ing a gradual expansion of the semantic- pragmatic contexts in which linguis- tic elements can be placed. .... prosodic pattern" (Dahl2oo 1: 100). ..... talk will be published under the tide "Focalizaci6n deI verbo en maya yueateeo". Markus Giger ...
What makes Grammaticalization? A Look from its Fringes and its Components Edited by
Walter Bisang Nikolaus P. Himmelmann .Björn Wiemer
Mouton de Gruyter Berlin · New York
What makes grammaticalization? An appraisal of its components and its fringes Björn Wiemer and Walter Bisang
1. Introduction 1.1. On the raison d' etre of grammaticalization and its components In the last thirty years the linguistic world has experienced an increasing number of studies on "grammaticalization" (or "grammaticization" according to some approaches; see also Himmelmann, this volume), which tumed into a real boom in the 1990s with publications such as Heine, Claudi and Hünnemeyer (1991), Hopper and Traugott (1993), Bybee, Perkins and Pagliuca (1994) and Lehmann ([1982] 1995). Research on grammaticalization as defined in these approaches deals with the evolution of grammar, or, more precisely, with the question of why originally autonomous linguistic items evolve into syntagmatically bounded morphemes deprived .o f referential content. This definition of the field goes back to Meillet, who introduced the term "grammaticalization" and described it as "le passage d'un mot au röle d'element grammatical [the transition of a word to the role of a grammatical element]" (Meillet [1912] 1948: 131). The boom period of research on grammaticalization was followed by scepticism in recent years. Every single basic tenet of grammaticalization theory became the subject of criticism not only from formallinguists but also from some functionallinguists. For instance, Haspelmath (1998) proposed that reanalysis should not be considered a component of grammaticalization processes, but had better be regarded as complementary t9 it. From the perspective offormallinguistics, Newmeyer (1998) presents a downright deconstruction of grammaticalization in the sense that "there is no such thing as grammaticalization, at least insofar as it might be regarded as a distinct grammatical phenomenon requiring a distinct set of principles for its explanation" (Newmeyer 1998: 226)} Newmeyer (1998: 225-295) divides grammaticalization into three components: morphosyntactic reanalysis, semantic change and phonetic reduction. He argues that none of these mechanisms is unique to grammaticalization, nor does any ofthem "require or entail any ofthe others" (Newmeyer 1998: 295). Moreover, Newmeyer (1998: 254-255) postulates that unidirectionality, i.e, the diachronic development of grammatical categories along a
4
Bjöm Wiemer and Walter Bisang
well-defined relative order of steps, is an epiphenomenon of more funamental principles. Thus, the cline discourse > syntax > morphology > morphophonemics > zero (Giv6n 1979: 209) is the result ofthe more generalleast-effort principle which leads speakers to shorten the linguistic expressions that are most commonly used (cf. Zipf 1935). The contributions to this volume share the common endeavour to look at grammaticalization from a broader perspective which goes beyond changes along grammaticalization "clines" ("channels", "paths") from morphosyntactically more complex to more reduced expression formats or from semantica11y more concrete to more abstract notions. They take grammaticalizationas a general perspective from which to analyse changes in the expression formats of grammatical structure or of the distribution of certain morphological or syntactic units in the languages of the world. From such a more general perspective, the fact that the three mechanisms discus~ed by·Newmeyer (1998) are correlated to a considerable extent in some languages, at least in those Indo-European languages which researchers in the area of grammaticalization looked at most closely, is an interesting finding as such, even if it is not of universal value at a11. This is also confirmed in one way or another by the chapters in this volume dealing with Slavonic languages (Böttger, Hansen, V. Lehmann, Weiss, Wiemer) and East and mainland SoutheastAsian languages (Bisang). If grammar is understood as a system ofmore or less stable, regular and productive form-function mappings, the field of grammaticalization in the above sense of a broader perspective is to be extended to all the processes involved in the diachronic change and in the emergence of such systems. 2 The dynamics of these processes take place between a set of mutually independent components which form the core of grammaticalization and a number of other fields which feed into grammaticalization and thus create its fringes. The core components are morphosyntactic change (reflected by the morpheme-based approach to grammaticalization), semantic or functional change (cf. categorybased approach) and constructions (cf. construction-based approach). These components interact with pragmatics, phonology and the lexicon. The interaction with pragmatics takes either the form of inferences, i.e., reanalysis in general, or the form of input from discourse-relevant morphosyntactic material which gets reanalysed for purposes of grammar (cf. Giv6n's cline mentioned earlier). Due to their interaction with grammaticalization, pragmatics, the lexicon and phonology create the fringes at the borderIine between grammaticalization and these other areas which are of particular interest for understanding what makes grammaticalization. Our broader concept of grammaticalization with its core components and its fringes will be discussed from the perspective of previous research in Sec-
What makes grammaticalization?
5
tion 1.2. For lack of space no exhaustive discussion is intended here; rather, we want to shed light on some basic problems which, in our opinion, any theory of grammaticalization has to deal with. We hope that it will become obvious that many of the differences which exist between the approaches mentioned here derive from differences in the model of language description (and of language use) adopted by their respective representatives. We would like to add that the overall approach of this volume still remains semasiologic. Onomasiological approaches such as the one by Detges (2000) and Detges and Waltereit (2002) will not be integrated into our concept of grammaticalization. The reason is not neglect but limited space, which precludes the more elaborate discussion which this approach deserves. In Section 1.3 we briefty present the history of the present volume, followed by a summary of each contribution.
1.2. Previous research on grammaticalization in the light of a broader notion The beginning of this section (Section 1.2.1) starts out with morphosyntactic change and with semantic change as reflected in the traditional morphologybased and category-based approaches. Given the notorious problem ofhow to distinguish between grammaticalization and lexicalization from the perspective of these two approaches, a large proportion of this section will be devoted to this topic plus related phenomena such as derivation vs. inflection and the problematic relevance of obligatoriness and paradigm formation as a definitorial criterion for grammaticalization. The third core component of grammaticalization - namely, constructions - will be discussed in Section 1.2.2. Since the lexicon as one of the fringe areas of grammaticalization is already discussed in Section 1.2.1, there will only remain two fringe areas. Pragmatics is the topic of Section 1.2.3, phonology is discussed in Section 1.2.4. Section 1.2.5, finally, deals with the role of language contact in grammaticalization.
1.2.1. Morphosyntactic change, semantic change and the question 0/ how to distinguish grammaticalization and lexicalization . 1.2.1.1. General remarks Meillet' s definition of grammaticalization in terms of "the transition of a word to the role of a. grammatical element" departs from a morpheme-based perspective, that is, it hinges on the existence of recoverable morphemes, each with its individual etymology. Grammaticalization clines are based on such etymologies and the morphophonological substance of their "carriers". 3 In
6
Bjöm Wiemer and Walter Bisang
addition to the morpheme-based perspective, MeHlet's key terms "word" and "grammatical element" imply a semantic or cognitive process of change as weIl as a development from the word as an item of the lexicon to the morpheme as an item of grammar. The involvement of the lexicon as weIl as some sort of semantic/cognitive extension ("increase of the range of amorpheme") in addition to the morpheme-based perspective becomes more evident in , Kurylowicz' s (1965: 69) seminal definition of grammaticalization: Grammaticalization 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 [i.e. derivational; BWIWB] formant to an inftectional one.
The involvement ofthe lexicon in Kurylowicz's (1965) definition gives rise to the question of where the borderline is between grammaticalization and lexicalizaHoii. As it wHI turn out in Section 1.2.1.2, the morpheme-based and category-based approaches alone cannot answer this question. What is needed is a structural framework provided by constructions. The distinction between derivation and inflection, which is of course related to the lexicon-grammar distinction, will be treated in Section 1.2.1.3. We show that derivational expression formats can expand their functional range without any corresponding change in morphosyntactic form. Finally, in Section 1.2.1.4 we argue that obligatoriness and paradigm formation may not be fully reliable indicators of a high degree of grammaticalization. 1.2.1.2. Grammaticalization vs. lexicalization . The problem of how to distinguish between grammaticalization and lexicalization is notorious in research on grammaticalization, although it seems to be badly neglected in some approaches. It is implicitly present in the prototypical definition of grammaticalization as the development of a lexical item to a marker of a grammatical category - a development which presupposes a continuum between lexicon and grammar. If we look at language change from the morpheme-based perspective as reflected in Giv6n's grammaticalization cline mentioned earlier, it turns out that similar morphosyntactic and morphophonological processes apply to grammaticalization and to lexicalization. Thus, the category-based perspective and the morpheme-based perspective both call for a clear-cut distinction between the two linguistic levels of lexicon and grammar and for a clear-cut borderline between what belongs to grammaticalization and what belongs to lexicalization. Himmelmann's chapter (this volume, see also Section 1.3 in the present chapter) is an important contribution to the distinction of grammaticalization
What makes grammaticalization?
7
and lexicalization. As he points out, grammaticalization has to do with changes in functional range and must be described in terms of constructions. Thus, grammaticalization is a (uni}directional phenomenon confined to changes that inftuence the functional range of morphemes (or their combinations) and it takes place within the framework of constructions (cf. Section 1.2.2), reftecting a gradual expansion of the semantic- pragmatic contexts in which linguistic elements can be placed. The relevance of constructions can be illustrated by a closer look at how semantic or cognitive maps are used in research on grammaticalization. 4 Cognitive maps are used as a tool in category-based approaches. The functions arranged in a semantic map are spatially represented as focal points within a continuum. The evidence for a specific spatial arrangement comes from a cross-linguistically consistent form-function mapping. In spite of this, the mere coincidence of functions within a semantic map reftected in formal properties is not sufficient for postulating the existence of a grammatical category. It is also necessary to account for the syntagmatic relations as they are manifested within constructions between the members on the form side of such a mapping relation.Otherwise principally any lexical element can be regarded as a grarnmatical marker provided it fulfils a function related to such focal points within a given semantic space in some particular context. The obvious negative consequence of such a reasoning would be that lexicalization cannot be distinguished from grammaticalization. The requirement of linking semantic properties as they are represented in semantic maps with clear-cut formal criteria as manifested in constructions is not recognized (nor accepted) in category-based approaches which try to explain grammaticalization as the outcome of cognitive problem-solving, triggered by reanalysis (metonymy-based) and conceptual "leaps" (metaphorbased) (Heine et a1. 1991). Leaving aside the question of whether such an approach can provide a plausible explanation of what makes grammaticalization, semantic (functional) change caused by conceptual or textual contiguity or by metaphorical reasoning is a motive underlying lexical variation, too. Purely category-based approaches that lack other formal and/or distributional criteria are thus systematically forced to treat grammaticalization and lexicalization as one and the same phenomenon. 1.2.1.3. Derivation vs. inftection If we look again at Giv6n's grammaticalization cline, we see that there is 00 place for derivational morphemes (but cf. Himmelmann, this volume: Section 2). It is quite possible to think of a development from derivational affixes to inflectional affixes as noticed repeatedly in the literature (e.g. the cases
8
Bjöm Wiemer and Walter Bisang
discussed by Kurylowicz 1964 in ancient Indo-European languages or by Giacalone Ramat 1998 on Italian data). But for most scholars it seems to be hardly disputable that as soon as some derivational affixes get involved in a regular paradigmatic (Le. grammatical) opposition they lose their derivational status and become part of an intlectionaI paradigm. The relation between inftection and derivation is thus automatically associated with a grammatical relation rather than with a lexical one. Given the tadt assumption in traditional theories of grammaticalization thatthere is a co-evolution between semantic bleaching and morphosyntactic integration (C. Lehmann [1982] 1995, Bybee 1985), it is impossible to think of a process in which a derivationaI pattern extends its functional range into a semantically more general (see Bybee 1985 on generality) and more productive category without any corresponding change in the derivational expression format of that pattern. If such a process exists, it is, paradoxical1y, a nice il1ustration of the independence of semantic change and phonetic reduction as postulated by Newmeyer (1998); see also n. 11. As the two chapters dealing with Slavonic languages show (V. Lehmann, Böttger), the category of aspect in this language family develops basically along the lines of such a process. In such a situation, a morpheme-based approach simply cannot make any adequate statement about the status of such a process between grammaticalization and lexicalization. As far as the Slavonic aspect opposition is concemed, the difference between the lexical and the grammatical status of stern derivation by affixes cannot be captured by looking at the form of the stern alone. It can be determined only on the premise that derivation does not necessarily change the lexical meaning of the deriving stern and by correlating pairs of lexically (almost) identical sterns with functions from actionality, temporallocation (semelfactivity, iteration, etc.), modality, etc. (see, inter alia, V. Lehmann 1999a, b; Wiemer 2001). From this perspective, the evolution of Slavonic aspect systems can reasonably be described and understood only within category-based approaches (as they have been developed, among others, by Bybee et al. 1994; cf. also Bybee and DahI1989). 1.2.1.4. The role of paradigm formation and obligatoriness in grammaticalization Much discussion has recently been devoted to the question of how much relevance should be ascribed to paradigm formation and the parameter of obligatoriness (both transparadigmatic and intraparadigmatic in terms of C. Lehmann 1995). Bybee (1985), Bybee et al. (1994), C. Lehmann (1995) and also Dahl (2000, 2001) are rather strong adherents of the view that the degree of obligatoriness is a reliable symptom of grammaticalization. Plungjan (2000:
What makes grammaticalization?
9
106) considers it a necessary and even sufficient parameter of the grarnmatical status of a given morphological opposition. 5 If one looks at the languages ofEast and mainland SoutheastAsia, there are hardlyany obligatory categories nor are there any paradigms. Bisang's contribution to this volume shows that the reason for this is linked to the strong impact of pragmatics (see Section 1.2.3) in these languages. From a more general perspective one may say that obligatoriness and paradigm formation are grammaticality parameters which are almost inevitable in a large number of languages, first of all Indo-European, but they are not absolute or universal criteria for measuring grammaticalization.
1.2.2. Constructions and grammaticalization Processes of erosion, as they are represented in grammaticalization clines, processes of semantic change (cf. Section 1.2.1), and of reanalysis are three central components of grammaticalization, but grammaticalization is more than that. Reanalysis does not happen out of the blue, it happens with regard to a concrete syntagmatic environment within which a certain linguistic item can be interpreted as belongiIig to a position differing from its erstwhile positional properties (cf. Bybee et al 1994: 11 et passim; Himmelmann, this volume). In the functional approach adopted in this volume, it is constru9tions that · form the syntagmatic framework which makes reanalysis possible. 6 Ifwe take constructions in the sense of Construction Grammar (see e.g. Goldberg 1995) as linguistic items with their own meaning which cannot be derived from their components, we also have to reckon with the emergence of new constructions and thus with the emergence of new grammatical structures which either replace older structures to express the same concept or which express new concepts for which there was no structure at an earlier time. 7 If grammaticalization is understood in terms of changes in the expression formats of grammatical structure in the languages of the world (cf. Section 1.1) such processes of rebuilding and building up grammatical structures should also be subsumed under the heading of grammaticalization. This assumption is also compatible with Dahl's (2001) "gram type"-based claim that grammaticalization 'processes should be considered even for "constructions without any overt markers, with the possible exception of a specific prosodic pattern" (Dahl2oo 1: 100). If constructions can have their own meaning it is to be expected that they have the potential to be grams, that is, expression formats for grammatical categories. Dahl's (2001) standard examples in this context are (productive) composition and coordinating constructions.
10
Bjöm Wiemer and Walter Bisang
The interaction of individual processes of grammaticalization and their integration into a new construction will be particularly discussed in Wiemer's chapter on passives (this volume).
1.2.3. Pragmatics and grammaticalization Pragmatics is discussed from two different perspectives. It is an important motor of grammaticalization in some approaches (Hopper and Traugott 1993; Bybee et a1. 1994) and it is seen as the starting point of Giv6n's grammaticalization clines presented earlier. Fina11y, the degree to which pragmatic inference is involved in the gradual consolidation of the expression of grammatical concepts seems to be subject to cross-linguistic variation. Bybee et al. (1994) discuss pragmatic inferences in the sense of Grice (1975). In a given situation of communication, the hearer is obliged to extract all the meaning possible from the message,which includes a11 the implications that are not controversia1. A semantic change can ~e place when a certain implication commonly arises with a certain linguistic form. That implication can be taken as part of the inherent meaning of the form, and can even go as far as to replace the original meaning of the form. Thus, "[w]hat makes inference interesting as a mechanism of change is the fact that inference a110ws the incorporation of new meaning into a gram" (Bybee et al. 1994: 289). Hopper and Traugott's (1993) approach is of particular interest for an approach which tries to integrate pragmatic inference and constructions. In their view, pragmatic inference is divided into two processes which both initiate processes of grammaticalization. The one which is more important and which is relevant for integrating constructions is metonymy in the way defined by Hopper and Traugott (1993).8 Metonymy is based on reanalysis and is the product of conversational implicature. It operates at the syntagmatic level and derives from abduction. This definition can be applied to constructions if they are taken to be the syntagmatic framework within which reanalysis takes place, i.e., within which a given morpheme is ascribed to another or a new functional one. The less important pragmatic inference is metaphor, which is defined as a product of conventional implicature and is linked to analogy. It operates at the paradigmatic level. Table 1 provides a survey of how the two inferential processes are defined and how they differ from each other. Discourse is seen as the starting point of grammaticalization clines. Thus, elements initially used for purposes of structuring discourse change into markers of syntax, and so on. However, developmental processes do not always start with discourse. There are a number of processes that lead from other sources
What makes grammaticalization?
11
Table 1. Mechanisms of language change: metonymy and metaphor
Metonymy
Metaphor
-
-
Syntagmatic level Reanalysis (abduction) Conversational implicature Operates through interdependent syntactic constituents
Paradigmatic level Analogy Conventional implicature Operates through conceptual domains
into discourse. Günthner and Mutz (see ch. 4) discuss two examples of such processes - from adverbial subordinators and from diminutive suffixes to pragmatic markers. The question is whether these processes can be subsumed under grammaticalization. Günthner and Mutz think not and prefer to call them "pragmaticalization". From the fact that the phenomena described by Günthner and Mutz do not reflect grammatical categories, it is certainly wise to treat them under a different heading. In spite of this, many pragmatic functions depend on a clear-cut morphosyntactic expression format which qualifies as a construction - in this case as a construction with its own pragmatic meaning. Since constructions clearly belong to grammar, the phenomena described by Günthner and Mutz still belong to the fringes of grammaticalization, even if their pragmatic function lies outside the range of grammatical function. From this, we can see that reanalysis can produce morphosyntactic structures (patterns) even in cases in which the functions involved are traditionally not related to grammatical categories. Finally, careful cross-linguistic analysis shows that there are languages in which what may be expressed by clear-cut grammatical categories in some other languages is the result of pragmatic inference rather than the result of clearly defined grammatical functions such as aspect or tense. The languages of East and mainland Southeast Asia asodescribed by Bisang (this volume) belong to this type of language. In these languages, aspectual and temporal notions are associated with certain markers by pragmatic inference. Thus, the expression of a number of grammatical categories in these languages originates in the fringes of grammaticalization.
1.2.4. Phonology and grammaticalization Grammaticalization gets its input not only from above - from pragmatics in the sense that discourse structures are subject to reanalysis - but also from below, that is, from the reanalysis of phonological structures and phonological
12
Bjöm Wiemer and Walter Bisang
rules. This aspect of grammaticalization is treated in Gaeta's contribution to this volume. His chapter also discusses the central position of morphology. While, under this view, grammaticalization aims at morphology from syntax as weIl as from phonology, there is almost no process leading away from morphology to either phonology or syntax. This statement partly contradicts Giv6n's grammaticalization cline from discourse to syntax to morphology to morphophonemics to zero (see also Section 1.1) because it excludes developmental processes further down from morphology and emphasizes developments from below morphology up to morphology. If we look at the instances of grammaticalization at the lower end of Giv6n's cline, we see that the majority of them have to do with economy: phonological substance is reduced for the purpose of making the pronunciation of an already highly grammaticalized linguistic item easier. What this approach does not consider is that reanalysis does not stop at syntactic or morphological patterns; it is also possible to take phonological patterns as a starting point for reanalysis in tenns ofhigher levels.
1.2.5. Grammaticalization and language contact Processes of grammaticalization may be reinforced by language contact in the sense that they are triggered by a language having them in a language which lacks them. This is particularly true of processes of reanalysis that may be mutually enhanced among the speakers of the languages in contact (Bisang 1996). Given the centrality of reanalysis as a process situated rather at the beginning of grammaticalization it may come as no surprise that the extent of contact-induced structural convergence may be subject to considerable variation within the individuallanguages, since grammaticalization may have stopped at different stages further down the grammaticalization cline for each language. Grammaticalization and contact-induced structural convergence are discussed in Weiss's and in Bisang's contributions to this volume.
1.3. On the structure ofthis volume The idea to put together a volume like this came up during the preparation of a workshop titled "Grammatikalisierung vs. Lexikalisierung", which took place from 1 to 3 February 200 1 in Constance (Gennany) and was organized by Björn Wiemer. 9 The original idea of this workshop was to present controversial cases of grammatica~ization that would open up a discussion guided by two questions: (i) In which way should grammaticalization be discrim-
What makes grammaticq.lization?
13
inated from lexicalization? and (in Ought we to widen grammaticalization theory in a way that would a110w the inclusion of types of structural change which can hardly be described by (or even contradict)prevalent parameters of grammaticalization, first of a11 those established by C. Lehmann (1995)? In other words, is it reasonable and viable to aim at a more comprehensive grammaticalization theory, or should we rather not "sacrifice" its mainly morpheme-based tenets? The participants of the workshop all feIt existing theories to be either insufficient or too much restricted to some specific phenomena of linguistic change (or to specific types of language). Consequently, they gave talks on phenomena which, in one way or another, challenged various tenets of well-known approaches to grammaticalization, or they tried to define more precisely how grammaticalization and lexicalization should be understood and distinguished from each other. One of the main results of the discussions during the workshop was that both the need for a more comprehensive theory of grammaticalization and the distinction between grammaticalization and lexicalization can be adequately dealt with only if distributional properties of linguistic units and the interaction between sterns and diverse functional morphemes are taken into account (see in particular Himmelmann's and V. Lehmann's contributions). We do not pretend that the participants were unanimous with respect to this point. Nor are the contributions collected here united under one common approach. In fact, they vary quite a lot as to the decisions taken. However, what we do claim is that the chapters included in this volume either present views on grammatical phenomena which hitherto have been neglected in studies ,on grammaticalization or discuss again seemingly well-known phenomena which turn out to be problematic if investigated with greater scrutiny. The provocative question then arises whether we have to widen the theory (in order to capture these phenomena), or whether we should rather not treat the respective phenomena as cases of grammaticalization at all (for "keeping the theory clean" of them). Many of the contributions also elaborate on the model of grammar to be adopted in order to adequately describe the phenomena to be analysed within the framework of grammaticalization (see the chapters by Bisang, Gaeta, Himmelmann, König and Vezzosi, V. Lehmann, and Wiemer). This volume consists of eleven contributions (including this introduction). lO Five of them are devoted to topics illustrated on the material of Slavonic languages (and written by Slavists), although they are by no means confined to phenomena exclusive to Slavonic languages. The large proportion of chapters dealing with data from these languages reftects one of the targets of the workshop - which has also become oneof the aims of this volurne - namely, to try to integrate studies in this European language family
14
Bjöm Wiemer and Walter Bisang
into ongoing research on grammaticalization. For it is quite remarkable that Slavonic languages have remained underrepresented in such research. One of the reasons might be a slower reception of work on grammaticalization and a different way of theorizing in Slavonic-speaking countries (for an overview see e.g. Plungjan 1998). More generally speaking, the very understanding of grammaticalization seems to have been different in a way which did not give rise to a theory specifically dealing with changes in grammatical structures (as is the case with the morpheme- or the construction-based approach outlined above; cf. Wiemer 2003 for a critical survey). Another reason for the low integration of Slavonic data into the literature on grammaticalization might be the fact that most published work and recent and ongoing research in Slavonic countries is not easily accessible to many western scholars. However, there is yet another reason for this state of affairs, which is more deeply rooted and has to do with the comparatively conservative character of most Slavonic languages with respect to diachronic changes in morphosyntax (see e.g. Hansen's and Wiemer's contributions to this volume). Weiss (this volume) shows that the two Slavonic languages Macedonian and Bulgarian "lag behind" the other Balkan languages with regard to the development of an indefinite article. The general impression arises that Slavonic languages change slowly, at least on- an average European level and regardless in which part ofEurope they are spoken (and with which languages they are in contact). Another "peculiarity" of Slavonic languages seems to be the capacity of their morphemes to resist phonological erosion in cases ofloss in referentiality (i.e., in cases of bleaching) and drift to the pole of functional markers. In fact, there are comparatively few cases in Slavonic in which attrition accompanies functional changes and the distributional expansion of morphemic markers. For instance, this does not happen with the Macedonian numeral eden 'one' (which is developing into an indefinite article; see Weiss's contribution), nor is it typical of Slavonlc modal auxiliaries (cf. Hansen's chapter), nor can we observe it with morphemes that have become components of passive constructions in Slavonic (and Baltic; see Wiemer's chapter). Finally, attrition is highly unlikely with derivationally based aspect (cf. V. Lehmann and Böttger). On the other hand, Slavonic languages appear to be more far-reaching in their functional development of verbal derivational affixation. According to Bybee and Dahl (1989) and Dahl (2000), the Slavonic aspect system is unique with its way of expressing the opposition between perfective and imperfective aspect by means of stern derivation. From the morphological point of view, the roots of this system are shared with Baltic and date back into prehistoric times. The same roots can even be detected in Germanic, where stern modification by prefixes (or preverbs) still is quite productive. But nowhere, except
What makes grammaticalization?
15
in Slavonic, has this productivity led to a grammaticalized opposition between perfective and imperfective aspect. This shows that productivity as such cannot be made a reliable criterion for (ongoing) grammaticalization. In any way, all the above "Slavonic peculiarities" make grammatically relevant changes much less transparent because they cannot be seen from mere changes of form, but have to be captured from changes in their distribution (mandatory, optional, or excluded) over inventories of functions. From this perspective, much data from Slavonic presents areal challenge for a theory that aims at describing and explaining the development towards (more) grammar, that is, a comprehensive theory of grammaticalization. 11 Let us turn now to the "non-Slavonic" chapters. The first, by Nikolaus Himmelmann, focuses on one of the central issues of grammaticalization theory wbich nonetheless has often been neglected. He develops the criteria relevant for distinguishing between grammaticalization and lexicalization from the perspective of the processes involved. On the one hand, bis argumentation starts from the rigid postula~e that grammaticalization is a (uni)directional phenomenon confined to changes that influence the functional range of morphemes (or their combinations). On the other hand, he strongly advocates a construction-based approach, which understands grammaticalization as a gradual expansion of the assembly of semantic-pragmatic contexts in wbich Iinguistic elements are placed. This kind of expansion can be (and normally is) accompanied by an expansion of the host-class (e.g. stems of a particular word class) and/or an expansion of the syntactic scope (from NP, VP or PP to c1ause, sentence or even a higher level). From this perspective, paradigm formation and erosion are to be regarded as epiphenomena. Walter Bisang' s chapter integrates the areal perspective and is thus related to the chapter by Daniel Weiss. His basic concern is that East and mainland Southeast Asian languages rarely build up paradigms because of their high degree ofindeterminateness and because in general there is a weak correlation between the lexicon and morphosyntax. He relates this state of affairs to the observation that in all these languages pragmatics plays a much greater role in encoding grammatically relevant information than in European in many other languages of the world. As a consequence, semantic and phonetic reduction cannot be used as a diagnostic key for assessing linguistic change in the structure of these languages. In addition, Bisang also questions whether functions must be uniformly mapped on grams (in Bybee's sense of this term). This does not invalidate the use of semantic maps (and of the category-based assumptions inherent to grammaticalization theory; see Section 1.2.1.2), but it strongly relativizes a conception of grammar (and grammaticalization) with more or less clear-cut and stable markers of categories.
16
Bjöm Wiemer and Walter Bisang
Whereas Bisang 's chapter can be seen as an illustration of (virtuaO grammaticalization "from above" - that is, from pragmatics - Livio Gaeta looks at grammaticalization "from below": on the reinterpretation·of phonological into morphological oppositions. His data derives from a broad variety of languages which clearly show that what he presents are rather common and frequent phenomena. Fro~ his look at grammaticalization from below, Gaeta argues for a morphocentric model of language which discloses a number of assumptions often hidden behind so-called 'grammaticalization clines'. Susanne Günthner and Katrin Mutz take another look at grammaticalization from the pragmatic perspective. Their theoretical discussion starts out from the differences between C. Lehmann's and Traugott's conceptions and parameters of grammaticalization. The first analytical section of their chapter is a case study of German conjunctional units, which develop from subordinative connectors to so-called discourse particles, that is, function words connecting larger parts of spontaneous discourse (monologues or dialogues). Their functional development shows an increase in pragmatically based function and can, thus, be subsumed underTraugott's notion "subjectification". The second analytical section deals with the semantic evolution of Italian modificationsuffixes. Starting out from their quantifying (diminutive, augmentative) functions these suffixes quickly acquire qualifying functions, which, in turn, develop into discourse functions. In this sense, they are comparable to German connective particles (though the authors also point out differences). The two cases discussed in this chapter demonstrate the "semanticization of a discourse-pragmatic function", but by this the suffixes simultaneously develop "from more grammatical elements (on the level of sentence-grammar) to less grammatical elements, functioning mainly on the discourse lever'. In addition, Günthner and Mutz also argue for integrating morphemes reanalysed for discourse purposes (particles or suffixes, respectively) into the lexicon. Finally, they advocate the view that different kinds of change related to grammar should be distinguished, namely, morphologization, syntacticization and pragmaticization; they deal with the latter of these phenomena. In their opinion, pragmaticalization should be set apart from grammaticalization proper. Ekkehard König and Letizia Vezzosi address the question of the autonomy of syntax. It thus takes up a central issue from the generative literature, which is relevant to grammaticalization inasmuch as clines from lexical (autosemantic) to grammatical (morphosyntactic, functionaO status are considered to be based on the loss of semantic structure ("bleaching"). In asense, the argument developed by König and Vezzosi turns this perspective upside down, since they ask whether there are semantically or pragmatically based restrictions on the applicability of grammatical elements to lexical sterns, that is, of reflex-
What makes grammaticalization?
17
ive markers to verbs. This point is developed on data from different languages, but first of all from English and some other Germanic languages. König and Vezzosi show not only that such restrietions do indeed exist, but also that they are ubiquitous both in diachronie development and in typologie al comparison. The crucial criterion sensitive to these restrictions is the feature [± otherdirected (action)]. This can be seen from the use of heavy vs. light reflexive markers, from the application vs. non-application of a reflexive marker in languages like modern English, from nominalizations and from the distribution of default reflexive and reciprocal interpretations in languages whose reflexive markers show a systematic ambiguity between these two functions (in the thi_rd-person plural). Of particular importance for grammaticalization is König and Vezzosi's finding that in the history ofEnglish a formerly marked emphatic strategy (+ self) became the unmarked strategy (with loss of emphatic function) by continually expanding its compatibility with an increasing number of verbs. Thus, the expansion of grammatical or functional markers through the lexicon (i.e. through the verbs compatible with them) again turns out to playa crucial role in grammaticalization.
Notes I. See also the contributions in the special issue of Language Sciences (Campbell 2(01). 2. We do not use the tenn "emergence" in the radical sense ofHopper (1987). What is mean~ here is the development of new structures such as a marker for a new grammatical category which was not present at an earlier stage, or the emergence of a new construction such as the passive (see Wiemer, this volume). J. Himmelmann (Ch. 2 in this volume) calls this the "element-based view on grammaticization" . 4. See e.g. Haspelmath (1997, 2(03) and Stassen (1997), among others, for the discussion of semantic maps. ~. The roots of this view can be traced back to lakobsonian views on obligatoriness and the nature of zero grammemes, which are reflected in the most influential morphological theories in Russia(cf. Wiemer 2003:59-61 for a summary). (,. From a generative perspective, Newmeyer (1998: 292-295) points at Roberts's ( 1993) approach towards a characterization of the circumstances in which reanalysis can take place. 7. 80th processes were labelIed 'exaptation' by Vincent (1995), although he did not rcfer to constructions. K. '(he terms "metonymy" and "metaphor" are used by different authors on grammalicalization. The concepts behind these words vary to a rather large extent in euch approach. Thus, the above definition of metonymy is limited to Hopper and
18
Bjöm Wiemer and Walter Bisang
Traugott (1993) and has nothing to do with the definition presented, for example by Heine et al. (1991). 9. We would like to thank the German Scienee Foundation (DFG) for supporting this workshop in connection with the SFB 471 at Constanee University. • 10. Some of the participants have published or will publish articles eonneeted to their talks at the workshop on other oeeasions and in other plaees: Ulrie,h Detges' eontribution appeared in "Studies in Language" (Detges 2(00), Christian Lehmann's talk will be published under the tide "Focalizaci6n deI verbo en maya yueateeo". Markus Giger has meanwhile finished a monograph on Czeeh (and Slovak) resultative eonstruetions (Giger 2(03). 11. To some extent, this type of semantie evolution without phonologieal erosion seems to eorroborate Newmeyer's eriticism of grammatiealization theory (see Seetion 1.1). However, in the ease of derivational aspeet we are dealing notjust with a simple proeess of semantie bleaebing, but with a eomplex interaction of lexical stern and derivational affixes extended over a whole word-class, in tbis ease the verb.
References Bybee, Inan L. 1985 Morphology: A Study 0/ the Relation Between Meaning and Form. (Typologieal Studies in Language 9.) AmsterdamlPhiladelphia: lohn Benjamins. Bybee, loan L., and Östen Dahl, The ereation of tense and aspeet systems in the languages of the world. 1989 Studies in lAnguage 13 (1-2): 51-103. Bybee, loan L., Revere Perkins, and William Pagliuea 1994 The Evolution 0/ Grammar: Tense, Aspect, and Modality in the lAnguages o/the World. Chieago/London: Cbieago U~versity Press. Campbell, Lyle (ed.) Grammaticalization: A Critical Assessment. (Language Sciences 23 2001 (2-3), special issue). Dahl, Östen 2000 The tense-aspeet systems of European languages in a typologie al perspeetive. In Tense and Aspect in the Languages 0/ Europe, Östen Dahl (ed.), 3-25. (Empirieal Approaehes to Language Typology. Eurotyp 20 (6).) Berlin/New York: Mouton de Gruyter. 2001 Grammatiealization and the life eycles of eonstruetions. RASK 14: 91133. Detges, Ulrieh 2000 Time and truth: The grammatiealization of resultatives and perfeets witbin a theory of subjeetifieation. Studies in lAnguage 24 (2): 345377.
What makes grammaticalization?
19
Detges, Ulrich, and Richard Waltereit 2002 Grammaticalization vs. reanalysis: A semantic-pragmatic account of functional change in grammar. Zeitschrift für Sprachwissenschaft 21 (2): 151-195. Giacalone Ramat, Anna 1998 Testing the boundaries of grammaticalization. In The Limits 0/ Grammaticalization, A. G. Ramat and P. J. Hopper (eds.), 107-127. (Typological Studies in Language 37.) Amsterdam/Philadelphia: John BenJanuns. Giger, Markus 2003 Resultativa im modemen Tschechischen: unter Berücksichtigung der Sprachgeschichte und der übrigen slavischen Sprachen. (Slavica Helvetica 69.) Beme: Peter Lang. Giv6n, Talmy 1979 On Understanding Grammar. New York: Academic Press. Goldberg, Adele 1995 A Construction Grammar Approach to Argument Structure. Chicago/ London: University of Chicago Press. Grice, H. P. 1975 Logic and conversation. In Speech Acts, P. Cole and J. Morgan (eds.), 41-58. (Syntax and Semantic 3.) New York: Academic Press. Haspelmath, Martin 1998 Does grammaticalization need reanalysis? Studies in Language 22 (2): 315-351. 1999 Whyisgrammaticalizationirreversible? Linguistics37 (6): 1043-1068. 2003 The geometry of grammatical meaning: Semantic maps and crosslinguistic comparison. In The new psychology o/language, M. Tomasello (ed.), 211-242. (Vol. 2.) Mahwah, NJ: Lawrence Erlbaum. Heine, Bemd, Ulrike Claudi, and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago/London: University of Chicago Press. Hopper, Paul J. 1987 Emergent grammar. In Berkeley Linguistics Society. Proceedings 0/ the Thirteenth Annual Meeting / General Session and Parasession on Grammarand Cognition, J.Aske et al. (eds.), 139-157. Berkeley, CA: Berkeley Linguistics Society. Hopper, Paul J., and Elisabeth C. Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Kurylowicz, Jerzy 1964 Th~ Inflectional Categories o/lndo-European. Heidelberg: Winter. 1965 The evolution of grammatical categories. Diogenes 51, 55-71. [Reprint in: J. Kurylowicz (1975), Esquisses linguistiques 11. Munich: Fink, 3854,]
20
Björn Wiemer and Walter Bisang
Lehmann, Christian 1995 Thoughts on Grammaticalization. (LINCOM Studies in Theoretical Linguistics 1.) Munich/Newcastle: Lincom Europa. Lehmann, Volkmar 1999a Aspekt. In Handbuch der sprachwissenschaftlichen Russistik und ihrer Grenzdisziplinen, H. Jachnow (ed.), 214-242. Wiesbaden: Harrassowitz. 1999b Lexikalischer Stamm und grammatische Kategorie. In Slavistische Linguistik 1998, R. Rathmayr and W. Weitlaner (eds.), 137-149. (Referate des XXIV. Konstanzer Slavistischen Arbeitstreffens. Vienna, 1518 Sept. 1998). Munieh: Sagner. (Slavistische Beiträge 380.) Meillet, Antoine 1948 L' evolution des formes grammaticales. In Linguistique historique et linguistique generale, A. Meillet (ed.), 130-148. Paris. [Original edition: Scientia (Rivista di scienza) 12, No. XXVI, 6, 1912.] Newmeyer, Frederick J. 1998 Language Form and Language Function. Cambridge, Mass.: MIT Press. Plungjan, Vladimir A. 1998 Problemy grammaticeskogo znacenija v sovremennyx morfologiceskix teorijax (obzor). Semiotika i informatika 36: 324-386. 2000 Obscaja Morfologija: Vvedenie v Problematiku. Moscow: Editorial URSS. Roberts, Ian 1993 A formal account of grammaticalization in the history of Romance futures. Folia LinguisticaHistorica 13: 219-258. Stassen, Leon 1997 Intransitive Predication. Oxford: Clarendon Press. Vincent, Nigel 1995 Exaptation and grammaticalization. In Historical Linguistics 1993 (Selected Papers from the 11th International Conference on Historical Linguistics), H. Andersen (ed.), 433-445. AmsterdamlPhiladelphia: John Benjamins. Wiemer, Bjöm 200 1 Aspektual' nye paradigmy i leksiceskoe znacenie russkix i litovskix glagolov (Opyt sopostavlenija s tocki zrenija leksikalizacii i grammatikalizacii). Voprosy jazykoznanija 2001 (2): 26-58. 2003 Grammatische Kategorien und Grammatikalisierung in der Forschung der Sowjetunion und Polens: Beitrag zu einem bislang nicht aufgearbeiteten wissenschaftsgeschichtlichen Kapitel. Zeitschrift für Slawistik 48 (1): 55-82. Zipf, George 1935 The Psychobiology ofLanguage. New York: Houghton Mifflin.