Abstract and Keywords
This introductory article discusses the theme of this volume, which is about grammaticalisation. It explains that grammaticalisation is considered a young sub-field of linguistics because the term was coined only in 1912 by Antoine Meillet and provides background information on modern studies in grammaticalisation which began in the early 1970s. This volume analyses grammaticalisation in relation to linguistic theory and discusses the domains of grammaticalisation. It also explores the grammaticalisation of form classes and categories and describes the different faces of grammaticalisation across languages.
Grammaticalization is believed to be a young sub‐field of linguistics. As a matter of fact, however, it is almost as old as linguistics, even if the term was presumably coined only in 1912 by Meillet.1 Many of the issues figuring in contemporary discussions on grammatical evolution were already discussed by German 19th‐century linguists such as Bopp (1816; 1833), Wüllner (1831), or von der Gabelentz (1961).
Modern studies in grammaticalization began in the early 1970s with the work of Givón, who argued that in order to understand language structure one must know how it has evolved. With his slogan ‘Today's morphology is yesterday's syntax’, he opened a new perspective for understanding grammar (Givón 1971: 12; 1979; see below). The first monographic treatments of grammaticalization were Lehmann (1995a) and Heine and Reh (1982). But perhaps a milestone in the history of modern grammaticalization studies can be seen in the symposium that Givón organized at the University of Oregon in 1988, resulting in two volumes on the topic (Traugott and Heine 1991a; 1991b). The two textbooks by Heine, Claudi, and Hünnemeyer (1991) and Hopper and Traugott (1993) then cemented the status of grammaticalization as an independent field of study within linguistics. Of similar importance to the Oregon symposium is the series of bi‐annual conferences that Wischer initiated in Potsdam in 1999 and the publications resulting from this (p. 2) meeting (Wischer and Diewald 2002), as well as from subsequent meetings (Fischer, Norde, and Perridon 2004; López‐Couso and Seoane 2008).
Since roughly the beginning of this century, grammaticalization studies have entered a new phase of development. On the one hand, they were subject to serious criticism (especially Newmeyer 1998; Campbell 2001a; Joseph, Chapter 16 below); on the other hand, they experienced an enormous expansion. Having been restricted primarily to core grammatical, semantic, and pragmatic analysis in the 20th century, they now attract interest in a wide array of related fields of linguistics, such as corpus linguistics, phonology, language acquisition, and sociolinguistics. Furthermore, while grammaticalization was initially practically exclusively the domain of functionally oriented scholars, it has increasingly been recognized as an important research topic by formal linguists as well. Moreover, grammaticalization research has spread beyond the traditional centres of linguistics to regions such as East Asia and South America. Facing this increasing expansion and diversification, we as editors believed that now would be a good point in time to take stock of the current state of grammaticalization studies, and simultaneously uncover possible directions for future research in this field.
Currently a wide range of approaches and theoretical orientations are in some way or other based on a grammaticalization perspective. This diversity is associated with a variety of different views on how this phenomenon should be defined. Going through the chapters of this volume, the reader will notice that grammaticalization is far from being a uniform concept, and various definitions have been proposed.
One kind of definition relies on pragmatic functions of linguistic material. Harder and Boye, for example, invoke the notion of competition for discourse prominence, and propose to define grammaticalization as ‘diachronic change which gives rise to linguistic expressions which are coded as discursively secondary’ (Chapter 5). And Nicolle concludes that what defines grammaticalization is the addition of procedural information to the semantics of an expression. In his approach, lexical items encode conceptual information, while discourse markers, pronouns, and tense, aspect, and modality markers encode procedural information (Chapter 32). Another aspect of grammaticalization concerns the frequency of use of linguistic material. In some of the definitions provided, frequency is portrayed as (p. 3) one of the driving forces, or the driving force of grammaticalization (see especially Chapter 6 by Bybee). We will return to this issue below.
On the other hand, there is also the view that grammaticalization concerns anything that relates to grammar. For Frajzyngier, for example, the term stands simply for any coding of a function within the grammatical system of a language (Chapter 51). Depending on which definition is employed, there are great differences with respect to the phenomena to be considered. In extremely general definitions, such as that proposed by Frajzyngier, for example, diachrony is not a major issue, and the ‘sources’ of grammaticalization are not restricted to lexical and other form–meaning units but also include tone, intonation, phonological changes affecting segments, linear order, and position. Still, when controversies arise many scholars agree in draw attention to the classic definition by Kuryłowicz to help settle the issue of what should be subsumed under the rubric of grammaticalization:
Grammaticalisation consists in the increase of the range of a morpheme advancing from a lexical to a grammatical or from a less grammatical to a more grammatical status, e.g. from a derivative formant to an inflectional one. (Kuryłowicz 1975: 52)
For most students of the field, grammaticalization is understood to be a diachronic process and, hence, findings can be verified or falsified by means of historical evidence. But it is also possible to analyse grammaticalization phenomena within a synchronic framework. This is demonstrated in particular by Langacker in his Cognitive Grammar account of a range of instances of grammaticalization (Chapter 7). As this volume suggests, there is no single approach or model that is predestined more than others to deal with grammaticalization phenomena, or that would account for all phenomena better than any other approach. To be sure, the questions asked differ from one approach to another and the answers given to central questions are not the same across the different approaches; but these answers are in most cases compatible with one another.
2. Delimiting the field
Each approach or ‘school’ of linguistics has its preferences as to the kind of linguistic phenomena that it is concerned with, and with respect to the way that it demonstrates its strengths and the advantages it offers over alternative approaches. Studies in grammaticalization also have their preferences. One noteworthy preference appears to be working on English and employing the English be‐going‐to future as a paradigm case of grammatical change. In a number of (p. 4) chapters, especially that by Bisang (Chapter 9), it is argued, implicitly or explicitly, that there is a need to take account of the typological diversity of the world's languages, more so than has been done in the past. This call for more diversity in the object of research is partially reflected in the design of this handbook, especially in Part V, which contains articles on a wide variety of languages, also outside the Indo‐European area.
The flip side of the question of how to define the phenomenon is of course: what counts as an instance of grammaticalization and what does not? One of the areas where this question has been hotly debated is that of discourse markers or particles (see e.g. Onodera in Chapter 50). Can they, or at least part of them, be described exhaustively within the framework of grammaticalization theory? Or, is a separate framework of ‘pragmaticalization’ required, as has been argued ever since Erman and Kotsinas (1993; Aijmer 1997: 2) proposed this term? Is it desirable to draw a boundary between ‘sentence‐grammatical phenomena’, to be treated under the rubric of grammaticalization, and ‘discourse‐pragmatic phenomena’, which are the subject matter of pragmaticalization studies (Günthner and Mutz 2004)? Diewald (Chapter 36) argues that it is possible to treat pragmaticalization as a sub‐process of grammaticalization. Note that already in 2000, Wischer (2000: 359) had proposed to treat the two as subtypes, referring to pragmaticalization as ‘grammaticalization on the text or discourse level’ and to orthodox grammaticalization as ‘grammaticalization on the propositional level’. Both processes have in common that language material undergoes recategorialization by changing from a more open to a closer categorial system. It is therefore obvious that grammaticalization theory provides a principled tool to bridge the boundary between two domains of linguistic analysis that tend to be treated as distinct, namely grammar and pragmatics. This is a point also brought home in much detail in a recent book publication by Ariel on the topic (2008).
More general, and this is an issue that comes up in a number of chapters, is the question of where the limits of grammaticalization lie. For example, is grammaticalization restricted to oral and written languages, or does it show up in other modalities of human communication as well? As Pfau and Steinbach show in Chapter 56, the behaviour of grammaticalization in sign languages is largely similar to that in oral languages. To be sure, there are modality‐specific differences. For example, in both kinds of modalities there are auxiliaries. However, whereas in spoken and written languages there is a major pathway from lexical verbs to the functional categories of tense, aspect, and modality (Bybee, Pagliuca, and Perkins 1994), in sign languages it is not only verbs but also nouns and pronouns that may give rise to auxiliaries.
There is reason to assume that grammaticalization most commonly arises in spoken, rather than in written language use. However, as Narrog and Ohori show Chapter 64, Japanese provides a number of examples where grammaticalization was triggered by the written rather than the spoken language, especially via translation. (p. 5) Strikingly similar developments have also been reported from European languages (see Heine and Miyashita 2008b).
On a more fundamental perspective, there are general human strategies, or mechanisms, that have been invoked for describing, delimiting, and understanding grammaticalization, such as analogy, reanalysis, generalization, or creativity; see especially Traugott (Chapter 2) and Fischer (Chapter 3). Among these conceptual mechanisms, reanalysis is one of the most frequently cited. But the significance of this notion has been challenged. Fischer, for example, argues that reanalysis is not something that speakers or hearers do. Rather, it is a concept of the analyst that is, at least with reference to language processing, being based on our ability to analogize (Chapter 3). Humans are analogical animals, as Anttila (2003: 438) puts it, but they also reanalyse the material they dispose of. They generalize, and they use linguistic forms and constructions creatively for novel purposes. The question then is: to what extent are these notions helpful for understanding or for defining grammaticalization, and, is any of these more relevant than others? Many different answers are volunteered in the following chapters, reflecting the conceptual diversity that characterizes the field of grammaticalization.
3. Central issues
Since the late 1990s, studies in grammaticalization have been the subject of critical discussions. Perhaps the most serious claim, first made by Newmeyer (1998) and taken up in this volume by Joseph in Chapter 16, is that grammaticalization is not a distinct process but merely represents a combination of independent linguistic processes (see also Campbell and Janda 2001, as well as the other contributions to Language Sciences 23.2–3). Another problem concerns what is most commonly referred to as ‘degrammaticalization’. Central to the problems of defining and delimiting grammaticalization is the question of what to do with what Hilpert calls ‘developmental U‐turns’ and other cases of degrammaticalization (Chapter 58); for some examples, see Narrog and Ohori's analysis of Japanese (Chapter 64). On the basis of detailed analysis, Norde concludes that while changes classified as degrammaticalization challenge the unidirectionality hypothesis, they also lend support to it in affirming it as a strong directional tendency in grammatical change as quantitatively limited exceptions (Chapter 38; see Norde 2009a for more details). That there is need for much further analysis of cases of suspected degrammaticalization, and more generally of degrammaticalization as such, is shown convincingly in Chapter 14 by Börjars and Vincent.
(p. 6) It may be useful to distinguish between two kinds of approaches to grammatical evolution. On the one hand there are approaches that focus on the initial phase leading from non‐grammatical, lexical structures to grammatical, non‐lexical structures. On the other hand there are also approaches that concentrate on a more advanced phase of the process relating to bound, typically inflectional structures, and the development of further advanced and abstract grammatical functions. An overview of the findings presented in this volume suggests that the kind of generalizations proposed are not the same, depending on which of the phases is highlighted by a given author (see e.g. the discussion in Chapter 5). Another issue that comes up in many chapters concerns the motivation(s) of grammatical change, and here a wide spectrum of views are voiced. At one end are adherents of schools of functional linguistics invoking discourse pragmatic and/or semantic principles. At the other end are students of generative models who tend to hold innate principles in children responsible for grammaticalization (see especially Chapter 4 by van Gelderen). What the two have in common is that both assume that, across languages, grammatical change is directional, leading, for example, from lexical to functional categories or structures, Hence there must be universal principles underlying them. Grammaticalization and generative grammar have had, as van Gelderen observes, ‘an uneasy relationship’, but due to the introduction of functional categories in the late 1980s and features in the 1990s, it has become possible to account for gradual unidirectional change in a generative framework.
Another central topic of linguistic theory concerns the nature of linguistic categories, and this is an area where the contribution of grammaticalization studies may have been of particular importance. When Ramat observes in Chapter 40 that it is not always easy to distinguish morphologically between adverbs and nouns or adjectives, or between adverbs, prepositions, and conjunctions, this points to an area where work on grammaticalization has come up with a range of new findings: clines (Hopper and Traugott 2003) or chains of grammaticalization (Heine 1992) are some of the constructs that have been proposed to describe and account for the overlapping nature of syntactic or morphological categories. That grammatical forms and constructions are best analysed as gradient categories is suggested in a number of chapters; Brinton, for example, presents evidence for a gradience view of lexicality and grammaticality in her discussion of English complex predicates (Chapter 45), and Krug concludes in his analysis of auxiliation and categorization in the domain of tense, aspect, and modality that ambiguous cases are the norm rather than the exception and that the borderline between lexical and grammatical items will always remain arbitrary to some extent (Chapter 44).
Syntax and morphology are in many theoretical frameworks of linguistics treated as phenomena belonging to distinct domains of analysis; still, it is well known to students of grammaticalization that it is hard to trace a clear boundary between the two. But the problem is even more serious than has previously been thought. (p. 7) Haspelmath argues in Chapter 27 that, ‘the non‐coincidence of the various criteria for syntactic vs. morphological status makes the very idea of a syntax/morphology distinction highly doubtful. Combinations of signs have different degrees of tightness, and it is not at all clear that this continuum can usefully be divided into two parts (syntax vs. morphology) or three parts (free words vs. clitics vs. affixes).’ One of the issues raised by Haspelmath, the categorial status of case marking in Hungarian, is also highlighted in König's discussion (Chapter 41): is Hungarian a language without case system or with an extremely rich case system?
This raises the question of whether students of linguistic analysis should decide on models that aim at accounting for the gradual nature of grammatical categories rather than insisting on classical models of discrete categorization in terms of necessary and sufficient criteria. Even if one were to decide on answering this question in the negative, Krug maintains in Chapter 44 that grammaticalization studies are helpful in improving the basis for decisions on where to draw relevant lines between categories.
Two other issues have figured prominently in earlier studies. One concerns the role of iconic coding, which is discussed by Haiman in Chapter 37, leading to the question: does iconicity influence grammaticalization processes? The second issue, one that has now attracted renewed attention, concerns what is commonly known as ‘the linguistic cycle’. That certain linguistic developments are cyclical has been claimed by scholars almost as long as linguistics exists as an independent discipline. In the history of grammaticalization studies, this claim has been put forward in various formats, perhaps the best‐known being Givón's cycle (1971; 1979: 209), reproduced in (1).
(1) Discourse 〉 Syntax 〉 Morphology 〉 Morphophonemics 〉 Zero
To what extent is grammatical evolution cyclical? This question is addressed in several of the chapters. A much‐debated case of cyclicity relates to negation and concerns what is widely known as Jespersen's Cycle; an analysis of this phenomenon, as well as that of a negative‐existential cycle, is discussed by Mosegaard Hansen in Chapter 46. Another kind of cycle concerns the rise and fall of grammatical subject and object agreement, which is van Gelderen's topic in Chapter 39. The same author has recently edited a whole volume on the topic of cyclicity (van Gelderen 2009).
Another feature that has received some attention concerns the behaviour of scope. Does grammaticalization entail a decrease in the scope that the entities concerned experience—hence, can scope be taken to be adopted as a definitional property (cf. Lehmann 1995a)? This issue is addressed in some of the chapters; Hengeveld in particular maintains that the diachronic development of expressions for tense, aspect, and mood leads from lower to higher scope (Chapter 47). The same stance is basically taken in generative grammar, which conceives of grammaticalization as ‘category climbing’, or in terms of Late Merge (p. 8) (cf. Roberts and Roussou 2003; van Gelderen 2004; Chapter 4 below). As has been pointed out however by some authors, for example by Norde in Chapter 38, scope is a problematic parameter. After all, the directionality of change with respect to scope largely depends on the particular notion of scope, and on the domain of grammar that it refers to.
Among the general questions that have so far not received the kind of attention they deserve is the following: how long does a grammaticalization process need to take its course? That such a process does not happen overnight is beyond any reasonable doubt. But what is the minimum and the maximum time required? Studies of pidgins and creole languages suggest that new grammatical categories can arise within less than a century. At the other end, however, there are also examples to show that the evolution of a category can extend over more than a thousand years. As Deutscher shows in Chapter 53, it took nearly two millennia for a fully independent speech‐introducing clause (‘this is what X said’) to grammaticalize into an obligatory quotative marker in Accadian. Furthermore, there is the question of how human languages evolved. Once an issue that was ignored or avoided by linguists for over a century, language evolution has recently become a hotly contested subject matter in some schools of linguistics. Smith in Chapter 12 argues that studies in grammaticalization can make a significant contribution to reconstructing the genesis and development of human language (see also Heine and Kuteva 2007). This is also a central issue in Dahl's discussion of how grammatical change relates to linguistic complexity. Approaching grammaticalization from the vantage point of complexity studies and distinguishing between system complexity and structural complexity, he is able to establish a number of correlations, for example between non‐linearity and high degree of grammaticalization (Chapter 13).
Finally, the volume is also concerned with a question that some might consider central in understanding grammatical change, but one that has also been discussed controversially: how does grammaticalization relate to first language acquisition? Does language acquisition recapitulate the diachronic evolution of grammar, as some have argued, or does grammaticalization originate in changes in child language? As Diessel argues on the basis of solid evidence from both domains, both questions have to be answered in the negative (Chapter 11). The two developments are in principle independent of each other. There is no causal link between them. Children seek to uncover the meanings of existing expressions. Grammaticalization, by contrast, involves the creation of novel meanings. Nevertheless, Diessel concludes that, while morphosyntactic and phonological changes in particular are different in language acquisition and in grammaticalization, the semantic and pragmatic developments of grammatical markers are based on the same mechanisms of categorization. In both cases, they are grounded in general perceptual and cognitive principles of the human mind.
(p. 9) 4. Domains and structures
What is the primary target of grammaticalization processes: meaning, form, or structure? Are the units to which grammaticalization applies lexical or non‐lexical items, constructions, or more generally, collocations of meaningful elements? These questions cannot be decided a priori but are necessarily linked to the particular theoretical framework within which they are raised. The issue of form vs. function is perhaps most relevant when grammatical change is viewed from the perspective of construction grammar. As Gisborne and Patten argue convincingly, the ‘constructional change’ of two constructions looked at in their Chapter 8 shares a number of properties with canonical processes of grammaticalization involving lexical items. Note that, like many (though not all) versions of grammaticalization theory, constructional models assume that lexicon and grammar form a continuum, and that grammatical change is gradual and incremental and leads to an increase in productivity and schematicity. It would seem that one either follows Noël (2006) in maintaining that schematization in constructions and grammaticalization are two different types of change, or one searches for an overarching theoretical framework that encompasses both. The case of the grammaticalization of quotative markers, as presented by Deutscher (Chapter 53) seems to provide support for the second approach. Deutscher suggests that speech‐introducing clauses rather than verbs such as ‘say’ or particles such as ‘like’ are the source material on which the path to the development of quotative markers is constructed. The lexical sources are only relevant in as far as they are used inside such a clause.
The nature of the process from lexical or less grammaticalized to more grammaticalized structures is a topic in many of the following chapters. One salient direction in grammaticalization leads from more concrete to more abstract meanings, as shown, for example, by Eckardt with reference to the emergence of scalar degree modifiers (Chapter 31). That such semantic processes need not be confined to one particular morphological category, such as the verb, is demonstrated by Ziegeler in her analysis of modality, where she argues in favour of what she calls a “more holistic semantic approach” to the study of modality (Chapter 48).
There are certain grammatical categories that time and again can be traced back mainly to one particular conceptual source only, while others derive from multiple conceptual sources. Both kinds of process are represented in this volume. The evolution of definite articles is of the former type, as De Mulder and Carlier show in Chapter 42. Many contributions to this volume observe that for most functional categories there is not just one source but an entire pool of different sources of grammatical development. Evidentials, for example, may not only come from grammaticalized verbs but may also go back to locative and deictic markers or members of other word classes (Chapter 51). The genesis of passive markers and (p. 10) constructions can be due to an even larger range of pathways (see Chapter 43 by Wiemer; see also Haspelmath 1990).
Grammaticalization takes place in discourse, and its most obvious outcome is to be found in morphology. Syntax, by contrast, is a domain that some do not centrally associate with grammaticalization theory. That such a view is in need of revision is demonstrated in a number of chapters. One of them is DeLancey's Chapter 29, where it is shown that grammaticalization theory is one of the two essential components of the functional‐typological approach to syntax. Grammaticalization, DeLancey states, ‘is not simply a mechanism by which morphological structure develops, it is the constant, universal tendency of language out of which all structure arises’. And in fact, for quite a number of students of grammaticalization, syntax is a central field of research. Therefore it is hardly surprising that there is a range of chapters in this volume analysing syntactic phenomena, like word order in Chapter 30 by Sun and Traugott.
A large part of research on grammaticalization relates to the interface area between semantics and pragmatics, and even approaches that focus on semantic issues tend to include a pragmatic component in addition, as can be seen for instance in Chapter 31 by Eckhardt. Of central importance for this issue is the following question, raised especially by Nicolle (Chapter 32): What is the contribution of context as opposed to inferential mechanisms in the rise of new grammatical meanings and constructions?
Another issue raised in a number of chapters concerns the question of whether a given phenomenon really qualifies as an instance of grammaticalization. Otherwise it may be more appropriately treated within some alternative field of analysis, or it may be best analysed as being located at the interface of two or more different fields of study. A paradigm example of an interface area concerns the relationship between grammaticalization and lexicalization. Having long been neglected as a distinct research field, lexicalization attracted considerable research between 2002 and 2005, where the central question was one of delimitation: where does grammaticalization end and lexicalization begin? We now know much more about the different manifestations of lexicalization processes, but, as Lightfoot observes in Chapter 35, the challenge remains for examining lexicalization in relation to grammaticalization.
Another interface area relates to the structure of predication. One of the test cases analysed in this volume concerns complex predicates in English, such as have a drink, make a call, give advice, which have been discussed in terms of lexicalization and idiomaticization. As Brinton shows convincingly, one type of complex predicates that involves the English light verbs make, take, give, have, and do exhibits changes that are characteristic of grammaticalization, being instances of grammaticalized phrasal constructions (Chapter 45).
Clause combining has attracted considerable attention in studies of grammatical change, and most aspects of combining are treated in the present volume. (p. 11) Clause subordination is discussed most of all by Ohori (Chapter 52), but coordination of clauses is also well represented, being in particular the subject of Giacalone Ramat and Mauri's Chapter 54. While there is the issue of how conjunctions and other elements of clause combining may arise, there is also the issue of what can happen further to such elements. Some new lines of research have shown that complementizers and other clause connectives can become final particles, for example, of utterances. Thompson and Suzuki demonstrate in Chapter 55 that this is potentially a cross‐linguistically regular process.
One domain that has, conversely, been somewhat neglected in past work is that of personal deixis. Personal pronouns, and more generally person markers, belong to the most conservative parts of grammar. Most of them are etymologically opaque. That speakers may create more than one new category of personal deixis is shown by Martelotta and Cezario in Chapter 60 on Brazilian Portuguese. Another domain that so far has perhaps not received the kind of attention it deserves is parenthetical constructions. This fairly new research topic appears to have many implications for the study of grammaticalization phenomena. Processes such as the rise of new discourse markers or particles are considered by some to be a test case for defining the limits of grammaticalization. Among the phenomena discussed by Hilpert (Chapter 58) are conjunctions in Germanic languages that come to be used outside their typical syntactic context, and undergo decategorialization. This is manifested in the development of independent intonation, strong restrictions on the initial or final position, and a replacement of earlier grammatical meanings with discourse‐pragmatic functions.
There are many different ways of explaining linguistic phenomena. Yet, when it comes to finding answers to the question of why languages are structured the way they are, grammaticalization studies provide insights that are indispensible for providing a satisfactory explanation. Mithun (Chapter 15) demonstrates how they help in understanding the morphosyntactic structure of extremely complex languages such as Navajo and other Athapaskan languages. Showing that the ordering structure of the Navajo verb structure was built up in stages over time via principles of grammaticalization, she is able to account for a number of morphosyntactic issues that have plagued preceding analyses of this language for decades. She rightly emphasizes that work on grammaticalization cannot replace synchronic language description. At the same time, she also points out that this work may, for example, lead to the conclusion that it is no longer necessary to decide whether a given morpheme is actually ‘lexical’ or ‘grammatical’, or whether subject and objects prefixes are ‘really agreement’ or ‘really pronouns’.
Grammaticalization theory can also shed light on the distinction between polysemy and homonymy. For example, there are languages where one and the same item serves on the one hand as both a passive and a causative marker and on the other hand as a lexical verb for ‘to give’. As Chappell and Peyraube show in Chapter 65, this situation is in no way odd or peculiar. Rather, it can be accounted (p. 12) for with reference to the grammaticalization processes that gave rise to it. Whether such situations should be treated as instances of polysemy or homonymy is a question that is notoriously controversial in linguistics. It is obvious, however, that from the perspective of grammaticalization theory they qualify as instances of heterosemy (Lichtenberk 1991), being the result of polygrammaticalization (Craig 1991), where there was one lexical structure that has given rise to different lines of development. Polysemy, or heterosemy, is in fact an area where grammaticalization studies provide both an important tool of analysis and some explanatory potential, and polysemy is also the focus of research on semantic maps. The question of how grammaticalization paths relate to (synchronic) polysemy as represented in semantic maps is a main topic of Chapter 25 by Narrog and van der Auwera.
In addition to the question of how grammaticalization can contribute to understanding the nature of language structure, there is also the question of what explains grammaticalization itself. While a range of different stances is voiced in this volume, two of them appear to be particularly prominent. On the one hand, there are explanatory approaches associated with what may be called the construction grammar paradigm that invoke frequency of use as one of the main forces, or the main force, driving grammaticalization, if not linguistic change in general. In the tradition of Bybee and Hopper (2001; see also Chapter 6), Torres Cacoullos and Walker, for example, define grammaticalization as the set of gradual semantic and structural processes by which constructions involving particular lexical items are used with increasing frequency and become new grammatical constructions, following cross‐linguistic evolutionary paths (Chapter 18). On the other hand, there are also approaches that highlight the speaker's communicative motivations and the way in which linguistic material is manipulated for finding optimal rhetorical solutions (see Chapter 33 by Waltereit).
One of the main challenges facing students of grammaticalization is the question why grammatical development is, at least to a large extent, unidirectional. Various ways have been proposed to explain unidirectionality. In generative linguistics, unidirectionality has been explained with reference to universal principles such as Late Merge, ultimately relating to the principle of Economy (van Gelderen 2004; Chapter 4). On the other hand, it is argued by Fischer (Chapter 3) that unidirectionality is not something necessarily intrinsic to grammaticalization on the speaker–listener level. In the usage‐based model of Bybee (Chapter 6), it is frequency of use that plays a central role. It is only when increases in frequency spur all the mechanisms to work together, she maintains, that we recognize an instance of grammaticalization: ‘Changes related to increases in frequency all move in one direction and even decreases in frequency do not condition reversals: there is no process of de‐automatization or de‐habituation, subtraction of pragmatic inferences, etc. Once phonetic form and semantic properties are lost, there is no way to retrieve them. Thus grammaticization is unidirectional.’
(p. 13) Rather than simply frequency of use, some students of discourse analysis see conversational routines as being central for the development of unidirectionality. Such routines have the effect that, for example, two or more independent units of language structure or meaning grow together into a single grammatical construction with interdependent, integrated components (see Couper‐Kuhlen in Chapter 34).
5. Studies across the world
Principles of grammaticalization have been claimed to apply to languages across the world irrespective of genetic or areal affiliation, and the question is whether this is appropriately reflected in the present volume. Unfortunately, the answer is not an unequivocal ‘yes’. As in grammaticalization studies in general, there is clearly a bias towards the major languages of the world. English in particular enjoys a privileged status, both in the discussions and in the exemplifications to be found in the chapters. While a number of chapters are devoted primarily to languages of European origin, the linguistically most complex regions of the world are clearly under‐represented. There is only one chapter devoted to the 2,000‐odd African languages (Chapter 57 by Heine), but other regions also showing a remarkable linguistic diversity, such as New Guinea, South America, or Australia, have not found the kind of attention they deserve. The reasons for this are obvious. Grammaticalization studies have traditionally focused on European languages, and to a lesser extent also on languages of Eastern Asia, i.e. Chinese (Chapter 65 by Chappell and Peyraube), Korean (Chapter 63 by Rhee), and Japanese (Chapter 64 by Narrog and Ohori). With respect to research on grammaticalization, languages in these areas are naturally at an advantage in the sense that they are historically relatively well documented. In contrast, relatively little information is available on grammaticalization processes, for example, in Papua New Guinea or Australia.
Work on grammaticalization thus shows a strong bias in favour of a few languages, while in most regions of the world this is a recent and yet underexplored field of study. On the other hand, there are also earlier academic traditions dealing with issues of grammatical development under a different heading, or under different theoretical premises. This is especially, but not only, the case in countries having a long tradition of written language use. In Korea, for example, such studies can be traced back to the 1960s (see Chapter 63). But even on a continent like Africa, studies in grammaticalization meanwhile have a history of roughly thirty years (Chapter 57). A large number of scholars around the globe now devote their work to issues of grammaticalization, thereby contributing to our knowledge of the (p. 14) typological diversity of grammatical change. This is reflected in particular in chapters such as that of Johanson on Turkic languages (Chapter 62), Wiemer on Slavic languages (Chapter 61), or Martelotta and Maura Cezario on Brazilian Portuguese (Chapter 60).
Note that grammaticalization processes usually concern individual categories or constructions of a language rather than languages as a whole. That it is nevertheless possible to determine the profile of grammaticalization for entire languages is claimed by De Mulder and Lamiroy in their comparative study of Romance languages (Chapter 24).
6. New topics and fields
As observed above, grammaticalization has more recently become the target of new fields of analysis. One of those new and promising fields can be seen in prosody. As Wichmann points out in Chapter 26, segmental attrition tends to be seen as a typical feature of grammaticalization, but she finds that it is a partial and secondary phenomenon, while the primary phenomenon is prosodic. The primary effect of frequency and habituation, she suggests, is not segmental attrition, but prosodic erosion or loss of prominence.
Grammatical change begins with individual speakers and affects specific social groups before it spreads to other individuals and social categories of speakers. While this is intuitively clear to students working in this field, work on the individual and the social dimension of grammaticalization has so far not received the attention it deserves. We are therefore glad that both dimensions are being considered in the present volume. That individuals provide the very first occurrences of phenomena that eventually develop into changes in language is pointed out above all in Chapter 20 by Raumolin‐Brunberg and Nurmi. At the same time, these authors observe that processes of grammaticalization tend to be slow, making it difficult to observe them in an individual's linguistic practices over her or his lifetime.
Much of the information on grammaticalization that is available is based on the analysis of standard languages or linguistic systems that are portrayed as being fairly uniform, while there is little information on how grammatical evolution relates to dialectal and demographic variation. Impressive insights into this issue can be found in recent sociolinguistic work, as discussed by Nevalainen and Palander‐Collin in Chapter 10, or in the treatment of English non‐standard varieties by Kortmann and Schneider (Chapter 21).
(p. 15) A question that is of interest in any academic discipline but that appears to be particularly relevant to grammaticalization studies is the following. What counts as evidence to support one's hypotheses and generalizations? Since grammaticalization is a diachronic process, evidence should first and foremost consist of historical ‘facts’. In this respect, students working on languages for which substantial written records on their earlier stages of development exist are in an ideal position. This becomes especially clear in the contributions on languages such as Chinese, as Chappell and Peyraube show in Chapter 65, or Ledgeway in his contribution on Latin and the Romance languages (Chapter 59). Languages without any written documents offer a less enviable prospect for finding appropriate empirical evidence.
Larger text samples and quantitative approaches are increasingly valued in the search for appropriate evidence. This development is reflected in many of the chapters. Clearly, research in grammaticalization increasingly relies on methods of analysis that allow for quantitative generalizations, most of all on corpus linguistics. Here the motto is: the larger the corpus is, the more likely that it will allow for a comprehensive analysis of the dynamics of linguistic change. Mair observes, for example, that small corpora may be sufficient to study grammaticalization of high‐frequency core‐grammatical categories, but they are insufficient when it comes to rarer phenomena, such as certain types of clausal subordination (Chapter 19).
Use of quantitative data, though, already has a distinguished tradition outside grammaticalization studies, concerning topics that are nevertheless relevant to grammaticalization. This is the case, for example, in work carried out by students of variation theory that seek to explain why one form is chosen over another to signal the same meaning or function in a given context. As Poplack shows in Chapter 17, variation theory can shed light on ongoing processes that are not within the scope of orthodox grammaticalization theory. Language change is commonly classified into whether it takes place entirely within a given language (i.e. internal change) or is influenced or caused by contact with other languages (external change). Grammaticalization, then, tends to be viewed as a paradigm case of internal language change. As more recent research has demonstrated, however, this view is in need of reconsideration. A large body of data shows that grammaticalization can be induced by language contact. A striking example is provided by the domain of evidentiality. For example, Aikhenvald observes that language contact and areal diffusion provide a major motivation for developing an evidentiality system and, consequently, grammaticalized evidentials are a feature of many linguistic areas (Chapter 49).
As is argued by Heine and Kuteva (2005; see also Chapter 23), cases of contact‐induced grammatical change are shaped essentially by the same mechanism as grammaticalization processes not affected by language contact. While grammaticalization exhibits the same kind of unidirectional behaviour irrespective of (p. 16) whether or not language contact is involved, Matras rightly insists that there is need for an overriding framework of language convergence, where grammatical change is viewed as internal to the individual speaker's language processing. In such a framework, he argues, contact‐induced grammaticalization is merely a sub‐category, even if an indispensable one (Chapter 22).