Approaches to Language Typology _ Shibatani

Approaches to Language Typology 1

1Approaches to Language Typology:

A ConspectusMASAYOSHI SHIBATANI AND

THEODORA BYNON

1. INTRODUCTION

The schools of language typology represented in this volume areall current and active. What unites them is a common goal and ashared scholarly tradition. They have all developed theoretical frame-works within which to account for the particular aspect of cross-linguistic variation they have selected to study, and they all havetheir roots in a shared European tradition of scholarship.

According to Greenberg (1974: 13), the word ‘typology’ gainedwide currency in linguistics only after circa 1928, but the researchactivities that can be brought under the rubric of ‘language typo-logy’ have a long history.1 Although it is not easy to ascertain thefirst formulations of a research programme of language typology,the underlying assumptions that run throughout the history oflanguage typology can be gleaned from the older passages of thenineteenth-century writings. Friedrich von Schlegel (1772–1829) wasamong the first linguists to propose a typological framework onthe basis of morphological characteristics. In Über die Spracheund Weisheit der Indier (1808: 45), he argues for a classification interms of the linguistic devices which languages employ in relatingconcepts to each other; the corresponding categories of relationalmeaning he terms ‘additional determinants of meaning’ (Neben-bestimmungen der Bedeutung):

We wish to thank Bernard Comrie for reading and commenting on an earlier version of thispaper. This work was in part supported by a grant-in-aid (04301059) from the JapaneseMinistry of Education, Science, and Culture to Masayoshi Shibatani.

1 But Gabelentz (1901: 481) contains a passage in which the author explicitly christens thefield ‘Typologie’: ‘Dürfte man ein ungeborenes Kind taufen, ich würde den Namen Typologiewählen.’ (Quotation cited in Plank (1991: 421.) Curiously enough, according to Plank (1991),the first edition of Gabelentz’s Die Sprachwissenschaft (1891), which Greenberg lists in hisbibliography, does not contain this particular passage.)

2 Masayoshi Shibatani and Theodora Bynon

Entweder werden die Nebenbestimmungen der Bedeutung durch innreVeränderung des Wurzellauts angezeigt, durch Flexion; oder aber jedesmaldurch ein eignes hinzugefügtes Wort, was schon an und für sich Mehrheit,Vergangenheit, ein zukünftiges Sollen oder andre Verhältnissbegriffe derArt bedeutet; und diese beiden einfachsten Fälle bezeichnen auch die beidenHauptgattungen aller Sprache. Alle übrigen Fälle sind bei näherer Ansichtnur Modifikationen und Nebenarten jener beiden Gattungen; daher dieserGegensatz auch das ganze in Rücksicht auf die Mannichfaltigkeit derWurzeln unermessliche und unbestimmbare Gebiet der Sprache umfasstund völlig erschöpft.(The additional determinants of meaning are indicated either throughinternal modification of the root, that is to say by means of inflectionor, conversely, in each instance by the addition of a separate word whichin itself signifies plurality, past, future obligation or some other such rela-tional concept; and these two simplest cases also represent the two maincategories of language. All other cases prove, on closer inspection, to bemere modifications and variants of these two categories; this is why thisopposition covers exhaustively the total domain of language which, asregards the variety of roots, is infinite and indeterminate.)

Though later typologists have elaborated on this simple morpho-logical classification based on the distinction between Sprachen durchFlexion and Sprachen durch Affixa—notably by adding a third iso-lating (monosyllabic) type in which the word is invariant andunanalysable—(see below), Schlegel makes it clear that the businessof language typology is, firstly, to classify exhaustively the lan-guages of the world according to specific grammatical criteria.

August Schleicher (1821–68) is better known as the founder ofthe Stammbaumtheorie, the genealogical tree model in historicaland comparative grammar, than as a typologist, but he too made animportant observation highly germane to contemporary typologicalpractice, namely the possible connections between morphologicalcharacteristics and the manner in which grammatical relations areexpressed. Schleicher (1848: 6–7) pointed out, perhaps followingWilhelm von Humboldt, upon whose work he relied heavily in typo-logical subjects, that in the isolating languages, which do not havemorphology, the grammatical relations of subject and object areexpressed by word order, whereas in agglutinative languages theyare expressed by affixes loosely attached to the root. In inflectionallanguages, on the other hand, grammatical relations are expressedfusionally with the unit expressing the root meaning. The signifi-cance of Schleicher’s observation lies in his recognition that linguis-tic properties show correlative patterns such that the presence of oneparticular property often implies the presence (or absence) of someother properties.


Our references to Schlegel and Schleicher were made not becausethese grammarians were the first typologists—it is most likely thatthey were heavily influenced by their predecessors—but because theirclear formulations of the relevant issues represent the two most pre-vailing concerns of typologists of the past and the present, namely(1) the typological classification of the world’s languages and (2) theidentification of correlative grammatical properties that define lin-guistic types. Various issues directly addressed or surrounding theundertakings by the various typologists and typological groups areall concerned with these two fundamental problems. In the follow-ing exposition, we shall elaborate on a number of salient sub-issuesthat emanate from the two goals set forth above.

2. CLASSIFICATION AND LANGUAGE TYPES

As mentioned above, Schlegel’s bipartite classification of the world’slanguages has been modified and extended by the successive at-tempts of such scholars as August Wilhelm Schlegel (1767–1845),Wilhelm von Humboldt (1767–1835), August Schleicher (1821–68),Edward Sapir (1884–1939), and Vladimír Skalicka (1909–91).Before we take up these developments in the classificatory aspect oflanguage typology, perhaps brief mention should be made of a sig-nificant shift in philosophical orientation towards linguistic com-parison that had taken place in the history of linguistics. The shiftin question is of interest not only from a historical point of view butalso from a contemporary methodological perspective in that thetwo current linguistic methods, namely generative grammar and typo-logical studies, reflect the two philosophical traditions that providedthe backdrop for eighteenth- and nineteenth-century linguisticstudies.

As explained by Paolo Ramat in his contribution to this volume,in the Age of Reason the diversity of the languages of the worldwas considered a superficial phenomenon behind which lay hiddena universe of ‘eternal ideas’ (or innate concepts) without whichrational thought was deemed impossible. These underlying mentalinvariants were said to be imperfectly reflected in the lexicon andgrammatical structure of the various languages and must thereforebe made apparent through linguistic comparison. This comparisonwas, in view of the postulated priority of the universal concepts, de-ductive and ‘constructive’ (Coseriu 1972: 214) in that the grammati-cal and semantic structure of different languages was interpreted interms of the basic categories which were deemed logically necessary


for rational thought. Language types would result from compar-ing linguistic to logical structure; for instance, cross-linguistic varia-tion in the ordering of subject, verb, and object would be relatedto the order of logical predication, in which the agent precedes theaction and the action is followed by the affected entity. Languageswhich reflect the logical order form one type, and those which donot, form another requiring an elaborate morphology to compensatefor ‘inverting’ the logical order.

For Ramat it was Wilhelm von Humboldt who represented theturning-point from this ‘philosophical’ to a new properly linguisticperspective. Humboldt supported the rationalist position to the pointthat thought depends on concepts, but departed from it in claimingthat concepts are language-specific. This is because individual lan-guages are historical entities which differ from one another in bothform and content and which continue to be developed by theirspeakers according to cultural needs. Each single language thusrepresents a unique segmentation of the external world and of theuniverse of human experience (see Ramat, this volume). Cognitionis achieved in the individual speech act when the speaker uses theforms of his or her language creatively in context-related utterance.The relationship between form and meaning, in other words, is notonly language-specific but also sufficiently elastic for new cognitiveacts to be created and communicated. From this perspective, then,linguistic comparison does not give access to, nor is it based on, auniversal logic. What is truly universal is the dependence of cogni-tion on ‘articulated sound’. That is to say, what all languages havein common is that they ‘achieve and represent cognition’ (Seiler, thisvolume).

Given that each language is a sign system in its own right linkinglanguage-specific forms and language-specific meanings, what as-pect of language is amenable to parametrization? In the passagequoted in the introduction to this chapter, Friedrich Schlegel arguesthat the inventories of lexical roots are large and incommensuratewhereas variation in the grammatical mechanisms employed in re-lating lexical concepts to each other is severely constrained cross-linguistically. It is the formal expression of relational meaning, then,which forms the basis of the so-called classical (or morphological)typology. As elaborated by Sapir (1921: chs 5–6), relational mean-ing ranges from the most abstract to the most concrete, comprisingthe basic syntactic relations (subject and object), such morpho-syntactic categories as gender, case, and tense, and paradigmaticrelations between related lexical concepts (compare farm, farms,farmed, farming, farmer).


Classical typologists have recognized three (potentially four) basicstrategies or techniques in encoding relational meaning. An inflec-tional (or flexional, fusional) language encodes relational meaningby modifying the lexical base by ‘true’ (that is to say, internal)inflection (as in English sang). This strategy was considered to achievea truly symbolic integration of conceptual and relational meaningbecause it represents relational meaning as the modification of lexi-cal units by means of ‘meaningless’ elements devoid of lexical asso-ciations. In an agglutinating (or agglutinative) language, on the otherhand, the individual exponents of relational categories are attachedone by one to the lexical base (as in Turkish ev-ler-im-de (house-PL-1SG.POSS-LOC) ‘in my houses’), leading to word structures which arerelatively complex but less integrated because, at least in places, thephonological shape of the affix may indicate its lexical origin. Anisolating language does not give overt expression to relational mean-ing or else does so by employing to this end the same kind of unitas is used for encoding lexical concepts (as the Chinese dative goalmarker gei, which derives from the full verb gei meaning ‘to give’).More marginally, an incorporating (or polysynthetic) language ischaracterized by incorporating constituents such as lexical objectsinto the verb, thereby compressing the content of a sentence into asingle word.

The shortcomings of this classical morphological typology as aclassificatory scheme under the strict sense of classification are alltoo apparent, as most languages possess forms exhibiting two ormore ‘techniques’ of encoding relational meaning. For example,English shows its isolating character in the encoding of modalmeanings by independent words such as will and may, its aggluti-native character in the regular plural formation (e.g. books), and itsinflectional character in the irregular plural and past tense forma-tion (e.g. feet, sang). Sapir (1921: ch. 6) asserts that languages intheir entirety cannot be neatly pigeonholed into a given class, thematter being a question of tendency (p. 134). It is the prevailingcharacteristics that determine the basic type of a language. Reflect-ing this assumption, Sapir accommodates the gradient characteriza-tions of linguistic types along the ‘degree of fusion’ such as ‘weaklyagglutinative’, ‘symbolic tinge’, and ‘mildly agglutinative-fusional’(see below). Moreover, languages may employ one technique inone domain, e.g. derivational concepts, and another method inanother domain, e.g. relational concepts. Languages could then be‘agglutinative-isolating’, ‘fusional-isolating’, and so on.

Clearly then typological classification, as envisaged by Sapir, whichis one culmination—the other being Skalicka’s attempt (see below)


—of classical (or morphological) typology, is inconsistent with theclassical theories of categorization and classification by, for ex-ample, Hemple and Oppenheim (1936), in which class membershipis determined categorically, together with the assumptions of theuniformity of members and of clear category boundaries. If any-thing, Sapir’s classification is much closer to that envisaged by themore recent prototype theory of categorization (e.g. Rosch 1977),which countenances a gradation from central members (prototypes)to peripheral members within a single category and fuzzy categoryboundaries.

Another innovation by Sapir was to separate from the parametersof technique the dimension of ‘synthesis’, the morphological com-plexity permitted in words. This dimension, which encompasses theparameters ‘analytic’, ‘synthetic’, ‘polysynthetic’, too, is gradient,and as with the degree of fusion, ‘mildly synthetic’, ‘mildly polysyn-thetic’, and other types of languages are recognized. The parametersalong the dimension of synthesis combine with the parameters oftechnique such that languages can be isolating and analytic (e.g.Chinese), fusional and analytic (English), agglutinative and poly-synthetic (Nootka), fusional and polysynthetic (Algonquin), and soon.

But, for Sapir, a more important classificatory scheme than thosebased on the ‘technical externals’ was the conceptual classificationbased on the following two kinds of question: (1) whether a lan-guage ‘keep[s] the basic relational concepts . . . free of an admixtureof the concrete (Pure-relational languages) or not (Mixed-relationallanguages)’, and (2) whether a language ‘keep[s] its radical conceptspure (Simple) or . . . build[s] up its concrete ideas by an aggrega-tion of inseparable elements (Complex)’, (1921: 138). In the totalclassificatory scheme arrived at by Sapir, languages can be ‘SimplePure-relational, Isolating, and Analytic’ (e.g. Chinese), ‘ComplexPure-relational, Agglutinative, Synthetic’ (e.g. Turkish), ‘SimpleMixed-relational, Fusional, Analytic (mildly fusional)’ (e.g. French),‘Complex Mixed-relational, Agglutinative (symbolic tinge), Poly-synthetic’ (e.g. Nootka), etc.

Sapir’s morphological classification took two radical departuresfrom the classical morphological typology. First, quantitative, asopposed to absolute, characterizations are recognized.2 Secondly,language types are defined in terms of combinations of properties,as opposed to single features. These two features of typologicalcharacterization, yielding quantitative characterizations (gradients,

2 See Greenberg (1954) for a rigorous quantitative approach to morphological typology.


scales, continua) and polythetic characterizations (Ramat 1987:12 ff.), are a hallmark of contemporary typology.

As is clear from the foregoing exposition, Sapir recognizes hier-archy in the importance of classificatory features, considering theconceptual classification to be the basis of fundamental types, whichcan be further subdivided according to the dimensions of techniqueand synthesis. The notion of hierarchy among the combinatoryfeatures defining types subsequently takes on a unique character,transforming itself into perhaps the single most important conceptcharacterizing contemporary typological practices, and we shall dwellon it presently; but for the moment, there still remain several areasneeding clarification with respect to the questions of classificationand language types.

First, concerning the notion of ‘type’, Sapir’s approach makes itclear that language type is to be defined in terms of a combinationof properties, which may be hierarchically ordered (see below). Type,in other words, is a ‘holistic, schematized structure’ (Seiler 1990:156) arising from a ‘cluster of properties’ (Greenberg 1974) that ex-hibit ‘preferred connections’ (Skalicka 1966) among them. WhereasSapir countenances direct gradient characterizations of actuallyoccurring types in terms of such descriptions as ‘weakly agglutina-tive’ and ‘mildly polysynthetic’, Skalicka (1935), perhaps followingHumboldt’s original idea, considers type to be an ideal referencepoint, ‘an extreme which is hardly ever realized (or never at all)’.In this framework, then, actual languages are approximations tothe ideal types, and the typological characterization of a givenlanguage can be made only in terms of the relative strength of thetypes involved in its structure.

Skalicka characterizes the flectional type, for example, as follows:

(1) as having polyfunctional endings (gender, number and case areexpressed by one single ending);

(2) no word appears without an ending, which contains both syn-tactic and semantic information (reg- never appears alone inLatin, but is always specified by an ending: regibus [‘king’-DAT-PL], reg-em [‘king’-ACC-SG] etc.; -(a)verunt in amaverunt [‘love’-3PL-PERF-INDIC-ACT] expresses membership of the word in theverb class, plus the grammatical determination of time, mood,person and diathesis);

(3) motion exists in word formation (niger [‘black’-MASC], nigra[FEM], nigrum [NEUTER] );

(4) a relatively free syntactic arrangement of words and so on.(Ramat 1987: 21 f.)


As noted above, in this Prague School typological framework, anindividual language can be characterized in terms of the relativecontribution made to its structure by the properties belonging todifferent types. Sgall (this volume) suggests that the predominanceof a particular type in the structure of a language may be inter-preted in terms of linguistic economy: adherence to a single type(either grammar words, or affixes, or alternation, or order) as themeans of encoding relational meaning would appear less costly thanan unprincipled deployment of several types side by side. That is tosay, the ultimate point of reference is a functional perspective, whichis an integral part of Prague School theory.

Defining language types in terms of a set of properties also con-stitutes the characterizations of the members of different typologicalclasses. Indeed, for Sapir it was the basic characteristics of singlelanguages that motivated grouping individual languages into morpho-logical types. Sapir’s famous passage describing the sense of thisspecific character of individual languages goes as follows:

[I]t must be obvious to any one who has thought about the question [ofthe general form of a language] or who has felt something of the spirit ofa foreign language that there is such a thing as a basic plan, a certain cut,to each language. This type or plan or structural ‘genius’ of the languageis something much more fundamental, much more pervasive, than anysingle feature of it that we can mention, nor can we gain an adequate ideaof its nature by a mere recital of the sundry facts that make up thegrammar of the language. (1921: 120)

Vilém Mathesius, a founding member of the Prague School, pur-sued this aspect of typological development. The practice, as exem-plified by his work (Mathesius 1928), is called the ‘characterological’approach, as it seeks to delineate the characteristics of individuallanguages or a group of genetically related languages.

This individualizing approach to language typology (Greenberg1974), which shares some underlying assumptions with the Hum-boldtian tradition, and which is still pursued in the anthropologicaltradition in America, must be kept apart from efforts to develop atypological framework in terms of a well-defined notion of types forcomparative purposes. Especially to be avoided is the confusion be-tween the practice of partial typology (see below) and linguisticcharacterology. As the quotation from Sapir given above says, apartial typological feature does not automatically lead to an under-standing of the underlying deep-seated character of the languagein question. Thus, contrary to the generally held intuition that erga-tive case-marking is a manifestation of some design of linguistic


structure fundamentally different from that underlying the familiarnominative-accusative languages, no consistent, deep-seated charac-ter associated with ergative case-marking has emerged so far in spiteof considerable efforts in recent years aimed at the discovery ofsome such correlation (see Plank (ed.) 1979 for the diversity ofergative phenomena).

Whereas language typology and the classification of the world’slanguages in terms of typological features imply classification of alanguage and its characterization as a whole, and while Sapir’s andthe Prague School’s typology aim at such goals, Sapir already re-cognized the difficulty of classifying the whole of a language into agiven type. He thus recognized the possibility of a single languagebelonging to two types defined by the dimension of technique.Polynesian languages, for example, are said to be ‘agglutinative-isolating’, while Cambodian is characterized as ‘fusional-isolating’—an utter contradiction in the classical typology.

It is because of this kind of oft-observed mixed characterizationthat languages allowed that scholars began to turn away from theattempt at holistic typology and to pay increasing attention to thepractice of partial typology, in which certain domains of grammarare targeted as the object of classification and characterization. Mostcontemporary typological studies fall into the practice of partialtypology, where specific constructions and grammatical phenom-ena, such as word order, case-marking patterns, relative clauses,passives, causatives, are examined, typologized, and classified. De-spite the fact that only a single domain is examined in partial typol-ogy, identification of clusters of properties and of their hierarchicalorganization plays the crucial role in this endeavour. But the focusof attention had shifted from the characterizations of individuallanguages or the specific type of languages to the drawing of cross-linguistic generalizations.

3. LANGUAGE-INTERNAL CORRELATIONS AND CROSS-LINGUISTIC

GENERALIZATIONS

Typological classification has a twofold goal, namely (1) ascertain-ing the entire range of variation, and (2) understanding the charac-ters of the members of each typological group. Of these two goals,the latter is more essential as a way of understanding the nature ofhuman language, though the former is equally important if we areto grasp the entire sphere of human language. An ideal situation isfor us to know the nature of the member languages from knowing


which typological group they belong to or from knowing a singletypological feature providing a basis for classification. Everytypologist’s dream in this regard is most eloquently expressed byGabelentz (1901: 481) in yet another famous passage in the typo-logical literature:

Aber welcher Gewinn wäre es auch, wenn wir einer Sprache auf den Kopfzusagen dürften: Du hast das und das Einzelmerkmal, folglich hast du dieund die weiteren Eigenschaften und den und den Gesammtcharakter!—wenn wir, wie es kühne Botaniker wohl versucht haben, aus demLindenblatte den Lindenbaum construiren könnten.

(But what an achievement would it be were we to be able to confront alanguage and say to it: ‘you have such and such a specific property andhence also such and such further properties and such and such an overallcharacter’—were we able, as daring botanists have indeed tried, to con-struct the entire lime tree from its leaf.)

It is precisely because of this desire that typologists have tradi-tionally sought a collection of properties as a defining feature of aparticular language type, rather than a classification based on asingle feature. But a simple aggregate of properties does not lead usto the utopian situation envisaged by Gabelentz. The propertiesmust be correlative in the sense that knowing the presence of oneproperty leads to predictions about the status of other properties.A particularly useful correlation between two properties is that ofimplicational correlation, such that the presence of one propertyimplies the presence of another.

The notion of a language type defined in terms of clusters ofconnected properties occupies a central place in both the descriptiveand generative traditions. From a descriptive point of view, a lan-guage is assumed, especially under structuralist influence, to be anorganic whole whose properties cohere together to form an inte-grated system. This assumption arises from the intuitive feeling thatdescriptive grammarians often have in actually describing individuallanguages and in observing systematic cross-linguistic similarities,but to some extent it is also a reflection of the desire for a perfectsystem. But, for generative grammarians, the idea that a language(or language type) is organized in terms of interconnected structuralproperties takes on a special significance. Their ultimate goal is toaccount for language acquisition by children—how children acquiregrammars in such a short period on exposure to an impoverishedstimulus. Surely children cannot be learning a complex grammaticalsystem bit by bit separately, which would take an enormous amountof time. It is more reasonable to hypothesize that a whole series of


grammatical properties are connected such that the acquisition ofone feature leads to the automatic acquisition of a whole arrayof related properties. It is this kind of assumption upon which thecurrent principles-and-parameters approach of Generative Gram-mar is built, and this approach also embraces a typological methodwhich attempts to account for the aggregate of typological pro-perties of a given language (group) in terms of the setting of aparticular typological parameter (see Fukui in this volume).

The utopian situation fancied by Gabelentz is what is aimedat by holistic typology, whose goal is a global characterization ofthe entire language on the basis of a small number of typologicalcharacteristics. The holistic approach is still practised by some re-searchers such as those under the Prague School tradition (see Sgallin this volume) and a number of Russian typologists such as Klimov(1974) and Yartseva (1979), who, for example, argues that ‘it is evi-dently insufficient to take some particular language level as a basisfor typological definitions; it is more advisable to seek those inter-connected and inter-level phenomena in which the specific featuresof various level markers can be displayed’ (p. 278). But attainingthe goal of holistic typology has proved quite difficult as more andmore aspects of individual languages have received scrutiny, andmost successful studies have been confined to specific domains, inwhich implicational statements make specific predictions only withina given domain or level of linguistic organization.

The discovery of implicational correlations between differentlinguistic objects or properties was first made in relation to a typo-logical domain that did not have any implication for other domainsof grammar, namely the typology of vowel systems by Trubetzkoy(1958). Roman Jakobson’s (1941, 1949) observation that the orderof acquisition of phonological units by children and of loss byaphasics follow certain patterns, and that these patterns are analogousto the typological patterns of the phonological systems of the world’slanguages, as investigated by Trubetzkoy, led him to the identi-fication of ‘irreversible solidarity’ (solidarité irréversible, einseitigeFundierung) between two elements such that the presence of a cer-tain phoneme, X, implies the presence of another phoneme, Y, in agiven phonological system. (Notice that the reverse, Y → X, is notthe case here.) Jakobson’s implicational laws (lois d’implication)are what typologists desire in realizing Gabelentz’s dream (even ifwhat is constructed is confined to a specific domain), and they aremost conspicuously employed in Greenberg’s typological works, inwhich he put forward specific (and hence falsifiable) claims aboutfeature combinations in the structure of languages. The parameters


investigated by Greenberg have produced cross-linguistic generali-zations (‘language universals’) of two kinds: (a) of the theoreticallypossible structures, only a subset occur (either absolutely or withsome frequency); and (b) implicational relations between differentfeatures hold such that the presence of one entails that of anotherbut not vice versa (unidirectional dependencies). (See Croft, thisvolume, on the nature of implicational statements.)

Greenberg’s (1966) study on word order typology has had a con-siderable appeal precisely because it has shown that a great deal ofprediction about the structure of a given language can be made oncethe basic word order of its major constituents is known. It has beenshown, for example, that if a language has dominant VSO order,(1) it has prepositions (G’s Universal 3), (2) the genitive follows thegoverning noun (G’s Universal 2, formulated originally in terms ofthe presence of prepositions), (3) question particles which are speci-fied in position by reference to a particular word in the sentence donot occur (G’s Universal 10), (4) it always puts interrogative wordsor phrases first in interrogative word questions (G’s Universal 12),(5) an inflected auxiliary precedes the main verb (G’s Universal 16),and (6) adjectives come after the noun (G’s Universal 17). (See Croft,this volume, for further developments in word order typology.)

While Greenberg’s word order typology has predictive power fora whole array of type-specific characteristics on the basis of a singletypological property, it is none the less confined to the specificdomain of word order. For example, from knowing that a givenlanguage has VSO order, we cannot predict whether it has agree-ment, whether it allows relativization on nominals other than sub-jects, whether it allows passivization of intransitive verbs, whetherit has the category of dual number, whether it allows long-distancereflexive binding, etc., etc. Word order typology, in other words, isstill a partial typology.3

Partial typological studies are thus confined in their predictive powerto specific domains, but they make far-reaching cross-linguistic pre-dictions. Greenberg’s statements are stated as universals preciselybecause of their assumed cross-linguistic validity. For example, fromhis Universal 3, we would not expect to find a VSO language havingpostpositions rather than prepositions. Indeed, as also pointed outby Ramat (this volume), seeking such cross-linguistically valid general-izations, rather than the characterization of a given type of language,is a primary concern of the majority of contemporary typologicalresearch.

3 But see Lehmann (1973), who draws both morphological and phonological implicationsfrom word order typology. See also the contributions to Lehmann (1978) for typologicalcharacterizations of individual languages classified in terms of word order.


Among the contributions to this volume, the St Petersburg/Lenin-grad School (as described in the contribution by Nedjalkov andLitvinov) and the Cologne School (as described by Seiler) mostclearly show the penchant towards partial typology aiming at cross-linguistic generalizations on well-defined, specific constructions andcognitive-conceptual domains such as causatives, passives, resulta-tives, possession, determination, and so on. Though the cross-linguisticgeneralizations drawn are based on the occurrence of different sub-types of a given construction within single languages, and accord-ingly it is possible to predict the occurrence of a certain sub-typewithin a given language on the basis of the occurrence of anothersub-type, interests are centred on the possibility of drawing cross-linguistic generalizations over the patterns of occurrence of a givenconstruction so that the entire range of possibilities of the con-struction and their possible deployment in individual languagescan be captured. Thus, for example, with regard to resultative con-structions, the St Petersburg/Leningrad group has found that acrosslanguages possessive resultatives (e.g. Japanese: Taroo wa boosi okabut-te iru (Taro TOP hat ACC wear-CONJ be) ‘Taro has a hat on’)are rarer than subjective resultatives (Taroo wa sin-de iru (Taro TOP

die-CONJ be) ‘Taro is dead’), which in turn are less common thanobjective resultatives (Taroo wa sibat-te aru (Taro TOP tie-CONJ be)‘Taro has been tied up’), yielding the implicational hierarchy of:possessive → subjective → objective. This hierarchy can be utilizedfor predicting the language-internal patterns of occurrence of dif-ferent types of resultatives. Thus, knowing that a given languagehas possessive resultatives leads us to the prediction that it also hasthe subjective and objective types.

Again, within the highly sophisticated framework of the Colognegroup, a large number of implicational statements can be gleaned.For example, in Seiler’s illustration of his method in terms of thecognitive-conceptual domain of possession, it is shown that if alanguage has possessive classifiers (e.g. ‘my as-a-pet dog’), it lackspossessive verbs (e.g. ‘own’, ‘have’). In fact a whole series of impli-cational statements can be drawn along the continuum exhibitedby the linguistic representations of the idea of possession. Needlessto say, these implicational statements apply language-internally aswell as capturing cross-linguistic generalizations.

As pointed out above, most contemporary typological works areof this kind of partial typology that seek cross-linguistic generali-zations, which can be applied in discovering language-internal pos-sibilities. In other words, contemporary partial typology strivesto draw generalizations that play the dual function of allowing usto make predictions on the language-internal properties of a given


language or a given type of language (a partial realization of theGabelentz dream) and of leading us to a general understanding ofthe nature of human language (though limited in scope).

4. COMPARATIVE TYPOLOGY

Although partial typology, seeking cross-linguistic regularities andconstraints on specific grammatical domains, is most widely pract-ised now, there are some recent resurgent moves that are perhapsmore consonant with the traditional typological approach in thatthey aim at comparing the overall grammatical structures of two ormore languages. This approach, which in some respects harks backto Mathesius’s characterological practice, can be said to be a branchof contrastive linguistics, but whereas contrastive linguistics has beenconcerned with comparison of languages with the utilitarian pur-pose of improving the methods of language teaching, the typologi-cal approach is more concerned with theoretical issues pertaining touniversal grammar and its relation to specific grammars. The movefor this unified typological comparison partly comes from dissatis-faction with the partial typological practice, as expressed by Hawkins(1986: 3):

At the same time this [partial typological] study . . . is probably missingimportant universal generalizations. It involves examination of a smallnumber of variant linguistic properties within large numbers of languages. . . In each case, small pieces of language are plucked out from the overallgrammar that contains them, and the range of attested variation is de-scribed, and universal generalizations, or truths, are proposed that arecompatible with all and only the observable patterns. Obviously, the moresuch pieces of language we study, the more universal generalizations wegain. But it is not clear that we are making much progress towards under-standing how the variants that an individual language selects in one areaof grammar are determined by, or determine, the variants that it selects inanother.

In contrast to partial typology, the comparative typology of spe-cific languages compares a large number of variant properties in asmall number of languages so as to identify the underlying princi-ples that unify the contrastive features that distinguish the languagescompared. The basic assumption, again, is familiar from the earlierassumption within typology that languages do not assemble theircharacteristic properties randomly; rather, language characteristicsare connected in a hierarchical manner such that the presence ofone characteristic may be responsible for the presence of others.


While Hawkins (1986) compares genetically related English andGerman without a specific theoretical formulation, more ambitiousattempts at both theoretical and descriptive levels have been takingplace.

The adoption of typological assumptions characterizes the signi-ficant shift in the orientation of Generative Grammar in the late1970s and the early 1980s (see Chomsky 1981). Rather than seek-ing only abstract universals that are assumed to be basic to allhuman languages, Generative Grammar began to pay more seriousattention to cross-linguistic variation. The new goal has become theuncovering of the set of universal principles (universal grammaror UG) and a unified account of the actual variant realizations ofthese principles in different languages. The new generative para-digm, known as the principles-and-parameters approach, hypoth-esizes abstract principles making up UG whose values areparameterized. Language variation results from the different valuesthat each language chooses in implementing the universal principlesgoverning human language.

Whereas the principles-and-parameters approach, or any system-atic comparison, appears to be more effectively practised with re-spect to closely related languages, it has in fact had a greater impacton comparative studies dealing with entirely unrelated languagessuch as English and Australian languages or American Indian lan-guages (see Hale 1983). Fukui, in this volume, also attempts a typo-logical comparison of two radically different languages, English andJapanese. A number of seemingly disparate properties of the twolanguages, such as the presence/absence of wh-movement, the pres-ence/absence of expletives, the freedom of word order, are broughttogether and an attempt is made to attribute the differences to twounderlying differences in the two languages: the head-parameter(English, head-initial; Japanese, head-final) and the presence (Eng-lish)/absence (Japanese) of agreement-inducing functional elements,e.g. AGR.

For a long time typological studies and Generative Grammarcountenanced different orientations and methodology: (1) the formersought features distinguishing languages, while the latter soughtcommon features; (2) the former examined a large number of lan-guages, while the latter dealt with a limited number of languages;and (3) the former confined itself to the actually observable fea-tures, while the latter posited abstract constructs rather freely.However, as the recent principles-and-parameters approach shows,the two fields are fast converging, though remaining differencesexist (see next section).


Before we turn to the next topic, it is perhaps worth summarizingthe issues concerning holistic versus partial typological practice. TheGabelentzian ideal of being able to construct the entire structure ofa language on the basis of a single or even a handful of propertiesis perhaps impossible to attain. Language structures do not seem toconsist of a simple aggregate of properties that can be drawn togetherby the presence of a certain fundamental property. Hawkins is quiteright in saying that a simple collection of cross-linguistic general-izations over bits of language in isolation does not lead us to adeeper understanding of how each generalization is related to therest of the language and how each language chooses those proper-ties permitted. Thus, to be more effective, partial typology mustorganize its domains of investigation into an interrelated networkso that possible hierarchical structuring of the domains may emerge.Indeed, some such possibilities are in the offing in the St Petersburgtypological framework and elsewhere, where inter-structural rela-tionships began to be recognized between different constructions,for example between causatives and transitive structures, betweenpassives, statives, and resultatives, and between benefactive con-structions and the basic ‘give’ constructions (see Shibatani, 1996, onthe last correlation).

The ultimate goal of typology and linguistics as a whole is tounravel the nature of linguistic properties: what are they, how arethey selected and distributed, and how are they organized? The goalcan be pursued by the methods of both partial typology and (mostlymodest) holistic approaches of comparison of two or a small numberof languages over several features. In fact, these two complementarymethods are not different substantively, as the difference betweenthem is largely a matter of degree.4

5. FORMAL VERSUS FUNCTIONAL APPROACHES

One of the major differences between the generative typologicalapproach and others is over the formal/functional dichotomy. Theterms ‘formal’ and ‘functional’, however, are used ambiguously inlinguistics. The ambiguity relevant to our discussion relates to theobjects to be typologized and to the nature of the explanationsinvoked.

Generative Grammar is said to be formal in two senses. In one4 See the proceedings of the plenary session on ‘Typology: Integral Typology versus Partial

Typology’ in the Proceedings of the Fourteenth International Congress of Linguists (Akademie-Verlag, Berlin, 1990).


sense, Generative Grammar is formal in that it deals with the ‘ab-stract formal skeleton’ (Chomsky 1975: 55) of syntactic structurespursued independently of semantic considerations in keeping withits autonomy thesis. In another sense, Generative Grammar is for-mal in that it offers explanations that are not functionally based.For example, Fukui’s contribution to this volume recognizes anabstract category ‘AGR’ that is not motivated by anything but thepatterning of formal properties relating to agreement, the distribu-tion of wh-elements, etc.5

When the term ‘formal’ is understood in the first sense of dealingwith formal objects in isolation from semantic and other functionalconsiderations, then, a fair number of typological investigations mustbe said to be formal. For example, Greenberg’s word order typo-logy takes no semantic or functional considerations into account,though the underlying semantic motivations for grammatical cat-egories and constructions are assumed. Contrast this with somemore recent approaches to word order typology such as Thompson(1978), Tomlin (1986), and Siewierska (1988), in which pragmaticand other functional considerations are given due attention.

The basic sense of the term ‘functional’ relates to the idea that thecentral function of language is to communicate experience andthought. Under this interpretation, ‘functional linguistics’ seeks todetermine how this central function is achieved in individual lan-guages and in human language in general. Some typological worksare truly functional in this sense, as they seek to determine the rangeof possible expression types over a particular conceptual sphere.

Perhaps the Cologne group countenances this functional approachmost strongly. As detailed in Seiler’s contribution to this volume,the domain of investigation firstly is not defined in terms of formalobjects such as grammatical constructions or formal elements suchas case markers, but rather in terms of cognitive-conceptual do-mains such as ‘possession’, ‘determination’, ‘apprehension’, and ‘ref-erencing’. The functional view in this paradigm is couched in thenotion of ‘problem-solving system’, that is, how languages concep-tualize a given cognitive domain and express the concepts by lin-guistic means. This functional stance has a long and venerablehistory linking the current works with Sapir’s and ultimately withthe Humboldtian conception of language as an ‘enérgeia’ (a creativeactivity) rather than an ‘érgon’ (a product).

Even though the St Petersburg group typically works with specific

5 It is ironical that Fukui terms these abstract categories as ‘functional categories’ inopposition to lexical categories whose members are associated with definite meanings.


grammatical constructions, it too shows a functional inclination inthe present sense. Thus, instead of formal characterization of con-structions, the domain of investigation is semantically defined interms of such notions as the ‘causative situation’ for causative con-structions and the ‘state of an object implying a previous event’ forresultatives. However, this approach in practice employs both func-tional and formal considerations in that the constructions investi-gated are often defined in connection with verbal or other formalcategories and their correlative semantics. For example, the StPetersburg studies on reflexives and diathesis define the domain ofinquiry more from formal than from semantic considerations.6 Thiscompromise is inevitable for some areas because it is often unclearwhether a coherent conceptual domain exists for a given linguisticfeature. Such is the case with passives and reflexives, where, espe-cially in the latter, a wide-ranging array of semantic correlates isobserved.

Among the research groups represented in this volume, the ParisRIVALC group (as described by Lazard in his contribution) ex-plicit-ly states its formal orientation. Thus, in the domain of theirinquiry ‘actancy’ is defined exclusively in morphosyntactic terms inkeeping with Tesnière’s (1959) model. The object of investigation,actancy variation, reflecting differential grammatical relationshipsbetween the predicate and the major nominal constituents, is deter-mined on the basis of morphosyntactic formal manifestations suchas verbal marking for cross-referencing, positional changes, andsensitivity to syntactic transformations.

However, the practice of the Paris group is not entirely formal.Rather than simply stating formal correlates of actancy variation,the group attempts to relate the variation with meaning and com-municative intent. In the final analysis, this group too aims at gain-ing an insight into the ‘relationships between, on the one hand,processes in the real world as reflected by the human mind (seman-tics) and the constraints of communication (pragmatics) and, on theother hand, the functioning of language with its own internal dia-lectics, its relative inertia and the unequal plasticity of its differentcomponents’ (p. 204).

The difference then lies in initial orientation—whether a concep-tual domain is first defined and then the linguistic methods thatsolve the problems of representation and expression are sought, orwhether one starts out with formal manifestations of a definite kindand then seeks their semantic and pragmatic correlates. In either

6 See also Knott 1988 on the St Petersburg methodology.


approach the ultimate goal is a functional account for the observedvariation. However, the problem discussed here relates to a morefundamental problem of cross-linguistic comparison. That is, theproblem of determining the basis of comparison, the tertium com-parationis, is a serious one for any comparative undertaking. Ramat,Seiler, and Croft address this problem (see also Seiler 1990).

Besides the basic notion of ‘functional’ discussed above, there isa more recent use of the term in reference to the approach, adoptedby many American typologists, that attempts to explain linguisticforms and their patterning in terms of external motivation. Thisapproach, as detailed by Croft in this volume, is contrasted with theformal explanation attempted in the generative paradigm. Thoughthis sense of the term ‘functional’ is more concerned with the prob-lem of explanation in linguistics, the approach is functional parexcellence in that its objective is the typology of the relation be-tween grammatical expressions and semantic as well as pragmaticfunctions.

6. TYPOLOGY AND UNIVERSALS

Typological research and language universals research are intimatelyconnected. But since the connection is sometimes blurred by a num-ber of factors, some relating to the interpretation of the notion oflanguage universals, a brief discussion seems appropriate. Initialreactions that while typology seeks differences among languages,universals research seeks what is universal, i.e. common to all hu-man languages, should give way to a more informed perspectiveonce it is recalled that typology in the first place seeks exhaustiveclassification of human languages (see §1 above) in terms of variantfeatures. Secondly, the current universals research seeks not onlywhat is com-mon to all human languages but also the range, andthe permitted variation within the range of possible natural lan-guages. In other words, the two fields have the same goal of char-acterizing human language and understanding it.

Typology should now be considered as a method in universalsresearch, as Generative Grammar has been since its original concep-tion. Typological studies define a domain of inquiry, which in atruly functional approach is defined conceptually, assemble possiblelanguage structures corresponding to the domain, and then drawcross-linguistic generalizations. The range of the observed variationand its internal structures defined by cross-linguistic generalizationsdelimit the range of possible (segments of) human languages. The


two most difficult tasks in this approach are establishment of tertiacomparationis that form a basis for the comparison and identifi-cation of a common functional denominator establishing the invari-ant on the basis of the generalizations drawn from the observedvariants (see Seiler 1990, and this volume).

As mentioned earlier, Generative Grammar has made a significantshift in its search for universals. Rather than seeking only what iscommon to all human languages, it has made serious attempts toaccount for cross-linguistic variation. The range of possible varia-tion within a given domain is represented by a parameter whosevalues vary. Universal grammar, UG, in this conception contains alarge number of parametrized principles that, together with con-stant principles, define the range of possible human languages.Languages vary within this limit as a result of different settings ofvarious parameters.

Both typological and generative methods aim at the same goal ofidentifying universal invariants and permitted variations. The differ-ence essentially is methodological, as pointed out in §4.

7. TYPOLOGY AND LANGUAGE CHANGE

Typology is often characterized as ahistorical classification of lan-guages and is contrasted with genetic classification, which is histori-cal. These two methods of language classification are indeedindependent of each other, genetic classification being based onshared lexical as well as morphological elements and typologicalclassification on shared structure. This does not, however, implythat typological and historical comparison belong to two entirelyunrelated subfields of linguistics, for, as elaborated most fully inGreenberg’s contribution to this volume—but see also those ofRamat, Sgall, and Croft, and Croft et al. (1990)—both evolutionand structure are constrained by general principles of language,and typological studies have important implications for historicallinguistics.

A diachronic perspective was soon to form an integral part of theclassical typology, isolation, agglutination, and fusion being inter-preted as successive stages in an evolutionary progression from sim-ple unanalysable to tightly integrated morphologically complexwords. Such change would be brought about by independent wordswith fairly general meaning being employed to convey relationalmeaning and in this function becoming agglutinated to others andprogressively fusing with them over time (Schleicher 1848; 1859).


This evolutionary perspective, first formulated by Horne Tooke inthe eighteenth century (Robins 1990: 172), was, however, basedon a priori reasoning rather than on actual empirical evidence and,in the positivistic climate of the late nineteenth century, this so-called agglutination theory was rejected as ‘glottogonic speculation’.As a result, historical and typological comparison went their ownways, much to the detriment of both.

The claim that new morphology reflects earlier syntax has re-cently been revived by Givón, who represents the grammaticalizationchain in the formula (1979: 209):

Discourse → Syntax → Morphology → Morphophonemics → zero

Givón argues that inherited morphology breaks down when mor-phologically related forms become separated through sound change(as is the case with the old past participles cloven and molten,which have lost their link with the verbs cleave and melt and becomeindependent adjectives; the ‘new’ participles cleft/cleaved and meltedfollow the pattern of the weak verb characterized by identical suffixesin past participle and past tense; etymologically, however, only theformer is a suffix while the latter goes back to the verb do used asan auxiliary following the main verb.) The loss of old morphologyis thus compensated by the morphologization of previous syntax,and the syntax is renewed in turn by the syntacticization of previousdiscourse structures. The diachronic processes of cliticization andagglutination, semantic bleaching and syntactic ‘tightening’ subsumedunder the notion ‘grammaticalization’ are at the forefront of currentresearch in historical linguistics (Hopper and Traugott 1994). Butwe do not now see grammaticalization as a mechanism which pro-pels entire languages from one type to another.

Joseph Greenberg, who has pioneered the integration of typologi-cal and diachronic research, has used the same approach of inter-preting types as evolutionary stages as did the tradition. Significantly,however, the notion of type is now defined by reference to a partial,and no longer an all-embracing holistic typology. Word order is aprime example. Greenberg’s word order universals (see above, §3)firstly allow us to interpret several seemingly isolated changes thatdistinguish the languages of modern Europe from their commonancestor (ultimately Proto-Indo-European) as related episodes of asingle ‘drift’ from verb-final to verb-medial order. This is because anumber of universals relate the relative order of verb and object tothat of head-noun and genitive attribute and of head-noun andattributive adjective (see Hawkins 1983 for a conspectus). Secondly,implicational universals of the form ‘If x then y’ impose constraints


on the relative order of changes. Assuming that in a given languagethe position of the object were to change from preverbal to postverbalposition, Greenberg’s Universal 25 (‘If the pronominal object fol-lows the verb, so does the nominal object’) predicts that while thenominal object can shift first (as in French, which shows variationbetween postverbal nominal object and preverbal pronominal object:Je vois Jean ‘I see John’, Je le vois ‘I see him’), the pronominal objectmay not change position first as this would violate the universal.

Typological considerations are, finally, relevant to historical re-construction in that if protolanguages were to violate any knownuniversals this would make them different from ‘real’ languages andwould make postulated diachronic developments based on themsuspect. In the comparative grammar of the Indo-European lan-guages, for instance, the so-called glottalic theory (Gamkrelidze andIvanov 1973) argues for ejectives in the place of the traditionallyreconstructed voiced plosives because phonological systems com-prising a voiceless, a glottalized, and an aspirated series (/t/, /t?/,/th/) are attested, while a system comprising a single aspirated seriesalongside voiced and voiceless unaspirated plosives (/t/, /d/, /dh/) issuspect for the same reasons (as already argued in Jakobson 1958).

The basic principles of design postulated by typology are obvi-ously reflected in both synchrony and diachrony. Historical linguis-tics has undoubtedly benefited from typology, but it must not beforgotten that synchronic states exhibit fuzziness and variation andstructural indeterminacy is often an indicator of ongoing change,which is where synchrony and diachrony meet.

8. CONCLUSION

The goal not only of language typology but also of a truly ‘gene-ral’ General Linguistics must lie in ‘explain[ing] the way in whichlanguage-specific facts are connected with a unitary concept oflanguage’ (Seiler 1990: 157). Typological comparison has movedfrom a search for universals of the kind ‘all languages have . . .’ tomore significant ones which are more telling. These are, firstly,the universals which state that of the x number of theoretically pos-sible combinations of properties, only a subset is actually found inthe languages of the world—be it absolutely or ‘with more thanchance frequency’ as Greenberg has tended to formulate it—and,secondly, the implicational type which makes the presence of onefeature dependent on that of another. Such recurrent constraintscannot be accidental and point to underlying invariants. For the


formalist these lie in the abstract categories and principles of Uni-versal Grammar; for the functionalist they lie in the task languagesneed to perform in the process of conceptualizing and representingthe spectrum of human experience.

In both the above paradigms unity and diversity are subsumedunder the higher principle of variation, which is reflected equally inthe synchronic and the diachronic dimension. The contributors tothis volume outline and illustrate with model analyses the differentpathways they are taking towards that common goal.

REFERENCES

Chomsky, N. (1975), Reflections on Language. New York: Pantheon Books.—— (1981), Lectures on Government and Binding. Dordrecht: Foris.Coseriu, E. (1972), Geschichte der Sprachphilosophie von der Antike bis

zur Gegenwart. Tübingen: Gunter Narr.Croft, W. et al. (eds.) (1990), Typology and Diachrony: Festschrift for

Joseph Greenberg. Amsterdam: John Benjamins.Gabelentz, Georg von der (1901), Die Sprachwissenschaft. Ihre Aufgaben,

Methoden und bisherigen Ergebnisse. Leipzig: Chr. H. Tauchnitz.Gamkrelidze, T. V. and Ivanov, V. V. (1973), ‘Sprachtypologie und die

Rekonstruktion der gemeinindogermanischen Verschlüsse’. Phonetika,27: 150–156.

Givón, T. (1979), On Understanding Grammar. New York: Academic Press.Greenberg, J. (1954), ‘A Quantitative Approach to the Morphological

Typology of Language’. In R. F. Spencer (ed.), Methods and Perspectivein Anthropology, 192–220. Minneapolis: University of Minnesota Press.

—— (1966), ‘Some Universals of Grammar with Particular Reference tothe Order of Meaningful Elements’. In J. Greenberg (ed.), Universals ofGrammar (2nd edn), 73–113. Cambridge, Mass.: MIT Press.

—— (1974), Language Typology: A Historical and Analytic Overview.The Hague: Mouton.

Hale, K. (1983), ‘Warlpiri and the Grammar of Nonconfigurational Lan-guages.’ Natural Language and Linguistic Theory, 1: 5–47.

Hawkins, J. (1983), Word Order Universals. New York: Academic Press.—— (1986), A Comparative Typology of English and German: Unifying

the Contrasts. London and Sydney: Croom Helm.Hempel, C. G. and Oppenheim, P. (1936), Der Typusbegriff im Lichte der

neuen Logik. Leiden: Sijthoff.Hopper, P. and Traugott, E. C. (1994), Grammaticalization. Cambridge:

Cambridge University Press.Horne, K. (1966), Language Typology: 19th and 20th Century Views.

Washington, DC: Georgetown University, Institute of Language andLinguistics.


Jakobson, R. (1941/1968), Child Language, Aphasia and PhonologicalUniversals (trans. A. R. Keiler). The Hague: Mouton.

—— (1949), ‘Les Lois phoniques du langage enfantin et leur place dans laphonologie générale’. Reprinted in Roman Jakobson: Selected Writings,vol. 1 (1971). The Hague: Mouton.

—— (1958/1971), ‘Typological Studies and Their Contribution to Com-parative-Historical Linguistics’. In Selected Writings, vol. 1, 523–532.The Hague: Mouton.

Klimov, G. A. (1974), ‘On the Character of Languages of Active Typo-logy’. Linguistics, 131: 11–25.

Knott, J. (1988), The Leningrad Group for the Typological Study of Lan-guages: Introduction and Translations. London: School of Oriental andAfrican Studies, University of London.

Lehmann, W. P. (1973), ‘A Structural Principle of Language and its Im-plications’. Language, 49: 47–66.

—— (ed.) (1978), Syntactic Typology. Austin: University of Texas Press.Mathesius, V. (1928), ‘On Linguistic Characterology with Illustrations

from Modern English’. Reprinted in J. Vachek (ed.), A Prague SchoolReader in Linguistics (1964), 59–67. Bloomington: Indiana UniversityPress.

Plank, F. (ed.) (1979), Ergativity: Towards a Theory of GrammaticalRelations. London: Academic Press.

—— (1991), ‘Hypology, Typology: The Gabelentz Puzzle’. Folia Linguistica,25 (3–4): 421–458.

Ramat, P. (1987), Linguistic Typology. Berlin: Mouton de Gruyter.Robins, R. H. (1990), A Short History of Linguistics. London: Longman.Rosch, E. (1977), ‘Human Categorization’. In N. Warren (ed.), Studies in

Cross-Cultural Psychology, 3–49. London: Academic Press.Sapir, E. (1921), Language. New York: Harcourt, Brace and World.Schlegel, Friedrich von (1808), Über die Sprache und Weisheit der Indier.

Heidelberg: Mohr & Zimmer. New edition by E. F. K. Koerner (1977),Amsterdam: John Benjamins.

Schleicher, A. (1848), Sprachgeschichtliche Untersuchungen I: zurvergleichenden Sprachengeschichte. Bonn: H. B. König.

—— (1859), Zur Morphologie der Sprache. Mémoire de l’Académie desSciences de St-Pétersbourg, series 7, vol. 1, no. 7, 1–38.

Seiler, H. (1990), ‘Language Typology in the UNITYP Model’. In W.Bahner et al. (eds.), Proceedings of the Fourteenth International Con-gress of Linguists, 155–167. Berlin: Akademie-Verlag.

Shibatani, M. (1996), ‘Applicatives and Benefactives: A Cognitive Account’.In M. Shibatani and S. A. Thompson (eds.), Grammatical Construc-tions: Their Form and Meaning, 157–95. Oxford: Oxford UniversityPress.

Siewierska, A. (1988), Word Order Rules. London: Croom Helm.Skalicka, V. (1935), Zur ungarischen Grammatik. Prague: Universitas

Caroliensis.


Skalicka, V. (1966), ‘Ein typologisches Konstrukt’. Travaux du Cercle Lin-guistique de Prague, 2: 14–28.

Tesnière, L. (1959), Éléments de syntaxe structurale. Paris: Klincksieck.Thompson, S. (1978), ‘Modern English from a Typological Point of View:

Some Implications of the Function of Word Order’. Linguistische Berichte,54: 19–35.

Tomlin, R. (1986), Basic Word Order: Functional Principles. London andSydney: Croom Helm.

Trubetzkoy, N. S. (1958), Grundzüge der Phonologie. Göttingen:Vandenhoek & Ruprecht.

Yartseva, V. (1979), ‘The Semantic and Structural Parameters of a Lan-guage Type’. Acta linguistica Academiae scientiarum Hungaricae, Tomus29 (3–4): 275–287.

Approaches to Language Typology _ Shibatani

Documents

Transcript of Approaches to Language Typology _ Shibatani