example, to determine why Japanese adopted a VO pattern for reflexivization alone but maintained all the other OV patterns. Similarly, study of Chinese dialects should illuminate our understanding of typological patterning, for southern dialects are more consistently VO than is the northern standard language, Putonghua. Aryan languages of India with their change toward OV order in the past two millennia will also provide insights into patterns modified under the influence of other languages. Such investigations will, moreover, provide data for statements of greater assurance about languages in change, supplementing observations made for Sinhalese and other languages so studied. Resulting conclusions will permit interpretation of many languages for which we have no lengthy series of materials.

8.5. Problems Resulting from Influence of Languages on Other Languages

Especially since Sandfeld's influential publication on Balkan languages (1930), attention has been given to characteristics of adjoining languages which may have been introduced by diffusion. While impressive for the extent of information about Greek, Albanian, Bulgarian, Rumanian, Serbo-Croatian, and Turkish of the present and past, their dialects, and further languages like Romany, Sandfeld's book illustrates that a great amount of further investigation needs to be carried out on areal influences. For even in the Balkan area, which has been studied more than a century and a half, specific syntactic characteristics, as well as lexical items, have been identified as due to diffusion, but not examined for their cause of adoption nor their possible further implications for individual languages.

For example, one such characteristic — the postposed article in Albanian, Bulgarian, and Rumanian — has been ascribed to a substratum, though not with general agreement (Sandfeld 1930: 165-172).

Other characteristics, such as the replacement of infinitives as verbal complements by subordinate clauses and the formation of the future with auxiliary 'will', are ascribed to the influence of Greek (Sandfeld 1930: 173-185); but their origin in Greek has not been accounted for with general assent. Found already in New Testament Greek, these constructions have been ascribed to Semitic influence. It is now clear, however, that use of auxiliaries like 'will' and preference for subordinate clauses are VO characteristics. Accordingly, use of an auxiliary to indicate future might have been expected for internal reasons in Greek, which had long been developing toward VO structure; even Homeric Greek includes many VO characteristics. Adoption and use of such patterns in other Balkan languages would be important to investigate, as well as any additional syntactic modifications following their adoption. For, apart from Turkish, these languages are VO; the adoption of specific VO patterns from Greek would not then have such dramatic consequences for their structures as were noted for Sinhalese or even for German. Nonetheless, study of the Balkan linguistic area should illuminate diffusion of syntactic characteristics in specific areas, in part because many studies have already been carried out on its languages.

Effects of languages in given areas on others have been observed, though reasons for the predominance of one language type over another or for the adoption of one typological pattern rather than another elude us. We can only speculate on the predominance of the Dravidian OV pattern in the Indian linguistic area: speakers of Dravidian languages most likely outnumbered speakers of Indo-European and Munda languages. Further, Indo-European languages had not yet eliminated all OV characteristics when their speakers came into contact with Dravidian speakers; while Classical Sanskrit has fewer OV patterns than does Vedic Sanskrit, its structure is comparable to those of Classical Greek and Latin, which also have OV relics. Presence of maintained OV patterns may then have eased adoption of Dravidian constructions.

Yet all linguistic areas should be examined, for the questions involved have fascinated linguists as well as other scholars, with no resolution of their views. One linguistic area of wide concern is Western Europe. Its languages today are so similar in structure that they seemed one in type to Whorf, who called them SAE, Standard Average European (1956: 138). Origins of similar constructions in these languages, such as compound tenses, are however disputed, as may be illustrated by the differing viewpoints of the great linguists Antoine Meillet and his student Emile Benveniste concerning origin of the "perfect with 'to have' in Germanic" (Benveniste 1971: 178-179). Meillet ascribed it to "imitation of Latin models." Benveniste on the other hand saw the solution of the problem not in "historical grammar" but rather in "consideration of the system." On this basis he concluded that "acquisition of a transitive perfect with 'to have' was an autonomous development in Germanic and owes nothing to the influence of Latin." For Benveniste nonautonomous development would have been possible only after long Germano-Latin bilingualism like the "Slavo-Turkish bilingualism that circumstances imposed in Macedonia for five centuries" (1971: 179). Benveniste's argumentation is persuasive, based as it is on "consideration of the system," that is, in accordance with the aims discussed above of examining and accounting for linguistic developments in terms of a typological framework.

Use of a framework is especially important when inter-relationships between languages of a given area are poorly known, or known only for very recent time-periods. Such areas provide intriguing problems, as noted for Amerindian languages by Franz Boas and Edward Sapir (Boas 1911: 36-37 et passim; Sapir 1963: 205). Although information on these languages is far less extensive than on long studied and well-documented languages, it may lead to insights into the origin of linguistic characteristics, such as use of person markers on verbs. According to Ingram, "in a structurally balanced system, a person marker will occur on the same side of a verb as the NP to which it refers" (Ingram 1975; see also Lehmann 1975b: 53). Representing person markers with lower-case s = subject and o = object, one would expect an SOV language with such markers to have the patterns SO (o-)s-V, a VSO language the pattern V-s(-o) SO/V-o-s SO. Ingram's principle is supported by forms like the Semitic perfective, for example Classical Hebrew, a VSO language:

If such a pattern, which had developed in a "structurally balanced system," came under the influence of an SOV pattern, these verb forms might be maintained, even though they no longer agreed with the new pattern. Such aberrant forms might then provide perspective on the language concerned, as well as on dominant areal relationships. Ingram proposed such a conclusion for some Salish languages. Parts of his argumentation have been disputed by Noonan, who points out the necessity of taking clitics into consideration, and thus the need of "great care" in "use of affixal morphology as a tool for reconstructing earlier word order patterns" (Noonan 1976). The importance of distinguishing between pronominal affixes as concord markers and as person markers has also been indicated (Lehmann 1975b). To the present far too little is known about person markers to propose hypotheses on their occurrence and origin. Besides Amerindian languages, investigations are also being carried out on African languages (Givón 1971; 1975).

In spite of the lack of firm conclusions, person markers have been discussed at some length here to illustrate how further insights may be obtained from study of such characteristics, both for individual languages and for linguistic areas. As further information is assembled, it may assist in illuminating interrelationships among past linguistic areas, such as the proposed Indo-European - Kartvelian area, or the Mesopotamian area of the fourth and third millennia B.C. Recent treatments of person markers are also instructive in illustrating the usefulness of a typological framework in disclosing further linguistic characteristics which may lead to improved understanding of languages and their history.

8.6. Identification of Further Typological Characteristics

Through scrutiny of patterns in accordance with a typological framework further characteristics of typological significance will be identified and incorporated in the framework. An additional pattern of interest is that of sentence-introducing particles. Such particles might be expected for VSO languages, since they are characterized by prefixation. Yet these particles attracted attention in the study of Indo-European languages, in which they have been treated quite independently of typological structure. The attention resulted from the observation that Irish and Hittite exhibit comparable particles. Irish, a VSO language, might well be assumed to include them, but Hittite is OV. Accordingly the languages in which they have been most thoroughly studied may seem to indicate that any type of language might have sentence-introductory particles as an indigenous characteristic.

Hittite and other Anatolian languages have come to be much better known since 1942, when Albrecht Goetze and Myles Dillon made their initial observations about the sentence-introductory particles (Dillon 1947). Of signal importance in further information about Hittite is recognition of a distinction between early texts and late texts, and consequent assumption of two stages in the history of the language: Old Hittite and Late Hittite. It is noteworthy that the particles are much more prominent in Late than in Old Hittite. On the assumption that such particles would be expected in VSO languages, we may ascribe their widespread use in Anatolian languages to the influence of a VSO language; Akkadian, widely used by the Hittite scribes, is the most likely source of that syntactic influence. The particles can therefore no longer be posited as a syntactic characteristic of Proto-Indo-European, even though the morphological elements can be related to particles like Sanskrit nu, Greek nûn, English now.

A further complication, however, must be noted in dealing with Anatolian languages. Akkadian — in contrast with other Semitic and Afro-Asiatic languages in general, which are strongly VSO — has many instances of verb-final sentences. This characteristic may be ascribed to influence from Sumerian, which while OV in many characteristics exhibits VO features as well.

In short, the materials which have been used to understand the functioning of sentence-introductory particles are themselves in need of elucidation. Data for such elucidation may now be available in the finds at Ebla in Northern Syria; but they have not yet been made available. Until new data on the early period of Semitic languages illuminate their use of sentence connectives, or until material is provided from other languages, we must reserve judgment on their use.

Other features are similarly in need of investigation. Verbal qualifiers, as one set of examples, require considerable clarification. These are categories which have traditionally been treated in analyses of verbal systems, generally in relation to morphological markers. They are also studied by logicians, as in modal logic. In recent linguistic publications the nomenclature of logicians has often been preferred to that of traditional linguistics. Thus verbal qualifiers are labeled CAUSE rather than Causative, CHANGE rather than Inchoative, and the like. Moreover, many studies have indicated modal features by representing them with "higher sentences": He causes something. The child lies down. > He lays the child down. Further, there seem to be few bounds on the verbal features that are proposed in some studies. By contrast, with their basis in empirical observations, typological studies give precedence to verbal qualifiers that are widespread, if not universal, such as Interrogative, Negative, and others treated in patterns listed in section 1.3. As further languages are investigated these may need elaboration and closer definition, though such studies should be carried out with reference to a framework.

Only Indo-European languages have been thoroughly analyzed for Q-features. Calbert, for example, has dealt in depth with "modality," particularly in German. His conclusion illustrates some of the difficulties involved in treating verbal features, for he states that "modality is ... a matter of degree ... Any act of saying (or any sentence) conveys a certain degree of modality which goes from zero (or near zero) in neutral reporting to the extremes of volition (will, command) and absolute certainty" (Calbert and Vater 1975: 55). One could thus scarcely expect systems using only binary characterization to represent fully nuances of utterances. Further, languages may structure modal categories differently from the patterning found in Indo-European languages. As noted above in section 1.8, Q-features proposed here may be expanded when additional typological investigations are carried out, especially when these deal with languages differing in structure from Indo-European.

Nominal constructions will also be explicated in further study. Patterning in compound numerals, for example, has been determined by Greenberg (1976). And the loose alignments of apposition may in time be comprehended within a specific structure (see Hauri 1976).

Besides constructions tied to central clause structure, there are some that seem to be free of such relationship, such as adverbial elements referring to time and place. In English, adverbials referring to place precede those referring to time, while in German the reverse order applies.

Study of such free patterns is of great interest, for they afford individuality to languages — in contrast with the constructions regulated by types.

Typological investigations must be amplified, to examine not only patterns discussed above in many further languages, but also constructions and patterns which seem independent of the central structures of language.

8.7. The Problem of Apparently Similar Patterns and Constructions in Different Languages

In the study of a wide range of languages problems result from contrasts between apparently similar constructions. The constructions labeled "passive" may serve as illustration. Even in two closely related languages like German and English the "passive" is not equivalent. German may use passive to topicalize verbs as well as nouns, even intransitive verbs:

Although English uses passive only to topicalize nouns, differences between the forms in the two languages are relatively minor, for in both, passive functions to indicate topicalization.

By contrast, in Japanese and other languages of Southeast Asia "passive" has totally different uses. It conveys the connotation of an unfortunate occurrence, in the so-called adversity passive (see section 2.3.7).

It may also indicate potentiality, and then may carry honorific meaning.

Besides these uses; the Japanese passive is also found without such connotations, with an element other than the agent as subject; this "normal" use of the passive has long been noted, especially in embedded clauses, as Kuno has pointed out.

Comparison of the "passive" in languages without informed understanding of its special uses would scarcely be illuminating. The difficulty is not simply brought about by application of specific labels, for in Japanese and other Southeast Asian languages the "normal" passive is comparable to the passive of English and German. But each of these languages further delimits the passive, associating additional characteristics with it, in this way leading to uses which differ considerably from language to language.

Characteristic delimitation or extension of widely found patterns is by no means limited to languages cited here, nor to the "passive." While the "passive" may provide a striking example, all patterns must be examined with reference to the language in which they are found, whether they seem clearly identified, like causatives, or only vaguely, like subjunctives. The causative in Arabic, for example, is associated with further modalities, such as volition. Modal categories especially have been discussed at length by grammarians dealing with intensively studied languages (Wackernagel 1926: 210-257). So have aspects and other categories, which have been of especial interest to many scholars because of shifting means of expression, as when the Indo-European inflectional paradigms indicating tense and aspect are replaced by compound expressions, and modal forms — subjunctive and optative — are replaced by verbal phrases with auxiliaries and by particles in various dialects. General linguistics will profit greatly from such studies, especially if data are analyzed in a more specific framework than that used in past studies — a framework based on subsequent typological study.

8.8. Relations between Language Types and Other Cultural Systems: Linguistic Relativity

The study of language has always been pursued to increase our understanding of people and of cultural systems. Many linguists concerned with typology have also maintained strong interest in this aim. For example, Finck prefaced his eight published lectures on "the German language as an expression of the German world-view" with the following quotation from Wilhelm von Humboldt (Finck 1899: my translation): "The characteristic intellectual features and the linguistic structure of a people stand in such intimacy of fusion with each other that if the one were presented the other would have to be completely derivable from it." Finck's lectures then proceed to illustrate the "fusion" for German. While one must respect Finck's caution in proposing broad generalizations as well as his extensive knowledge of many languages, his conclusions have not been well received, in part because they were stated in terms of medieval psychology, and of linguistics concerned with surface structures.

Comparable interest in the relations between a language and the world-view of its speakers was manifested by Edward Sapir and his contemporary Benjamin Whorf. Their work has been so strongly identified with such a conception that the proposed relationship between language and culture is often referred to as the Sapir-Whorf hypothesis, alternatively as linguistic relativity. Although the writings of Sapir and Whorf aroused considerable interest, leading to attempts to test the hypothesis in carefully designed experiments, linguistic relativity is widely regarded with skepticism. The esteem accorded it by typological linguists may also have diminished regard for this approach to language.

Yet further advances in typology and in linguistics may well provide guidelines for advantageous attention to linguistic relativity. For example, patterns to be investigated must not be taken from deep structure, nor even from those regulated by the principle stated above. Since deep structure patterns are universal, they apply to all humanity, and thus do not reflect possible specific influences of a given language on the thinking and the world-view of its speakers. Moreover, inferences drawn from such patterns must be more than speculation.

An example of a particular pattern may be supplied from Finck's concern with German. This pattern is the use of endingless adjectives for adverbs as well as predicate adjectives (1899: 53).

One may accept Finck's conclusion that the pattern tends to give the copula a fuller meaning than it has in English and the other Germanic languages because it is accompanied by the same form of the adjective as are other verbs. With greater reluctance one might also honor his interpretation that German in this way tends not to ascribe permanent characteristics to objects, but to view them as active and changing — a characteristic which he sees also in the maintenance of grammatical gender in German. But the further interpretation relating these linguistic patterns with the "strength of will and intellect" of their speakers strikes one as an imaginative flight.

In the same way, widely discussed proposals of Whorf relating cultural with linguistic characteristics in Amerindian languages, and their implications for English and SAE, have been regarded with skepticism; see especially Whorf 1956: 134-159. English has indeed a tense system in the verb, and SAE languages objectify time, speaking of "three hours" in the same way as of "three chairs." Moreover, western culture ascribes a great deal of weight to:

1. Records, diaries, bookkeeping, accounting, mathematics stimulated by accounting.
2. Interest in exact sequence, dating, calendars, chronology, clocks, time wages, time graphs, time as used in physics.
3. Annals, histories, the historical attitude, interest in the past, archeology, attitudes of introjection towards past periods, e.g., classicism, romanticism. (Whorf 1956: 153)

Yet Whorf's proposed "relation of habitual thought and behavior to language" scarcely rises above the level of speculation, even if poetic and intriguing. For when apparently comparable situations or comparable linguistic structures are examined, the relationships may not be the same as in the SAE area. Chinese culture, for example, exhibits similar attention to "records ... chronology ... histories," but Chinese does not have a tense system — though it does objectify time. Turkish on the other hand has a comprehensive tense system, but the accompanying culture does not assign the same importance to "records" and the like. Intuitively we may sympathize with Sapir's statement which Whorf cited as a foreword to this article "that the 'real world' is to a large extent unself-consciously built up on [our] language habits" (Whorf 1956: 134). At the current stage of our knowledge of language and of its role in society, however, we cannot declare such statements confirmed. Their significance for understanding ourselves, our culture, and other cultures, however, adds additional weight to Whorf's subsequently stated plea: "We must find out more about language! Already we know enough about it to know it is not what the great majority of men, lay or scientific, think it is" (1956: 250).

So much of linguistic activity consists in the attempt to "assemble and arrange ... the whole body of linguistic phenomena" that only a few linguists pursue Whitney's third goal of "explaining" them, let alone relating language systematically to other cultural matters (Whitney 1892: 6). In this situation linguistic relativity must be viewed as an intriguing hypothesis, to be confirmed, whether in part or as a whole, or to be rejected or to be refined by scholars of the future (Penn 1972).

8.9. Approaches to Language

Linguists assume as the goals of their discipline the three tasks indicated by Whitney: assembling, arranging (i.e., describing), and explaining the phenomena of language. They carry out this aim through the production of grammars accompanied by dictionaries. It is widely agreed that a grammar is a theory of a language. A grammar should then have the attributes of a theory, including predictive capability.

In seeking to meet these goals various approaches to the study of language are proposed and maintained, often referred to as "different 'theories'." But, as Dixon has stated recently: "The term 'theory' is in fact, inappropriately used in modern linguistics; many so-called 'theories' have no predictive power and are little more than systems of notation, 'general plans' in terms of which the grammar of a language may be formulated" (Dixon 1976: 75). While even the prominent "theories" or "general plans" cannot be reviewed here without undue expansion of this chapter, the contributions of typological study to the aims of linguistics may be briefly noted with reference to widely used approaches in the formal study of language.

Typological study, like much other study of language, is based on the view that linguistics is an empirical science. Theory is formulated in accordance with the data of languages, and tested by means of such data. In carrying out investigations, observation and description are certainly directed by one's theory, which, whether transformational or other, provides "discovery procedures." But the strength of a theory lies in explanations it affords, or in understanding it contributes.

In the nineteenth century, explanation was sought in determining earlier forms. If a question was raised about an element in the sound system, like the /r/ in were as opposed to the /z/ in was, or about a specific form, like the past tense sang as opposed to the present tense sing, attempts were made to determine the situation in earlier stages of the language and the conditions under which the concerned elements arose. If these questions were solved, linguists assumed that the elements were explained. Such "explanations" are indeed satisfying, but they have their limitations. Even for well-documented languages like English and its congeners, explanations can be secured only for elements and events subsequent to the so-called parent language, Proto-Indo-European. Moreover, elements and events of languages with little documentation, like those of America or Africa, can be given only meager explanation with the historical approach. Further, linguists found it impossible to explain the facts of language as a phenomenon through its history; occasional intrepid scholars indeed proposed deriving all the languages of a continent from one, or all the languages of the world from one, but the resulting views are very vague, and thus are regarded largely as curiosities (e.g., Swadesh 1971). The explanatory power of historical linguistics thus has severe limitations.

Moreover, predictive possibilities of historical linguistics are restricted to proposing hypothetical forms a given item might have developed to in a certain language. For example, if the cognate of Latin equus and Greek híppos, Old English eoh 'horse', had been maintained in Modern English it would presumably be ee, cf. OE feoh > NE fee, OE trēo > NE tree. Further predictions could scarcely be ventured, as of the loss of eoh. For similar monosyllables were maintained, such as /ay/ eye, /ow/ owe, /ɔ/ awe. Possibilities of prediction in historical linguistics, like its explanatory capabilities, then provide only meager results among general expectations for a theory.

The most prominent recent approach, transformational grammar, seeks its explanations for the phenomena of language in treatment of linguistic structures by the mind. In the words of Chomsky: "The theory would suggest an explanation for the linguistic intuition of native speakers as regards John is easy to please and John is eager to to please. This explanation would rest on the assumption that the concepts of grammatical structure and 'significant generalization' made explicit in this theory constitute the set of tools used by the learner in constructing an internal representation of his language (i.e., a generative grammar), on the basis of presented linguistic data" (Chomsky 1964: 928). Transformational grammar assumes that rules for complex syntactic structures, the so-called transformational rules, differ from those for relatively simple structures. Evidence for some of the posited simple rules is indeed found in investigations of child language learning, in the study of aphasia, in experimental investigation of speakers, most spectacularly of speakers with bisected brains. But in spite of extensive psychological investigation, no psychological validity has been found for transformational rules (Fodor et al 1974). Accordingly, hopes of explanatory capabilities for transformational grammar have been severely reduced, especially since treatment of simple structures in transformational grammar differs little from that of traditional grammar. Moreover, transformational grammar has never been demonstrated to have predictive power. Many linguists are accordingly turning to other approaches, none of which have been elaborated to the extent of transformational grammar; discussion of them will then be omitted here.

While transformational grammar has yielded only limited gains, it is noteworthy that its chief proponent, Chomsky, has recognized increasing kinship with linguistics of the past (1975: 196 et passim). Such recognition supports the view that linguistics is indeed a cumulative science. As some findings of earlier linguists retain validity, so will some contributions of transformational grammar. Among these is the emphasis on abstract underlying structures. Such structures are represented above in the universal rules of section 1.7, and in the patterns exemplified throughout the book.

As these chapters illustrate, the typological approach permits predictions for these patterns. If for example a certain type of comparative construction is found in a language, the prediction can be made that the language is OV or VO; one needs little more, for example, than evidence of the pattern in the Japanese sentence below to conclude that it is an OV language (section 2.1.9, sentence 58a).

The other patterns listed in section 1.4 also permit such prediction, whether they concern the simple clause or complex structures involving embedding or complementation and the like. OV languages, for example, include characteristic relative constructions and other nominal modifiers, as well as characteristic verbal modifiers. A language including postverbal interrogative markers, like Japanese ka, or postverbal negative markers, like Japanese -nai, permits the prediction that it is OV in clause structure. The identified patterns would be used similarly in VO languages, as illustrated in Chapters 3, 4, and 6. When unmarked patterns of a language have been recognized, their structure may also be determined.

The principles formulated in this book must then be incorporated in any theory of language. Besides permitting prediction, the principle in section 1.3 also yields an explanation of its patterns once a language is determined as VO or OV. For other constructions are so arranged that verb + object sequences are not interrupted; and nominal modifiers are placed nearest the noun (object) while verbal modifiers are placed nearest the verb. By leading to such predictions and explanations the approach presented in this book is demonstrated to be essential for an understanding of language as well as fundamental for observation and description of specific languages.

Besides these contributions, study undertaken with the typological approach has illustrated the immense complexity of language as a social system. Even the numerous patterns which have been singled out here fail to encompass its structures. Moreover, like any social system, language is subject to change. The elements, rules, and principles identified have then a statistical basis; they cannot be expected to apply in the rigorous sense used of elements and laws in the physical sciences. But even in change, languages of a given type are modified in accordance with the principles presented here. If, for example, a language changes from OV to VO structure, it will introduce characteristic VO patterns, such as prepositional constructions, preverbal devices for indicating interrogation, VO patterns of complementation, and so on. Such modifications have not been discussed at any length here. They have not yet been adequately investigated. Recent papers deal with some of the problems, such as those examining complementation in early Germanic and Greek (Kurzová 1968; Ureland 1973; Li, ed. 1975; 1976; Lehmann 1976b). The principles proposed above permit explanation of many phenomena attested in the languages investigated, but much remains to be done.

An illustration of explanatory contributions may be taken from reflexivization, which seems to be expressed by a bewildering variety of surface structures when treated simply from a morphological or lexical point of view. But the progression from an OV type of reflexive construction in the Proto-Indo-European middle to the lexical and morphological devices developed in Sanskrit, Greek, Germanic, and the other Indo-European dialects is clarified when reflexivization is viewed as a syntactic category accompanying verbs (Lehmann 1974a). Moreover, the diverse devices by which reciprocal and reflexive relationships are indicated in languages of the several types are readily interpreted when these are viewed as expressions of the middle category. Conclusions derived from the study of language as a phenomenon accordingly clarify syntactic change and reasons for different surface expressions for a given syntactic category.

Although such achievements have resulted from its application, the approach presented here and its findings need elaboration through study of many more languages. As such study is carried out, additional attention must be given to morphology and phonology, which partly for lack of space could not be treated at any length here. Moreover, patterns identified in recent treatments of discourse analysis must be examined with regard to language types (Grimes 1975; Longacre 1976). Further, while exploration of syntactic patterns discussed here has demonstrated remarkable interrelationships in the syntactic structure of language, fuller understanding of these interrelationships will result as other languages are investigated with the procedures illustrated here. Such investigations will also clarify the development of patterns which are produced to compensate for the rigidities, and for short-comings of individual language types.

Linguists who like Saussure have thought deeply about language have come to be convinced of a "profound unity" underlying diversities apparent among the many languages used now and in the past. This book has been produced to demonstrate that such convictions are not illusory, and to encourage additional studies which will disclose more of that profound unity.