Linguistic Typology Research Paper

Typological classification is frequently contrasted with the genetic classification of languages and is often characterized as ahistoric in its orientation. This was also not the case in the beginning of the typological studies in the early nineteenth century, when genetic and comparative approaches were not clearly distinguished from the typological one. Indeed, it was in the context of developing a program of comparative historical research in his 1808 book Uber die Sprache und Weisheit der Indier (On the Language and Wisdom of the Indians) that Friedrich Schlegel (1772–1829) developed a typological framework most famous in the history of typology. It was mainly due to the structuralist context, in which modern typology arose, that it countenanced a synchronic outlook. In contemporary typological scenes dynamicization of typology, as propounded most forcibly by Greenberg (1969, 1995), is fast obliterating the view that typology is a purely synchronic discipline.

1. Classical Typology

In the above-mentioned book Friedrich Schlegel proposed what is believed to be the first typological framework, according to which the languages of the world are classified into two types—flexional (or inflectional) type and affixing (or agglutinative) type. Though later typologists, most notably August Wilhelm Schlegel (1767–1845) and Wilhelm von Humboldt (1767–1835), elaborated on his bipartite classification, Friedrich Schlegel, alluding to the ultimate goal of typological classification, made a bold claim that the opposition between Sprachen durch Flexion and Sprachen durch Affixa represents the two main categories of language, all other cases being mere modifications and variants of these.

The typological studies of the early nineteenth century are also known as morphological typology since the classificatory basis was the morphological structure of words. A beginning linguistics textbook typically introduces four types of language from this early endeavor: (a) inflectional (or fusional) type: relational meaning is expressed by internal modification of the lexical base such that component parts are not easily isolable (e.g., Indo-European languages, English sing:sang:sung); (b) agglutinative type: words consist of easily segmentable parts, each indicating a distinct grammatical meaning (e.g.,Turkish e -ler-imde house-PL-1SG.POSS-LOC ‘in my house’); (c) isolating type: words do not give overt expression of relational meaning or else do so by employing to the end the same kind of unit as is used for encoding lexical concepts (e.g., in Chinese the verb gei ‘to give’ is used also as an indirect object marker); (d) incorporating (or polysynthetic) type: nominal arguments (typically objects) and pronominal forms referring to them are incorporated in the verb form (e.g., Ainu awakka-ta-re1SG-water-draw-CAUS ‘I make someone draw water’).

Classical typology embodied system internal inconsistencies and had associated value judgements that caused its decline in the second half of the nineteenth century, when the study of genetically related languages was viewed as the only possible scientific comparative study of languages. Different languages were thought to represent different stages of the development of the languages and language families with the belief that the inflectional type, which fused distinct morphological information into single morphemes, represented the highest degree of linguistic sophistication. This assumption put classical IndoEuropean languages at the top of the evolutionary ladder and isolating languages of Asia, e.g., Chinese, at the bottom, contradicting the long history of the Chinese language and the advanced civilization associated with Classical Chinese.

The quadripartite classification has shortcomings as a true classificatory scheme in that a language may have morphological characteristics of several types simultaneously. English, for example, shows an inflectional characteristic in some irregular forms (e.g., bring:brought, man:men), an agglutinative characteristic in regular verbal and nominal inflections (e.g., walk:walked, book:books), an isolating characteristic in certain modal expressions (e.g., can not walk), and an incorporating feature in some compound forms (e.g., to hand-knit a sweater). Edward Sapir’s multidimensional typology developed in his 1921 book Language (Sapir 1921) was in part response to the need to accommodate gradient characterizations of linguistic types, as he saw that languages in their entirety could not be neatly pigeon-holed into a given type, the matter being a question of tendency. While classical typology saw the culmination in Sapir’s highly elaborate classificatory scheme, his followers have not generally put it into practice. Indeed, the only place where classical, morphology-based typology has been seriously pursued is Prague, where Vladimir Skalicka (1909–1991), who recognized five types of languages (the four previous types plus the introflexive type), was a pillar of the so-called Prague School of Typology (see Sgall 1995).

While morphological typology as formulated by the nineteenth century grammarians is no longer considered to be adequate in meeting the ultimate goal of language typology, classical typology saw several developments of contemporary relevance. It is perhaps even fair to say that the followers of classical typology have contemplated most contemporary issues of language typology. For example, August Schleicher (1821–1868) recognized the possible connections between morphological characteristics and the manner in which grammatical relations are expressed, thereby pointing to the current view that a language type is to be defined in terms of a combination of properties that exhibit correlative patterns. The evolutionary chain of isolating-agglutination-inflection envisioned earlier is now recast as a cyclic development and such a model is integrated in the current study of grammaticalization (see Bybee 1985, Hagege 1988). The Humboldtian claim that the structure of a particular language reflects its speakers’ Weltanschauung has stimulated a line of research on the relationship between language types and the speakers through the nineteenth century leading up to the so-called Sapir– Warf Hypothesis in the early twentieth century. This hypothesis claims a connection between the linguistic structure and the way the speakers perceive the outside world. As typological studies take on more cognitive outlook (see DeLancey 1981), a reappraisal of the Humboldtian tradition may be in order. An interesting development along this line is the recent revival of interest in ethnopsychology and the development of what is dubbed as ‘ethnosyntax’ by Wierzbicka (1992) and others.

2. Modern Typology

A major difference between classical typology and modern typology, developed in the middle of the twentieth century, lies in its scope and methodology. Classical typology sought a global characterization of the entire language according to a small number of typological characteristics. This holistic approach has proved to be too ambitious, and modern typologists practice partial typology, where specific phenomena or individual grammatical constructions are studied. Methodologically, modern typology heavily relies on cross-linguistic distributional patterns. Implicational universals are drawn from the distribution of existing types and non-attested types.

Roman Jakobson (1896–1982) is generally regarded as the founding father of modern typology. In his seminal paper, Jakobson (1958 1971) recognized that typological analysis leads to the discovery of the dependency relation between two elements such that the presence of one element, X, implies the existence of another element, Y, in the same system (but not vice versa). For example, the parameters of voicing and aspiration in consonants yield the following four types of consonant system (illustrated with dentals); (1) /t/, /d/; (2) /t/, /t^h/, /d/; (3) /t/, /t^h/, /d/, /d^h/; (4) /t/, /d/, /d^h/. Of these four logically possible types, only the first three are actually observed—no language has voiced aspirates in the absence of voiceless aspirates, i.e., the fourth system. From this, it is possible to formulate an implicational statement that the presence of voiced aspirates implies the presence of voiceless aspirates. This statement constitutes a universal on human language if no language indeed violates the pattern implied by this statement—that no consonantal system can contain voiced aspirates in the absence of voiceless aspirates.

Developing Jakobson’s insight in a systematic and over-arching manner, Joseph H. Greenberg (1915– 2000) has changed the whole perspective of linguistic typology transforming it into a new method of universal research. Greenberg’s (1966) study on word order typology drew a large number of implicational universals encompassing both morphological and syntactic levels, and some transcending them. That languages differ in word order is a common place experience of those who are exposed to foreign languages. English is an SVO (Subject-Verb-Object) language, while Japanese is an SOV language. The less frequently observed VSO order is instantiated by Welsh. The impact of Greenberg’s work was very strong because of the far-reaching predictions it made on the basis of this kind of simple, basic feature of language. For example, it has been shown that if a language has dominant VSO order, (a) it has prepositions (G’s Universal 3); (b) the genitive follows the governing noun (G’s Universal 2, originally formulated in terms of the presence of prepositions); (c) question particles which are specified in position by reference to a particular word in the sentence do not occur (G’s Universal 10); (d) it always puts interrogative words or phrases first in interrogative word questions (G’s Universal 16 ); (e) an inflectional auxiliary precedes the main verb (G’s Universal 16 ); and (f ) adjectives come after the noun (G’s Universal 17).

Greenberg’s universals have the following two characteristics. They are implicational, such that they are restricted universals. The universals cited above obtain only with regard to VSO languages, and they have nothing to say about those languages having a different word order. Secondly, Greenberg recognizes statistical universals or high cross-linguistic tendencies, which do not obtain absolutely. For example, Greenberg’s Universal 4 is formulated as: ‘with overwhelmingly greater than chance frequency, languages with normal SOV order are postpositional,’ which allows a language like Persian, which has SOV order with prepositions. Beside implicational and statistical universals, linguists in general recognize unrestricted universals, which are believed to obtain absolutely; e.g., all languages have vowels and consonants; all languages have nouns and verbs.

Implicational statements are also obtainable from hierarchies that grammatical categories exhibit. The most well known grammatical hierarchy is that of grammatical relations formulated by Keenan and Comrie (1977): Subject>Direct Object>Indirect Object>Genitive>Oblique Objec>Object of Comparison. This hierarchy, derived from the crosslinguistic relativization patterns, predicts, for example, that if a language allows relativization on an oblique object (i.e., an object of a preposition), then it also allows the same relativization strategy to apply to all those nominals holding the grammatical relations higher in the hierarchy than the oblique object.

Another hierarchy with far-reaching ramifications is the so-called animacy hierarchy: first /second person>third person>pronoun>human noun>animate noun>inanimate noun. In a language such as the Australian language Dyirbal, where an agentive nominal of a transitive clause may receive either ergative or nominative marking—split-ergative marking—those high on the hierarchy tend to be marked nominative, while those low on the hierarchy tend to be marked ergative, such that we can draw implicational statements to the effect that if an agentive nominal X in the hierarchy is marked nominative, all those higher in the hierarchy are also marked nominative. The animacy hierarchy is also at work in voice phenomena such that an agentive nominal high in the hierarchy tends to be expressed in subject position, i.e., in the active voice, while when a patient nominal is high in the hierarchy, passivization tends to occur. A corollary of these tendencies is that those nominals high in the hierarchy are less likely to be expressed overtly as an agentive phrase (as in the English by agentive phrase) in the passive clause.

Recent studies of prototype categories have revealed that grammatical categories and constructions form continua rather than exhibiting rigid category boundaries. For example, studies on transitivity have shown that the traditional binary distinction between intransitive and transitive verbs (or constructions) needs to be recast as a multi-valued continuum (see Hopper and Thompson 1980). A simpler and much coarser form of this continuum yields the following hierarchy: Transitive>Active intransitive>Inactive intransitive, where active intransitive verbs/constructions are those expressing activities (e.g., ‘to dance’), and inactive intransitive verbs simple processes (e.g., ‘to fall’). An implicational statement can be also derived from this hierarchy; e.g., if a language permits passivization of an inactive intransitive, then it also permits active intransitives and transitive verbs to undergo passivization. Also noun incorporation (see above) basically follows this hierarchy such that the object of a transitive clause is the easiest to incorporate, followed by the subject of an inactive intransitive clause, the subject of an active intransitive clause, and finally the subject of a transitive clause. Thus, if a language incorporates the subject of an active intransitive clause, then it is likely to incorporate the subject of an inactive intransitive clause and the object of a transitive clause.

3. Functional Typology

The term typology is used in reference to a theoretical and methodological approach that contrasts with other, especially Chomskyan generative, approaches. The Chomskyan generative approach is said to be formal in two senses. In one sense, Generative Grammar is formal in dealing with the ‘abstract formal skeleton’ (Chomsky 1975p. 55) of syntactic structures as an autonomous system independent of semantic or functional factors. In another sense Generative Grammar is formal in that its explanations are in terms of the system internal elements and structural relations. The typological approach sharply contrasts with these two characteristics of Generative Grammar. Its basic orientation is functional in that it assumes that the central function of language is to represent cognition and to communicate experience and thought. The task of typology is to determine the range of variation that language permits in achieving this central function. In this framework a language is seen as a problem-solving system; how languages conceptualize a given cognitive domain and express the concepts by linguistic means (see Seiler 1995). Studies dealing with specific constructions such as causative and possessive constructions thus start out with the formulation of the concepts or the cognitive domains to be expressed and typologize the relation between grammatical form and semantic function (see Shibatani 1976 and Seiler 1995). This cognitive-functional stance has a long and venerable history linking the current works with Sapir’s and ultimately with the Humboldtian conception of language as an ‘energeia’ (a creative Activity) rather than an ‘ergon’ (a product).

The term ‘functional typology’ is used in recent years in order to underscore the functional orientation the current typological research countenances. This approach also differs from Generative Grammar in its recourse to external explanations. Explanations for cross-linguistic patterns of linguistic structure are sought in functional motivations such as iconic motivation—preference for the structure of language to reflect and the conceptual structure in a parallel manner (see Haiman 1985). Haiman has shown that conceptual distance is reflected in grammatical distance. Thus, inalienable possession, which involves an inseparable or a closely connected whole-part relation, is expressed by a more tightly integrated form than inalienable possession; e.g., m-polu ‘my back’ vs. ra perei ‘my house’ (Kpelle). Similarly, in causative expressions, shorter forms tend to express direct causation and the longer forms indirect causation, e.g., kill vs. cause to die. Another line of research along the functional typological premise has been developed by Hawkins and his associates in which processing mechanisms are invoked as an explanation for observed linguistic patterns (see e.g., Hawkins 1990, Hawkins and Cutler 1988).

4. Typology And Language Change

The importance of typology to one of the major concerns of historical linguistics, namely reconstruction of proto-languages, has long been demonstrated. The basic idea here is that where a reconstruction conflicts with the known typological facts, there is a prima facie reason to suspect its validity. The most celebrated phonological example of this use of typology as a control on reconstruction is concerned with the Proto-Indo-European consonant system. The standard reconstruction posited a three-fold system of voiced and aspirated consonants (illustrated with dentals) /t/,/d/,/dh/. However, Jakobson (1958/1971) noted that typological investigation revealed that no language would contain /dh/ in the absence of the voiceless counterpart /th/, casting doubt on the validity of the Proto-Indo-European system reached by the standard reconstruction. This has led to either reinstating the Neogrammarians’ earlier four-point stop system of /t/, /th/, /d/, dh/, or developing alternative hypotheses. The so-called glottalic theory, which replaces the traditionally reconstructed voiced stops with ejectives, has also been informed of typological results. Phonological systems comprising a voiceless, a glottalized, and an aspirated series ( /t/, /t^?/ , /t^h/) conform to synchronic typological findings, but the system arrived at by the conventional re-construction does not (see Gamkrelidze and Ivanov 1973).

The move to bring typology to the heart of diachronic linguistics, namely linguistic change itself, has been spearheaded by Greenberg (1969, 1995), who advocates the dynamicization of a synchronic typology. Greenberg offers what he calls the ‘stateprocess model,’ which views attested language types as possible stages through which languages can pass and non-attested language types as impossible stages into which a language can change. In other words, just as synchronic typology puts constraints on language types, constraints on typological change can be viewed as limitations on language change.

The connection between synchronic typology and possible change can be illustrated by the typology of number categories in the noun. It has been shown that the category of dual is only found in a system having a singular/plural distinction—no language has a singular/dual system or a plural/dual system. When a language with no number distinction develops a number system, it involves a singular/plural distinction; no language develops a system with a tripartite singular/dual/plural distinction without going through a stage in which a singular/plural distinction has been established. In this way, diachronic typology seeks to identify allowable and non-allowable transition from one state to another so as to determine possible paths of language change.

In the Greenbergian diachronic typology, typological states and transitions between states are characterized by two logically independent factors— stability and frequency. Stability represents a probability that a language, which is in a particular state, will exit from that state, while frequency is the likelihood that a language will enter a particular state. For example, a state of having nasal vowels is characterized as having high frequency but low stability, because a vowel frequently acquires the nasality feature from an adjoining nasal consonant, but once a system acquires nasal vowels, it tends to retain them. The dimensions of frequency and stability of typological states render the state-process model a probabilistic one, which countenances the gradualness of the transition from one language state to the next. In this model synchronic exceptions are accounted for as transitional states arising from gradual diachronic process of change from one state to another.

The two main concerns of diachronic typology— gradualness of historical change and its directionality—have received focused attention in the recent developments in grammaticalization theory (see Bybee 1985, Heine et al. 1991, Hopper and Traugott 1994, Lehmann 1982). Grammaticalization refers to the process by which lexical items come to take on grammatical functions such as marking of tense, aspect, grammatical relations, definiteness, modality, etc. In Chinese the verb gei ‘to give’ is used as an indirect object marker, and the verb iru ‘to be/to exist’ in Japanese has developed an aspectual usage of indicating an on-going Activity, the progressive aspect.

Grammaticalization typically involves synchronized change in several different grammatical dimensions. As a lexical item gets grammaticalized, it tends to undergo phonological reduction, which in turn renders the form less independent, downgrading it to a morphologically dependent status as in the case of enclitics and affixes. Semantically, a lexical item gradually loses its original referential meaning and acquires such relational meanings as those previously noted. As in the case of the Chinese verb gei and the Japanese verb iru, Grammaticalization is a gradual change such that these forms co-exist as verbs along with the newly developed grammatical functions. The two main concerns of grammaticalization theory are the relationship between the input and the output of grammaticalization—what kind of lexical items give rise to what kind of grammatical functions?—and the directionality of change—is the change unidirectional from concrete to abstract?

The gradual nature of historical change brought to the fore by diachronic typology and grammaticalization research has an important implication to the matter of definition as well as the synchronic characterization of a language. Many phenomena are in transition from one state to another, and as such they are not amenable to categorical definitions. For example, whether Chinese has prepositions or not is not an easy question to answer, as forms such as the indirect object maker evolving from the verb gei ‘to give’ can be construed either as verbs or prepositions. The existence of transitional phases of this kind calls into question the traditional distinction between synchronic linguistics and diachronic linguistics. Dynamicization of a synchronic description is the next logical step that diachronic typology suggests.

5. Typological Differences In Clausal Organization

As described in Sect. 4, modern typology typically examines specific domains or individual grammatical constructions in order to ascertain possible variation in human language. This leaning toward partial typology stems from the difficulty encountered in attempting to typologically characterize languages in their entirety. Despite this, there is a lingering sense among typologists that a holistic characterization of a language is a desideratum. Georg von der Gabelentz most eloquently expressed this feeling at the beginning of the twentieth century:

But what an achievement would it be were we to be able to confront a language and say to it: ‘‘you have such and such a specific property and hence also such and such further properties and such and such an overall character’’—were we able, as daring botanists have indeed tried, to construct the entire lime tree from its leaf. (Gabelentz 1901, p. 481)

Whether or not the ideal state entertained by Gabelentz can ever be reached remains to be seen. However, one area that merits further investigation is the differences that languages show in the organization of clause elements. In compensation for the emphasis on methodological matters in the preceding expose, the following allows a glimpse of what actual typological investigation or comparison of languages in general involves.

In expressing a transitive situation involving an agent and a patient participant, languages typically distinguish these in some way, so that who is acting on whom is made clear. English, for example, achieves this by using different pronominal forms; e.g., ‘He hit him.’ In the case of intransitive clauses, its sole nominal argument (S) can be assimilated either to the form of an agentive nominal (A) or that of a patient nominal (P). English assimilates it to the nominal form of A—‘He fell, He ran.’ A clause that treats S and A alike and distinguishes P from these is said to display the nominative-accusative coding pattern (S=A≠P), and the languages possessing such clauses predominantly are called nominative-accusative type languages.

Some languages organize clauses differently from the nominative-accusative pattern. Many Australian aboriginal languages follow the absolutive-ergative pattern, whereby S is identified with the P nominal a transitive clause, while distinguishing A from these; i.e., the S=P≠A pattern. Clauses may be organized in yet another way. The American Indian language Lakhota, for example, splits S nominals into an agent- like category (S_a) and a patient-like category (S_p), yielding the coding pattern, known as active-type: S_a =A≠P=S_p.

All of these languages have a single type of dominant transitive clause, where the coding of A and P is fixed. Some languages have more than one type of transitive clause making a unique comparison of intransitive clauses with transitive clauses problematic. Philippine languages are such languages, where a simple transitive clause comes in two variant forms. Cebuano, for example, renders the sentence ‘Juan read the book’ in two ways; Ni-basa si Juan sa libro (AF-read TOP Juan the book), Gi-basa ang libro ni Juan (GF-read TOP book ACT Juan). That is, either A or P can be coded as what is called Topic (marked si for proper names and ang for common nouns) of a transitive clause. The S nominal of an intransitive clause is coded as a Topic as in: Ni-dagan si Juan (AF-run TOP Juan) ‘Juan ran.’ (AF=ctor-focus marking; GF=goal-focus marking; TOP=Topic marker; ACT=Actor marker).

The focus marking in the verb indicates what semantic role is being coded as Topic, such that the prefix niin Cebuano indexes an agentive Topic, and gia patient topic. This applies to both transitive and intransitive clauses such that an intransitive clause with a patient S Topic would trigger gimarking; Gi-kapoy si Juan (GF-tired TOP Juan) ‘Juan is tired.’ This split treatment of S Topic marking in the verb is reminiscent of the active-type-coding pattern. However, Philippine languages are not active-type languages in that the nominal marking is consistent for S—it is always realized as si/ang-marked Topic—and that there are two competing forms of a transitive clause (see Shibatani 1985 on the Philippine voice system).

The four types of coding pattern examined have important implications for some basic grammatical categories. First, the notion of Subject is problematic in those clause types displaying coding patterns other than the nominative-accusative pattern. Subject, as found in European languages, refers to the category obtained from the S=A grouping in the nominativeaccusative pattern. It is a category deriving from the assimilation of S to A, and accordingly is agent-based.

Morphological identification of S and P obtains in an ergative-type language. A long-standing question in typology is whether this grouping of S and P should be identified as Subject. One obvious problem with this move is that whereas Subject is an agent-based category, the S=P grouping is a patient-based category, since it obtains by assimilating S to P. Thus a conservative view is that the S=P grouping forms a category distinct from Subject. Since S and P typically involve zero marking, the term Absolutive suggests itself as the label for this category.

In active-type languages, there is no unique identification between the totality of S and A or P; and therefore, neither Subject nor Absolutive category obtains. In Philippine linguistics, Topic is often likened to Subject, but since it is based on neither A nor P—either can be Topic and S can be identified with either kind of Topic—it is neither an agent-based nor a patient-based category.

While different clausal types of the world’s languages may display grammatical relations resembling Subject, they are not entirely identical. Subject is agent-based, whereas Absolutive is patient-based. Philippine Topic is neutral in its alignment with the semantic roles. These different grammatical relations, however, can be identified at a functional level, in that they are the essential relations that a nominal must bear in clause formation. That is, Subject, Absolutive, and Topic are an indispensable nominal category in a clause over which a verb predicates. Predication will not be complete without a nominal category about which the verb says something. The primary status of these grammatical relations is reflected in the primacy they enjoy in such grammatical processes as relativization and focusing. Subject, Absolutive, and Topic, in other words, can be identified as a Primary Relation at the functional level, despite the differences at the substantial level.

The differences in grammatical relations have some additional implications. Firstly, it predicts that those languages that do not have the Subject category do not show the active-passive voice opposition. Active voice (as expressed by ‘John hit Bill’) is defined in terms of the alignment of A and Subject, while passive voice typically involves the alignment of P and Subject, as in ‘Bill was hit by John.’ From this it is obvious that languages without the Subject category cannot have the active-passive opposition. Still, other languages may show a different kind of voice contrast. Ergative languages often display ergative-antipassive voice opposition, characterized by the realignment of the absolutive category over P and A.

In both active-passive and ergative-antipassive oppositions, there is a basic voice associated with the basic form of the verb. In the former, active is the basic voice, and in the latter, the ergative voice is basic. Thus, a transitive situation is normally expressed either as an active form or as an ergative form in the languages having these systems. Derived voice forms, the passive and the antipassive, contain derived verb forms, and they are used far less frequently than the unmarked, basic voice forms. Philippine-type languages, in contrast, have no unmarked voice category—the alignment of Topic with A or P is equally marked in the verb. A characteristic of the fluid-voice system of this kind is that both A-Topic form and PTopic form are used with equal facility such that there is no lop-sided distribution of voice forms as in the other two systems discussed previously. In contradistinction to these types of languages, active-type languages are known to lack voice opposition of the kind discussed here. The three voice mechanisms studied—the active/passive system, the ergative/antipassive system, and the fluid-voice system of Philippine languages—are different in their substance. Yet they are comparable at the functional level in that they all involve diathetic realignment of a Primary argument.

Languages can differ with respect to the nature of grammatical relations and voice mechanisms depending on the way clauses are organized. Besides these, cross-linguistic differences in other grammatical phenomena have been reported, some of which may be attributable to the typological differences outlined here. The question of whether these differences have profound implications for the overall organization of grammar aside, it is becoming increasingly clear that languages are likely to differ at the level of grammatical organization far more greatly than was believed in the latter half of the twentieth century. This suggests that cross-linguistic similarity be sought in the functional equivalence or similarity rather than in the specific manners in which languages define and organize grammatical elements.

Bibliography:

1. Bybee J L 1985 Morphology: A Study of the Relation between Meaning and Form. John Benjamins, Amsterdam
2. Chomsky N 1975 Reflections on Language. Pantheon Books, New York
3. Comrie B 1989 Language Universals and Linguistic Typology, 2nd edn. University of Chicago Press, Chicago
4. Croft W 1990 Typology and Universals. Cambridge University Press, Cambridge, UK
5. DeLancey S 1981 An interpretation of split ergativity and related patterns. Language 57: 626–57
6. Gabelentz G von der 1901 Die Sprachwissenschaft. Ihre Aufgaben, Methoden und bisherigen Ergebnisse. Tauchnitz, Leipzig, Germany
7. Gamkrelidze T V, Ivanov V V 1973 Sprachtypologie und die Rekonstruktion der gemeinindogermanischen Verschlusse. Phonetika 27: 15–56
8. Givon T 1984/1991 Syntax: A Functional-Typological Introduction. 2 Vols. John Benjamins, Amsterdam
9. Greenberg J H 1966 Some universals of grammar with particular reference to the order of meaningful elements. In: Greenberg J H (ed.) Universals of Grammar, 2nd edn. MIT Press, Cambridge, MA
10. Greenberg J H 1969 Some methods of dynamic comparison of linguistics. In: Puhvel J (ed.) Substance and Structure of Language. University of California Press, Berkeley, CA
11. Greenberg J H 1974 Language Typology: A Historical and Analytic Overview. Mouton, The Hague, The Netherlands
12. Greenberg J H 1995 The diachronic typological approach. In: Shibatani M, Bynon Th (eds.) Approaches to Language Typology. Clarendon Press, Oxford, UK
13. Greenberg J H, Ferguson C A, Moravcsik E A (eds.) 1979 Universals of Human Language. 4 vols. Stanford University Press, Stanford, CA
14. Hagege C 1988 Contribution des recherches typologiques. In: Joly A (ed.) Linguistique genetique. Historie et theorie. Presses Universitaire de Lille, Lille, France
15. Haiman J 1985 Natural Syntax: Iconicity and Erosion. Cambridge University Press, Cambridge, UK
16. Hawkins J 1990 A parsing theory of word order universals. Linguistic Inquiry 21: 223–261
17. Hawkins J, Cutler A 1988 Psycholinguistic factors in morphological asymmetry. In: Hawkins J (ed.) Explaining Language Universals. Blackwell, London
18. Heine B, Claudi U, Hunnemeyer F 1991 Grammaticalization: A Conceptual Framework. The University of Chicago Press, Chicago
19. Hopper P, Thompson S A 1980 Transitivity in grammar and discourse. Language 56: 251–99
20. Hopper P, Traugott E 1994 Grammaticalization. Cambridge University Press, Cambridge, UK
21. Jakobson R 1958/1971 Typological studies and their contribution to comparative-historical linguistics. In: Selected Writings. Mouton, The Hague, The Netherlands, Vol. 1
22. Keenan, E, Comrie B 1977 Noun phrase accessibility and universal grammar. Linguistic Inquiry 8: 63–99
23. Lehmann C 1982 Thoughts on grammaticalization. A programmatic sketch. Arbeiten des Kolner Universalien-Projekts 48
24. Ramat P 1987 Linguistic Typology. Mouton de Gruyter, Berlin
25. Sapir E 1921 Language. Harcourt, Brace and World, New York
26. Seiler H 1995 Cognitive-conceptual structure and linguistic encoding: Language universals and typology in the UNITYP framework. In: Shibatani M, Bynon Th (eds.) Approaches to Language Typology. Clarendon Press, Oxford, UK
27. Sgall P 1995 Prague school typology. In: Shibatani M, Bynon Th (eds.) Approaches to Language Typology. Clarendon Press, Oxford, UK
28. Shibatani M (ed.) 1976 The Grammar of Causative Constructions (Syntax and Semantics 6 ). Academic Press, New York
29. Shibatani M 1985 Voice in Philippine languages. In: Shibatani M (ed.) Passive and Voice. John Benjamins, Amsterdam
30. Shibatani M, Bynon Th (eds.) 1995 Approaches to Language Typology. Clarendon Press, Oxford, UK
31. Wierzbicka A 1992 Semantics, Culture, and Cognition. Oxford University Press, Oxford