Introduction: the typological and generative approaches to language universals

Typology and Universals

Typology represents an approach to the study of linguistic structure that differs in certain important respects from the generative and the functionalist approaches. The most important difference between typology and these other approaches to linguistic structure is that the typological approach is fundamentally crosslinguistic in nature. A formalist linguist can analyze a single language in the search for universals of language structure. There are formalist analyses of many different individual languages, but relatively few crosslinguistic formalist studies (a notable exception is the work of Mark Baker, e.g. Baker 2003; but see Croft 2008). A functionalist linguist can analyze a single language in the search for universals of the relationship of language structure to language function. Some functionalist theories are more extensively supported by crosslinguistic data, notably Functional Grammar (Dik 1997) and Role and Reference Grammar (Van Valin and LaPolla 1997). These theories recognize the variation in grammatical structures across languages, but their frameworks generally focus on what is believed to be common to all languages (see §5).

Typology, on the other hand, is fundamentally comparative. A genuinely typological

analysis of a grammatical construction, or a phonological pattern, or other aspects of

language, examines the variation across a large number of languages. In this respect,

typology resembles comparative historical linguistics. The goals of typology and comparative historical linguistics are very different, however, although the results of each are essential to the other (see §2). Comparative historical linguistics seeks genetic (family tree) relationships among languages, in order to discover the history of the languages and their speech communities (Joseph, this volume). Typology examines a broad sample of languages in order to discover universals of language structure and propose explanations of those language universals.

For this reason, typology is linked to language universals. For some linguists, typology simply means the description of variation, that is, how languages differ in their structure. For example, a simple descriptive typology of the word order of numeral and noun would divide the languages of the world into three broad types:

•Those in which numerals normally precede the noun they modify, as in English (Indo-European) two women;

•Those in which numerals normally follow the noun they modify, as in Ma’di (Nilo-Saharan, Uganda-Sudan) à g O@ su@ ‘men four’;

•Those in which numerals may either precede or follow the noun they modify, as in Wardaman (Australian, Australia) guyaminyi mulurruwuyayi ‘two old.women’ or marluga lege ‘old.man one’

However, typology in the linguistic sense is more than a classification of how

languages differ in their structure. A descriptive typology leads to generalizations that

constrain how much languages can vary; those generalizations are language universals.

Several examples of language universals will be given in this chapter.

The search for language universals is shared by typology and generative grammar.

However, the language universals derived from typological research are quite different from those derived in generative grammar, although the generative and typological approaches arose at around the same time. Although the belief in language universals has considerable modern currency, it is by no means a necessary fact or universally-held opinion, and in fact the opposite view was widely held in American linguistics until around 1960. To a considerable degree, the difference between the generative and typological approaches to language universals can be traced to the different traditions to which their respective founders, Noam Chomsky and Joseph Greenberg responded. The generative approach represents a reaction against behavioristic psychology, while the typological approach represents a reaction against anthropological relativism.

The behaviorist view of language, in particular language learning, is anti-universalist

in that it posits no innate, universal internal mental abilities or schemas. In the behaviorist view, linguistic competence is acquired through learning of stimulus-response patterns. In contrast, the generative approach posits the existence of innate internal linguistic abilities and constraints that play a major role in the acquisition of language. It is these constraints that represent language universals in this approach. The argument used by Chomsky (e.g., Chomsky 1976) for the existence of innate universal linguistic competence refers to the ‘poverty of the stimulus’. It is argued that the child has an extremely limited input stimulus, that is, the utterances that it is exposed to from the mother and other caregivers.

This stimulus is incapable of permitting the child to construct the grammar of the adult’s language in a classic behaviorist model; therefore, the child must bring innate universals of grammatical competence to bear on language acquisition. Hence the primary focus on universals in the generative tradition has been on their innate character.

The anthropological relativist view of language is that the languages of the world can

vary arbitrarily: ‘languages could differ from each other without limit and in

unpredictable ways’, in a famous quotation from the linguist Martin Joos (Joos 1957: 96).

This view of language was particularly strong among anthropological linguists studying North American Indian languages, which indeed differ radically in many ways from socalled Standard Average European languages. However, the comparison of one “exotic” language or a limited number of languages to English only indicates diversity, not the range of variation, let alone limits thereto. Greenberg and others discovered that a more systematic sampling of a substantial number of languages reveals not only the range of variation but constraints on that variation. Those constraints demonstrate that languages do not vary infinitely, and the constraints represent language universals. Hence the primary focus on universals in the typological tradition has been on their cross-linguistic validity, and on universals that restrict language variation (see §5).

The innate universals posited by generative grammar are intended to explain

linguistic structure. The poverty of the stimulus argument is essentially a deductive

argument from first principles (although it does make assumptions about the nature of the empirical input, and what counts as relevant input). The poverty of the stimulus argument is one aspect of Chomsky’s more generally rationalist approach to language. The universals posited by typology are intended to represent inductive generalizations across languages, in keeping with typology’s empiricist approach to language. The generative grammarian argues that the discovery of innate principles that the child brings to bear in learning a single language can be extrapolated to language in general (Chomsky 1981).

The typologist argues that a grammatical analysis based on one language or a small

number of languages will not suffice to reveal language universals; only a systematic

empirical survey can do so.

The typological approach to discovering language universals, like any empirical

scientific approach, is basically inductive. In order to discover what language in general is like, or the universal character of language, one should look at a large number of languages. For example, consider the two English examples in (1)-(2):

(1) They talked about the war.

(2) What did they talk about?

The question in (2) differs from the statement in (1) in three ways: the interrogative

pronoun corresponding to the war in (1) is found at the beginning of the question; the

preposition about nevertheless remains “stranded” in its position after the verb; and the auxiliary verb did occurs, positioned before the subject. Looking only at English, one cannot tell what the significance of these three differences are. In fact, the initial position of the interrogative pronoun is extremely widespread across the world’s languages, though by no means universal. In contrast, the stranding of the preposition and the insertion of an auxiliary that is absent in the corresponding statement are extremely rare and idiosyncratic grammatical traits of English, and not causally connected to each other or to the initial position of the interrogative pronoun. It is only by examining a broad sample of languages that the significance of different properties of linguistic structure can be assessed.

Likewise, one can only judge whether a causal connection exists between two

grammatical properties by examining a large and widely distributed sample of languages.

For example, languages differ in the constructions used for nonverbal predication.

English requires a copula (in boldface in (3)) for nonverbal predication at all times,

whereas Russian lacks a copula in the present tense:

(3) She is a doctor.

(4) Ona vrač.

She doctor

A common explanation for the absence of a copula, or zero copula, in a variety of

theories is that the reason for the occurrence of a copula is a “need” to place verbal

inflections that cannot occur on nonverbal predicates such as the nouns in (3)-(4) (see

Stassen 1997: 66 for the history of this proposal, and also Baker 2003: 40). Zero copulas

therefore are claimed to occur when verbs lack inflection or when the inflection is zero; nonzero inflections require a “dummy” copula to carry them. Thus the zero copula occurs in Russian present tense because that is the “unmarked” inflectional category.

This hypothesis appears to be a plausible one, assuming that the purpose of the copula

is to carry verbal inflections for nonverbal predicates. It appears to be valid for English, Russian, and a number of other languages. However, Stassen demonstrates that in fact it is not crosslinguistically valid in general (Stassen 1997: 65-76). There are many languages with overt inflections but a zero copula, such as Sinhalese (Indo-European, Sri Lanka; see the Appendix for abbreviations of grammatical elements in the examples):

(5) mahattea e-nә w-a

gentleman come-NPST-IND

‘The boss comes/will come.’

(6) mahattea a-aw-a

gentleman come-PST-IND

‘The boss came.’

(7) unnæ hee hungak prә siddә kene-k

3SG.M very famous person-NOM

‘He is/was a very famous person.’

And there are many languages with (at least some) zero inflections but always an overt copula (translated as COP in example 9), such as Cambodian (Austroasiatic, Cambodia):

(8) vì: ә tɤ u phsa: r

he go market

‘He goes/went/will go to market.’

(9) mә n-s nù h cì: ә kru:

man that COP teacher

‘That man is a teacher.’

Instead, the universal governing occurrence of a copula is the semantic type of

predication: if predicate adjectives (predication of a property) requires a copula in a

language, then predicate nominals (predication of object class) requires a copula as well (Croft 1991: 130; Stassen 1997: 127).