1.6.1 Processes versus results
Some authors, in particular Fischer (2000), Joseph (2001, 2003, 2004) and Newmeyer (1998, 2001), have argued that grammaticalization is not a separate process of change. According to Newmeyer (1998: 232), the term ‘process’ should refer to ‘a distinct phenomenon requiring an inherent set of explana- tory devices’, but in the case of grammaticalization changes, ‘[t]he term ‘‘process’’ is often used informally to mean nothing more than ‘‘phenomenon to be explained’’ ’. Instead, Newmeyer (1998: 235) characterizes grammatica- lization as ‘essentially an epiphenomenal result of independent historical developments each of which falls out of some independent theory.’ Since semantic change, phonetic reduction, and reanalysis, Newmeyer argues, may occur independently from one another, there is no reason to assume a separate dynamic force (i.e. grammaticalization) when these changes happen to co-occur. For the same reason, he argues, there is no need for a ‘gramma- ticalization theory’.
In his discussion of some stock examples of grammaticalization (among them the well-known development of the Greek future marker tha, ultimately from a full verb the´lo: ‘want’) Joseph (2001, 2003, 2004), too, argues that this change can be fully explained within the framework of ‘traditional’ historical linguistics, and that grammaticalization is at best ‘just a label given to a
29 Nevertheless, some approaches to grammaticalization regard it as primarily semantically- pragmatically motivated. See Fischer (2008: 340) for a convenient overview.
particular type of outcome of independently needed mechanisms of change’ (Joseph 2004: 51).
Fischer (2000: 151, 152) similarly argues:
One of the problems I have with the way grammaticalization has been dealt with in the literature is that the mechanistic side of it has been overemphasised, with the result, I think, that the mechanism has become too powerful as an explanatory tool or as a description of a diachronic process of linguistic change . . . Although I would agree . . . that reanalysis and analogy, or metonymic and metaphorical processes, are important in language change . . . I still cannot see that there is room for a separate or ‘independent’ process of grammaticalisation. Where most linguists see a unidirec- tional process from concrete to abstract, a process that cannot be cut up into segments, I can only see a more or less accidental concurrence.
In a more recent paper Fischer (2008: 337) maintains that ‘the notion of grammaticalization can be useful as a heuristic device to discover changes that have taken place in language and to understand why changes often follow similar pathways’, but that it is ‘incorrect to elevate grammaticali- zation to some higher status, to some independent mechanism of change’.30
This line of criticism is rejected by Haspelmath (1999a: 1062), and Heine and Kuteva (2002: 2f.). According to the latter, ‘the main task of grammati- calization theory is to explain why grammatical forms and constructions are structured the way they are, and these four mechanisms [see (31) below], as opposed to many other conceivable mechanisms, have been found to be relevant to achieve such explanations’.
(31) The principal mechanisms involved in grammaticalization (Heine and Kuteva 2002: 2):
1. Desemanticization or semantic bleaching – loss of (concrete) meaning 2. Extension or context generalization – use in new contexts
3. Decategorialization – loss of morphosyntactic properties (e.g. inXection) 4. Erosion or phonetic reduction – loss of phonetic substance
The mechanisms in (31) are clearly interrelated according to Heine and Kuteva (2002: 3f.). The initial stage of grammaticalization involves context-induced reinterpretation of concrete items as abstract items. Such reinterpretations can often be described as metaphorical extensions, for example when a body-part
30 Instead, Fischer (2008: 338) argues that grammaticalization is essentially derived from more fundamental, cognitive principles which are at work in both language change and language acquisition. This is a novel and interesting approach, but one that cannot be discussed in detail in the present work.
noun such as ‘back’ is used to denote a spatial concept ‘behind’. Semantic bleaching is thus the result of the transition from concrete meaning to abstract, grammatical meaning. Once such forms have acquired grammatical meaning, they are subject to decategorialization and phonetic reduction (because grammatical forms are usually unstressed). In other words, even though the four mechanisms in (31) have been observed to occur independ- ently, they are not independent of one another in grammaticalization. In a pertinent comment on Linguist List, Dahl (1996) puts it this way:
This to me seems like saying that since love and sex can occur without each other, they are totally diVerent phenomena. For [Newmeyer’s] argument [in an earlier posting, against grammaticalization as a process] to go through, he would have to show not only that the processes can occur independently but also that they are unrelated even in the well-documented cases when they show up together. What some of us have claimed is that the things that happen in grammaticalization do so in an orderly fashion which not only predicts what changes can occur but also puts constraints on what synchronic grammatical systems are found.
Heine (2003b: 579) also acknowledges the criticism by Newmeyer and Campbell that the changes involved are not unique to grammaticalization, but he points out that ‘jointly they are responsible for grammaticalization taking place’, and thus, that ‘they can be said to constitute diVerent compon- ents of one and the same process’.
To sum up the discussion thus far, I think that the arguments brought forward by Heine, Kuteva, and Dahl are suYciently convincing to allow grammaticalization to be regarded as a process with a dynamics of its own. Nevertheless, three caveats are in order. First, acknowledging that gramma- ticalization is a process is not tantamount to saying that grammaticalization is a ‘driving force’. To my mind, Bybee, Perkins, and Pagliuca (1994: 298) hit the nail on the head when they write that ‘The processes that lead to grammaticalization occur in language use for their own sakes; it just happens that their cumulative eVect is the development of grammar’ (emphasis mine).
Secondly (and this has frequently been stressed in grammaticalization literature), there is nothing deterministic about grammaticalization (see e.g. Traugott 2001). In other words, grammaticalization is a process in the sense that it involves a series of subsequent steps that occur in a certain order, as we have seen above, but not in the sense that each step inevitably leads to the next (this lack of inevitability was Newmeyer’s (1998: 251) main argument for rejecting the view of grammaticalization as a ‘distinct process’). Traugott (2001: 3) puts it this way: ‘Changes do not have to occur. They also do not have to go to completion, in other words, they do not have to move all the
way down a cline, or even continue down it once they start out on it.’31In addition, grammaticalization need not involve change at the level of phon- ology (at least not initially). For instance, there is no diVerence in pronun- ciation between the participle considering (We are considering going to Denmark) and its grammaticalized form as a conjunction (Considering her age she is still quite active). In fact, it may take ages before a grammaticalized form is phonetically reduced, if this happens at all. The Modern Swedish preposition bland ‘among’, for example, which grammaticalized over half a millennium ago from a (now obsolete) noun bland ‘blend’ (Norde 2000), shows no signs of reduction to the present day.
Finally, and perhaps most importantly: saying that grammaticalization is a process does not imply that there are universal pathways of grammaticaliza- tion. Heine and Kuteva’s World Lexicon of Grammaticalization (2002) may seem a veritable treasure-trove of cross-linguistically similar grammaticaliza- tion changes, but it has to be noted that many of the cases quoted in this work have not been studied in suYcient detail to substantiate claims about such universal pathways. In languages with little or no written histories this has of course not been possible, but even some of the pet examples from well- documented Indo-European languages do not stand up to scrutiny. Enger (2002, 2003), for example, has shown that the well-known development of the Proto-Scandinavian reXexive pronoun *sik (3sg.acc) into a verbal suYx -s(t) with passive meaning in the modern Mainland Scandinavian languages, was not a simple development from free form to clitic to suYx, as is generally assumed (e.g. Hopper and Traugott 2003: 159V.). Enger convincingly argues that many Old Norse st-verbs are to be considered derived forms rather than inXected ones, which implies an unusual pathway, namely free form> clitic > derivational suYx > inXectional suYx.
Another case in point is the Modern Finnish conjunction kun, which has both temporal and causal meaning. On the basis of this homonymy alone, Traugott and Ko¨nig (1991: 197) and Heine and Kuteva (2002: 291, quoting Traugott and Ko¨nig) conclude that the causal meaning was inferred from the temporal one. From a recent paper on this conjunction however (Herlin and Kotilainen 2004), in which the authors did examine the history of kun, it becomes evident that the state of aVairs in Modern Finnish is not at all the result of an internal development following a universal semantic pathway, but the result of standardization, inXuenced by cognates from Indo-European
31 Traugott’s Wrst point, that changes need not occur, is beautifully depicted in the following quote from Kuryłowicz (1966: 174, cited in Lass 1997: 302): ‘Il en est comme de l’eau de pluie qui doit prendre un chemin pre´vu (goutte`res, e´gouts, conduits) une fois qu’il pleut. Mais la pluie n’est pas une ne´cessite´.’
languages. The causal meaning may very well have developed from the one- time comparative and similaritive functions of kun or its usage as an em- phatic particle. The comparative and similaritive meanings were restricted to the conjunction kuin, originally a variant form of kun. In addition, kun was assigned contrastive meaning. Thus an artiWcial distinction between kun and kuin was introduced by nineteenth-century Finnish language reformers, inspired by the same distinctions in Indo-European languages (primarily Swedish).32
What these three examples (and the list could easily be expanded)33show is that even in grammaticalization studies people tend to base their analyses on results (a superWcial comparison of initial and Wnal states) rather than on a detailed examination of the change itself. It is therefore more appropriate to speak of tendencies than of ‘universal pathways’, ‘laws’, or ‘principles’ (Traugott 2001: 1; Joseph 2003: 486f.; Fischer, Norde, and Perridon 2004: 13; see also section 1.6.3).
For degrammaticalization, Wnally, no one would wish to claim that there exist universal pathways, because degrammaticalization changes are rare and mostly unique to a single language. It is not even clear indeed whether degrammaticalization is a process. I will return to this issue in Chapter 7.34
1.6.2 Grammaticalization as a ‘theory’
The study of grammaticalization has frequently been referred to as ‘gramma- ticalization theory’ (Newmeyer 1998: 234V. and references there), which according to Newmeyer (1998: 240) merely reXects ‘carelessness of usage’. In what is probably the most vehement critique of grammaticalization to date, Janda (2005: 47) puts it as follows:
The fervency of grammaticalization ‘theorists’ makes grammaticalization studies feel almost like a religion, while criticisms by perceived outsiders provoke reactions suggesting that critics are heretics . . . he [i.e. Richard Janda] should, for the record, state: Grammaticalization phenomena exist, and occur frequently, but all valid
32 Compare Swedish medan ‘while’ (both temporal and contrastive), a¨n ‘than’ (comparative), and som (similaritive).
33 See for instance Detges (2004) and Tsangalides (2004).
34 This being said, I will continue to use phrases such as ‘X degrammaticalizes’ or ‘Y is a degrammaticalization’, assuming that most (if not all) readers will understand that these are meta- phors for hopelessly tiresome formulations such as ‘X is the result of changes brought about by (unintentional) actions of language users.’ When I use formulations such as ‘changing morphemes’, or ‘changing constructions’, by no means do I imply that language is conceived of as having a ‘ ‘‘trans- individual, trans-generational’’, (near-)immortal ‘‘life of its own’’ ’ (Janda 2001: 290), entirely beyond the control of language users (see Newmeyer 1998: 239 for remarks similar to Janda’s).
generalizations concerning them result from the interaction of other, primary linguis- tic elements, because grammaticalization is not a (single) thing, but the outcome of other processes, and so grammaticalization is neither ubiquitous nor monolithic . . . Many alchemists made non-optimal use of scholarly time due to distraction by the notion ‘philosopher’s stone’; grammaticalization ‘theorists’ have an equivalent obsession: Wxed grammaticalization ‘path(way)s’/ ‘chains’ as putative guarantees for accurate reconstruction.
Janda surely exaggerates, but I agree that the use of the term ‘theory’ may lead to confusion, for instance in Heine (2003b), who states both (32) and (33): (32) Grammaticalization theory is neither a theory of language nor of
language change; its goal is to describe grammaticalization, that is, the way grammatical forms develop through space and time, and to explain why they are structured the way they are. (Heine 2003b: 575) (33) As observed earlier, grammaticalization theory is a theory to the extent
that it oVers an explanatory account of how and why grammatical categories arise and develop. (Heine 2003b: 578)
Haspelmath (2004: 23) Wnds the criticism that grammaticalization is not a theory ‘partially justiWed’, since it is not a ‘theory in the sense of a well-deWned system of interconnected falsiWable hypotheses’. He also notes that grammati- calization researchers do not subscribe to a single monolithic theory, but that they try to understand and explain large classes of similar semantic and morphosyntactic changes. Therefore, Haspelmath argues, the term ‘theorizing about grammaticalization’ would perhaps be more appropriate.
To conclude: although grammaticalization is not a theory, its merits as a descriptive framework, I think, are beyond dispute. What is more, the relevance of grammaticalization studies is not conWned to the description of grammaticalization phenomena itself. For instance, it may feed other branches in linguistics, such as evolutionary linguistics, as is evidenced by the following quote from Tomasello (2003: 13f.):
Generative grammarians believe that the human species evolved a genetically based universal grammar common to all peoples and that the variability in modern languages is basically on the surface only . . . The alternative is the usage-based view, in which there is no need to posit a speciWc genetic adaptation for grammar because processes of grammaticalization and syntactization can actually create grammatical structures out of concrete utterances – and grammaticalization and syntactization are cultural-historical processes, not biological ones. Thus, it is a historical fact that the speciWc items and constructions of a given language are not invented all at once, but rather they emerge, evolve, and accumulate modiWcations over historical time as
human beings use them with one another and adapt them to changing communica- tive circumstances.
In section 2.7.3, I will return to the advantages of a usage-based approach over a generative approach to grammar change.
1.6.3 Changes and correspondences
If I drive from Edinburgh to London, make some stops for petrol, and take a brief trip east on the way to visit a friend in Cambridge, I can still be said (from the point of view of ‘the accomplishment’, or juxtaposition of initial and Wnal states) to ‘have driven from Edinburgh to London’. How I got there is another (kind of) story . . . We seem usually to be thinking of macro-stories when we talk about ‘change’; but the micro-stories are of enormous theoretical importance as well. (Lass 1997: 288)
The above quote nicely illustrates an important distinction in historical linguistics, namely the distinction between, in Andersen’s (2001: 228) terms, ‘diachronic correspondences’, and ‘changes’. A diachronic correspondence is the ‘relation between homologous elements . . . belonging to two chronologic- ally separate synchronic states in a linguistic tradition’. Andersen makes a point similar to Lass’s when he notes that language historians often speak of ‘changes’ when what they mean are diachronic correspondences, which are not changes, but the results of change. ‘True’ changes, on the other hand, are ‘the historical events in a linguistic tradition by which practices of speaking vary over time’. Changes can be observed when they occur (though they often go unnoticed by members of a speech community) and they are reXected in synchronic variation or diachronic correspondences. The problem with much work on grammaticalization is that the distinction between changes and correspondences is not always observed:
To some extent, a basic problem here, as I see it, is that either by working from synchronic data and drawing even well-reasoned historical inferences (i.e., making a claim of grammaticalization based on internal reconstruction) or by comparing two stages somewhat distant in time and trying to infer what the pathways from one stage to the other were, one is falling into the trap discussed by Andersen . . . of confounding a diachronic correspondence with an actual innovation or change. While this is not a problem that is restricted to grammaticalization studies, it is a real one, and, speaking just impressionistically here, it is one that proponents of grammaticalization, with its appeal to universally applicable pathways of change that in essence do the work of historical investigation for one, seem rather prone to fall victim to. (Joseph 2004: 52)
Janda and Joseph (2003: 13) even propose a three-way distinction: ‘diachronic correspondence (juxtaposing two potentially non-adjacent times) versus in- novation (initiated by an individual person at one particular time) versus change (requiring adoption, over time, by all – or at least much – of a
group).’ Although it is unquestionably true that these three are essentially diVerent phenomena, it is not always easy, to put it mildly, to distinguish the latter two in diachronic linguistics. In most cases it is indeed impossible to detect innovation, since new features are generally not recorded until they have spread to other members of a speech community. (Who, for instance, was the Wrst speaker of English to use the reduced variant gonna instead of going to?) I will therefore take ‘change’ to comprise both innovation and spread.
An additional problem with complex phenomena such as grammaticaliza- tion and degrammaticalization is that they typically involve ‘micro-stories’ (in Lass’s terminology) on several linguistic levels (semantic-pragmatic, morphosyntactic, phonological). In the remainder of this work, I will there- fore use the term composite change to refer to grammaticalization and degrammaticalization changes, and primitive change to refer to a change at one linguistic level. A composite change, then, always involves several primi- tive changes. In Chapter 3, I will outline a classiWcational model to distinguish primitive changes in degrammaticalization, in order to avoid the ‘corres- pondence trap’.
This implies that, wherever possible, historical evidence of primitive changes must be provided. This is not standard practice, as noted by Joseph (2004: 47), who accuses much work on grammaticalization of being ‘ahistor- ical, not giving due consideration to the full range of information about the steps in a particular development and attempting to work out the history of various phenomena from synchrony alone’. This caveat obviously holds for languages with little or no written histories, but even for abundantly docu- mented languages such as Greek or the Romance languages. As an example of such avoidable oversimpliWcation I will discuss one of the pet examples in grammaticalization studies, which can be paraphrased as ‘the development of the Romance suYx -ment(e) from the ablative singular of the Latin noun mens ‘‘mind’’ ’ in section 1.6.5.
1.6.4 Reconstruction as evidence
For languages without written histories, changes can only be reconstructed. A clear example of reconstruction in grammaticalization studies is the case of person-number endings on the verb in Buryat Mongolian (cited in Hopper and Traugott 2003: 141, based on Comrie 1980), which bear a striking resem- blance to pronouns with corresponding person and number, as is shown in Table 1.2.
Such reconstructions are very common in theorizing about grammaticali- zation, but they have been heavily criticized, in particular for their potential
circularity. Newmeyer (1998: 279V.), for instance, points out that it is inad- missible to use reconstructed pathways as ‘evidence’ for universal pathways, which subsequently serve as the basis for further reconstruction. According to Janda (2005: 46), reconstruction forces the linguist ‘to understand the data-rich present based on the lacunar past’. And for once, Lehmann (2004: 156) appears to agree with grammaticalization critics when he writes: ‘Such cases of diachronic variation where [an earlier form] is reconstructed do not count as historical evidence either.’