The concept of markedness was first developed in the Prague School of linguistic theory. The notion of marked and unmarked values of a category was first developed for phonological systems by Trubetzkoy (1931; 1939/1969) and first applied to morphosyntactic categories and semantics by Jakobson (1932/1984; 1939/1984; see Greenberg 1966b:11). Markedness has since been adopted by both the generative and the typological approaches to linguistic theory, not surprisingly in rather different ways. As a consequence, markedness in generative grammar is considerably different from markedness in typology (compare Battistella 1996). In fact, in adapting the concept of markedness to cross-linguistic universals, Greenberg (1966b) introduces significant theoretical innovations to markedness (Croft 1996). For this reason, we will use the rather cumbersome locution typological markedness in this book.

Like implicational universals, typological markedness is a fundamental concept underlying much contemporary work in typology, even though it is not overtly referred to very often. Much current typological work is supported by typological markedness (see chapters 5–7). Also, the phenomena described as typological markedness represent an important manifestation of the interplay between two major competing motivations, economy and iconicity, in linguistic expressions. Finally, typological markedness plays a significant role in an influential model of morphological representation, that of Bybee and her associates, which in turn is closely associated with recent developments in syntactic representation, particularly construction grammar (see, for example, Bybee and Thompson 1997; Croft 2001).

Typological classification is the process of describing the various linguistic types found across languages for some grammatical parameter, such as grammatical number or the formation of relative clauses. Typological classification is historically the first manifestation of typology in modern linguistics, starting with the morphological classification of languages in the nineteenth century. The notion of a linguistic type has changed somewhat since that time, particularly under the impact of structural linguistics (the term ‘typology’ was first used in linguistics in 1901; Gabelentz 1901/1972:481). The following section will describe the current concept of a linguistic type, or strategy as it is sometimes called, while the concluding section will discuss morphological typology and the major conceptual changes that have occurred in the evolution of the concept of a linguistic type.

The usual procedure for initiating a cross-linguistic comparison of a particular grammatical phenomenon for the purposes of a typological analysis is to survey the range of structures used for the phenomenon in question. In morphosyntax, the phenomenon is generally a grammatical construction, defined on an external basis precisely because of the degree of structural variation actually found in languages (see §1.3). Thus, given a particular external definition of a category, such as that proposed for the relative clause, one may then classify the linguistic structures found across languages to express or manifest that external definition. These structures are called types or strategies. This is typology in the first sense, a cross-linguistic structural classification of morphosyntactic phenomena.

This volume is an introduction to the concepts and methodology of linguistic typology. It complements other introductory volumes on typology, particularly Comrie 1989 and Mallinson and Blake 1981, in that the material is organized by theoretical concept (implicational universal, markedness, prototype) rather than by topic area (word order, grammatical relations, relative clauses, animacy). Also, the range of concepts covered is somewhat broader, mostly because of the need to describe developments in functional–typological explanation and diachronic typology in the last decade. Needless to say, there is some overlap with the aforementioned volumes. From a pedagogical point of view, however, this volume is intended to complement, not supplement, the more topic-oriented introductions. In particular, breadth in theoretical coverage has meant that detailed examples of typological generalizations, complete with qualifications, possible counterexamples and explanations for those counterexamples, could not always be included (though I have tried not to oversimplify examples without at least citing more detailed studies). The material in this volume has been used in courses in conjunction with Comrie 1989, Greenberg 1966a (the original article on word order), Greenberg 1966b (the monograph on markedness) and other articles on more specific topic areas.

I believe that an essential part of any linguistics class, and above all any class on typology, is for the student to encounter one or more ‘exotic’ languages. For practical reasons, in an introductory typology class this encounter must be somewhat limited.

Languages do not occur in static or stable states. All languages exhibit some degree of grammatical variation, and they change over time; in fact, much synchronic variation represents language change in progress. Changes in linguistic structure are changes in the grammatical properties that enter into one or more of the cross-linguistic patterns described in the preceding chapters. This fact suggests two extensions of typology. First, if language types fall into universal cross-linguistic patterns, then it is worth investigating if the cross-linguistic patterns also govern changes in language type. Second, it should be possible to classify typologically the linguistic changes themselves, and look for relationships among linguistic processes in the same way that typologists seek relationships among linguistic states. This latter process has been especially fruitful and has led to intensive study of a family of associated language processes called grammaticalization.

The dynamicization of synchronic typology

We may return to the foundations of synchronic typology in order to see its consequences for language change. Synchronic typology is founded on the typological classification of logically possible linguistic types (chapter 2). Once this is established, one then constructs a language sample in order to determine which of the logically possible language types are actually attested in the languages of the world, and formulates universals (implicational universals, typological markedness, prototypes, conceptual distance, formal complexity, etc.) that restrict human languages to the actually occurring types. These universals are then explained in external terms.

In chapter 1, we described the three types of typological analysis – typological classification, typological generalization and (functional–)typological explanation – as the three stages in the process of doing empirical science. We also noted that these three stages are interleaved in all scientific practice, typology included. In the subsequent chapters, we made use of this by discussing explanations along with the typological universals (generalizations) that were described in those chapters.

Typological generalization presupposes a basis of cross-linguistic comparison, and in §1.4 we argued that the basis for cross-linguistic comparison is ultimately external, that is, semantic and discourse function for morphosyntax, and phonetic reality for phonology. Thus, the types of universals that one can identify through cross-linguistic comparison are universals of the relationship between linguistic form (morphosyntactic or phonological) and external function or reality. This is the sense in which the typological approach to grammar is functionalist.

The functional–typological approach does not eliminate linguistic form from analysis; far from it. The functional–typological approach does not deny the existence of arbitrariness in grammar (Croft 1995a:504–09). It takes the existence of variation across and within languages as evidence that grammatical structures result from an interplay of convention (which is arbitrary; Lewis 1969) and functional motivation. In fact, typology is more inclined to accept the existence of arbitrariness in language than formalist approaches which seek to explain virtually every idiosyncratic fact, or (nontypological) functionalist approaches that attempt to do the same thing (see §§1.3, 9.2).

The conceptual categories that we have seen to be organized into typological markedness and hierarchy patterns do not occur in isolation. Any given noun phrase has values for number, case and animacy/gender, for example, and any given verb has values for tense, aspect, modality and so on. In other words, conceptual categories always occur in combination in utterances. Hence, it is reasonable to examine the possibility of grammatical interactions between conceptual categories, and to seek typological patterns in those interactions. This chapter will explore some of these patterns.

In fact, we have already seen one instance of interaction of categories. Behavioral potential was described in §4.1.2 as the potential for a category value to inflect for values of other categories, or to have a distribution in a wider range of syntactic contexts. It was assumed that the typological markedness ranking of values of a category would be the same no matter what potential is examined. But in fact the rankings can and do vary depending on what other category is being used for measuring behavioral potential. These are illustrated in §6.1

This is not the only type of interaction between categories, however. The second type is an interaction of values in two cross-cutting categories. Here, a value in one category is typologically unmarked only if it co-occurs with a particular value of another category. The values on the two categories form a cluster or typological prototype.

What is typology?

The term typology has a number of different uses, both within linguistics and without. The common definition of the term is roughly synonymous with ‘taxonomy’ or ‘classification’, a classification of the phenomenon under study into types, particularly structural types. This is the definition that is found outside of linguistics, for example in biology, a field that inspired linguistic theory in the nineteenth century.

The most unassuming linguistic definition of typology refers to a classification of structural types across languages. In this definition, a language is taken to belong to a single type, and a typology of languages is a definition of the types and an enumeration or classification of languages into those types. We will refer to this definition of typology as typological classification. The morphological typology of the nineteenth and early twentieth centuries is an example of this use of the term. This definition introduces the basic connotation that the term typology has in contemporary linguistics: typology has to do with cross-linguistic comparison of some sort. Methodological issues in cross-linguistic comparison will be discussed in §§1.3–1.6, while chapter 2 will be devoted to the notion of a linguistic type, including morphological typology, and its refinements in twentieth-century research.

A second linguistic definition of typology is the study of patterns that occur systematically across languages. We will refer to this definition of typology as typological generalization. The patterns found in typological generalization are language universals. The classic example of a typological universal is the implicational universal.

Restrictions on possible language types

The first step beyond typology as the classification of types and toward the explanation of the cross-linguistic variation that classification describes is the discovery of restrictions on possible language types. Linguistic theory in any approach, formalist or functional–typological, has as its central question, what is a possible language (§1.2)? This question can in turn be paraphrased as: of the logically possible types of languages, how do we account for what types actually exist?

One of the features that distinguishes the typological method of discovering constraints on possible language types is the empirical method applied to the problem. If a typologist wants to find restrictions on possible relative clause structures, for example, he or she gathers a large sample of languages and simply observes which of the possible relative clause types are present and which are absent. That is, the restrictions on logically possible language types are motivated by the actually attested language types. If there is a gap in the attested language types, then it is provisionally assumed that the gap represents a constraint on what is a possible language, and explanations are sought for the gap. This is the inductive method, which must be used in constructing generalizations from empirical data. In contrast, the generative approach uses a rationalist deductive method, in which it is argued that certain analyses of a single language represent universals of human language because they cannot possibly be learned by a child (the ‘poverty of the stimulus’ argument; §1.2).

Grammatical hierarchies and implicational universals

In §3.3, we observed that the implicational universals for noun–adjective order were all of the form ‘X ⊃ NA’, and that this pattern of implicational universals represents the phenomenon of dominance of NA order. There is another even more frequent pattern found in implicational universals, which is taken to represent another type of deeper phenomenon, a grammatical hierarchy.

If one examines languages with prepositions, the following pattern in implicational universals for noun modifiers are found (Hawkins 1983:75; Dryer 1992b):

1. (1) Prep ⊃:

2. (1a) NNum ⊃ NDem

3. (1b) NDem ⊃ NA

4. (1c) NA ⊃ NG

5. (1d) NG ⊃ NRel

The series of implicational universals in 1a–d represents a chain in which the implicatum of each universal is the implicans of the subsequent universal in the list. This pattern leads to a sequence of grammatical types ranked by their position in the chain of implicational universals. This sequence is called a grammatical hierarchy. Hawkins christened this particular hierarchy the Prepositional Noun Modifier Hierarchy, represented as follows (the > symbol is oriented in the same direction as the implication ⊃ in 1a–d):

1. (2) Prepositional Noun Modifier Hierarchy: NNum > NDem > NA > NG > NRel

The chain of implicational universals in 1a–d can be reformulated to make direct reference to the Prepositional Noun Modifier Hierarchy in 2:

1. (3) If a modifier–noun order on the Prepositional Noun Modifier Hierarchy occurs in a prepositional language, then any order to the right also occurs in the language. If the opposite order to one on the Prepositional Noun Modifier Hierarchy occurs in a prepositional language, then any order to the left also occurs in the language.