Preamble1
The notion of grammaticalization has generated a huge amount of scholarly interest in the past 30-plus years,2 and will surely continue to do so for many years to come. I offer here my own, admittedly perhaps idiosyncratic, perspective on grammaticalization, examining its very nature and discussing some key aspects of this framework for examining language change.
I start with a fairly neutral characterization of grammaticalization, focusing on what happens in such developments, that is, on grammaticalization as a linguistic effect:3
the embedding into grammar, i.e., the taking on of grammatical status, of once-non-(or less-)grammatical phenomena.
In this characterization, it should be noted that I do not say “the process by which once-non-(or-less-)grammatical phenomena become embedded into grammar”. I do this deliberately, since it is my firm belief that one thing that grammaticalization is not is a process; in this view, which admittedly is not a position shared by all linguists, grammaticalization is not seen as a mechanism of change.4
Rather, for me, grammaticalization of this sort is a result, a state that comes into existence through the action of well-known and well-recognized processes/mechanisms of change, in particular regular sound change, analogy, borrowing (taken as a cover term for language contact effects more generally), and metaphorical extension in meaning.5
The term “grammaticalization” is also used as a cover term for “grammaticalization studies”, i.e. for work done within a framework that takes grammaticalization as its leading idea, and for a methodology that informs the research done within that framework.
In what follows, I elaborate on these points, drawing on material and viewpoints that I have published elsewhere — see especially Joseph 2004, 2011a, 2014, and 2017 — but I attempt to present them here in a fresh manner. Moreover, this presentation allows for different observations to be gathered together in one place. My ultimate goal, as the title suggests, is to shed light on and clarify just what grammaticalization is and what it is not.
1. Grammaticalization as a methodology
As noted above, the notion of grammaticalization can be seen as a guide to a methodology for understanding certain types of grammatical change. Various works in grammaticalization — see especially Hopper & Traugott 2003 — lay out a set of principles that inform analyses and give ways in which accounts of grammatical change can be developed. In this way, it is possible to take the view that grammaticalization is essentially a type of internal reconstruction.
To understand what internal reconstruction is, consider this somewhat informal scenario (see also Joseph 2010):6 you walk into a classroom and notice that the desks and chairs are all in different places and arranged differently from how they were when you were last in the room; how is this situation to be explained? One could suppose that the movement of the furniture could have been caused by the intervention of aliens, or by the effects of a windstorm; alternatively, the chairs could have staged a rebellion against the desks that had been oppressing them, or another teacher could have rearranged the furniture in the classroom in order to offer a setting for a movie or to stage a play or simply to promote discussion in their class. In trying to decide among these possibilities, it must first be admitted that all are conceivable scenarios under certain assumptions; crucially, though, not all of them are equally plausible, and in fact, some of these can be ruled out rather easily as they involve improbable or even impossible developments in order to be effected. In particular, chairs are simply not capable of holding the propositional attitudes or carrying out the actions necessary for staging a rebellion, and visits by extra-terrestrials are highly unlikely and even if extra-terrestrials do exist and did visit Earth, one has to ask why they would bother to pull a prank like changing around the furniture. The remaining possibilities have greater degrees of probability. That is, an exceedingly strong windstorm could certainly wreak havoc in a room if windows were left open or were blown out, but that is not a likely event, and in any case, an absence of broken glass would allow one to eliminate that possibility. This reasoning leaves the best hypothesis remaining as the one that explains the alterations as the result of human intervention, actions by another instructor sometime before your (re-)entry into the classroom.
The basic structure of this exercise is that of trying to deduce the historical events that led to a particular synchronic state, guided by our sense of what sorts of events are likely and unlikely to have created the observed synchronic state of affairs. This is a mode of inference that we engage in all the time: we see a puddle or wet pavement in the morning and can hypothesize that it rained overnight even without directly experiencing the rain; we see a friend’s hair in disarray and guess that he had run out of shampoo, had a turn with a hostile barber, or perhaps just lost his comb; we see a colleague yawning and seemingly going through the motions in the workplace and guess that he did not sleep well or enough the night before. And so on.
In each case we are attempting to reconstruct some aspect of the past that is not directly observable but which, rather, is inferable from the outcome and from what we know about how such outcomes generally arise. Applied to language, this mode of reasoning gives us internal reconstruction, hypothesizing about the causal historical underpinnings to a particular configuration of facts in a language, with the evaluation metric for the most reasonable hypotheses being whether they are supported by what is known about language and about language history in general.
In any historical investigation, including grammaticalization-based explorations of how some grammatical element has developed, we generally have to fill in gaps in the historical record. No matter how detailed the historical record is, there almost always are gaps in coverage that in the absence of direct evidence, including evidence of a comparative nature from other dialects or other languages, require the inferential reasoning that amounts to internal reconstruction. William Labov 1994:11 refers to this aspect of methodology when he states that historical linguistics is the “art of making the most of bad data” (though cf. Janda & Joseph 2003:14 suggest that one should substitute “imperfect” for “bad”!).
Typically, one has an inkling, an idea, that a given morpheme, e.g. a suffix, might have something etymologically/historically to do with some other morpheme, e.g. a free word, and then one uses that as a starting point for further investigation to fill in the details of the development. An example from Spanish illustrates this: adverbial marker -mente as in claramente ‘clearly’ appears on the face of it, due to the similarity in form, to have something to do with the free word mente ‘mind’.7 A full account, however, needs to go beyond such a simple observation of a possible connection and to try to motivate the connection in question and work out the steps that led to the particular formation and formative. A guess here would be that phrasal forms like clara mente, literally ‘with a clear mind”, i.e. ‘clearly’, would be a way of linking the adverbial use of the bound form with the lexical use of the free form. It turns out that such phrases do occur in adverbial usage in Classical Latin, with mente clearly being the ablative singular form of the stem ment- ‘mind’.
How might one then strengthen this hypothesis? Helpful here would be the recognition that the movement from a free word to an apparent affix via a phrasal use is a common development cross-linguistically; this is really the essence of a grammaticalization account. Moreover, positing such a development here would be in accord with the “principle of unidirectionality” that the grammaticalization framework offers as a condition on grammatical change; this principle states that movement in grammatical change is always from lexical to grammatical, e.g. in this case from free word mente in a phrase to grammatical formative -mente as a word-level formative. This scenario would then represent an internally arrived at reconstruction of the development leading to modern adverbs like claramente, internal reconstruction in that it draws entirely on data from within Spanish, augmented by a sense of how grammatical change in general can proceed.
So far so good – but then reality sets in. In particular, Old Spanish reveals not just phrases like clara mente, but also clara miente, clara mientre, and a few other variants. Furthermore, from Latin mente, the regular expected outcome in later Spanish is miente (as in some of the Old Spanish forms); compare modern Spanish miento ‘I lie’ from Latin mentior ‘I lie’ (infinitive mentīri) or pierdo ‘I lose’ from the Latin root perd-.
These Old Spanish facts mean that the -ment- form, in both the free word and in the adverb, is not the direct lineal development out of Latin ment-, neither in its free form nor in its phrasal combination that yielded the adverbial usage. Rather, ment- must represent a learnèd borrowing into Spanish that replaced the regularly developing (-)miente (the -r- forms like mientre have a different story involving analogical influence of other adverbs in -r-). Moreover, the learnèd borrowing that interrupted the “flow” from the phrasal combination of Latin to the adverbial form of this variety of later Romance in a sense started the devolution from word to affix via phrase all over again in the Middle Spanish period; that is, there was not a straight-line unidirectional movement from word to affix, but instead one that got interrupted by the re-introduction of a newly borrowed form into the adverbial. So the reality is somewhat different from what the internal reconstruction hypothesis based on modern Spanish -mente/mente might suggests.8
Thus, internally arrived at reconstructions can be wrong, but that is all right methodologically because, in a sense, internal reconstruction is nothing more than a hypothesis-generating procedure; that is to say, it is not a way of giving definitive answers but rather a way of offering possibilities to be explored further.
An example from Ancient Greek makes this clear. In particular, based on the following forms:
NOM.SG ónoma ‘name’
GEN.SG onómatos ‘of a name’
NOM.SG méli ‘honey’
GEN.SG mélitos ‘of honey’
NOM.SG poimḗn ‘shepherd’
GEN.SG poiménos ‘of a shepherd’
it is straightforward to segment the genitive forms as:
onómat-os
mélit-os
poimén-os
and then, working with an alternation between the stem form with -t and the nominative form without a -t to internally reconstruct the paradigms as having *mélit and *ónomat as the historical NOM.SG forms, with a final stop-deletion sound change giving the attested forms. At this point, it is appropriate to turn to external data in the form of comparisons from other Indo-European languages, specifically the word for ‘honey’ in Hittite and Gothic:
Hittite milit
Gothic miliθ
These external comparanda confirm the internally arrived-at reconstruction *melit, since they show a final dental obstruent.9
However, for ‘name’, the relevant comparison forms external to Greek are the following:
Hittite laman
Gothic nama
Sanskrit nāma
Latin nōmen
It is noteworthy that, unlike the word for ‘honey’, none of these forms have a final -t, suggesting that the internally generated reconstruction *ónomat is wrong. Interestingly, if one were to have taken a greater range of data from Greek itself into consideration from the outset of this exercise, a difference between the ‘honey’ word and the ‘name’ word might have been evident, as the denominative verbs formed from these nouns actually show different patterns: blíttō ‘to make honey’ (a regular outcome from *mlit-yō) versus onomaínō ‘to name’ (a regular outcome from a form that in Greek terms would have been *onomn̩-yō).
Thus caution is in order with internal reconstruction more generally, as it can lead to ahistorical hypotheses and accounts always need to be tested against as wide a range as possible of data, especially of an external kind. If grammaticalization accounts are essentially using the tenets of grammaticalization as part of an exercise in internal reconstruction, then the need for caution becomes evident in those accounts as well, as the Spanish (-)mente case illustrates. Grammaticalization as a methodology in reconstructing linguistic history must be viewed as generating hypotheses, not necessarily giving definitive answers.
2. Something grammaticalization is not
As a counterpoint to the previous section discussing something that grammaticalization is, I turn here to something that it is not. That is, grammaticalization is not a theory of morphology, or at least not much of a theory of morphology, despite the fact that it is avowedly interested in the origins of grammatical morphemes. I illustrate this claim with several case studies, showing what is needed for a viable theoretically based account of various morphological phenomena, in some instances with a direct comparison with a grammaticalization-based study.
2.1. “Phonogenesis” and English handiwork10
Hopper 1994 presents a case for what he terms “phonogenesis”, the devolution of a morpheme into mere phonological “bulk” in a word with no morphemic status, a state which for him represents an “advanced stage of grammaticalization” as it is at the far end of a cline of development for once-lexical or once-morphological/grammatical material.11 He offers up as a case in point the modern English word handiwork, defined by Merriam-Webster on-line (https://www.merriam-webster.com/dictionary/handiwork) as ‘work done by the hands; the product of handiwork’. In so doing, Hopper gives an implicit morphological analysis of the word.12
As to its history, handiwork derives from Old English hand-ge-weorc, where hand- is the stem for ‘hand’ and ge- is a collective prefix that combines with weorc to give a word meaning ‘a collection of work’, so that handgeweorc is ‘a collection of work (done by) hand’. For Modern English handiwork, though tri-morphemic from an etymological standpoint, Hopper argues that it is to be analyzed as built up of hand-, as the stem for ‘hand’, and work, as the stem for ‘work’, with the -i- being something that just “adds to the phonological bulk” of the word; that is, the -i- is synchronically non-morphemic.
It is useful to consider here the principles of American Structuralist morphological analysis in order to have a basis for deciding whether -i- is morphemic or not. In particular, in doing morphological analysis and identifying morphemes, in the structuralist framework, one is to look for recurring partials of form that match recurring partials of meaning; applying that to handiwork allows one to segment off hand- and -work in the modern form, leaving -i-. Another principle of analysis in this framework is also that there must be an exhaustive parsing of a word, such that all pieces of form must be identified and accounted for. It is not possible to simply omit -i- from analysis here and to treat it as having no analytic relevance. That is, one has to say that -i- is a morpheme, admittedly perhaps one with no clear meaning; note that if morphology focuses on form, then it should indeed be possible to have morphemes that are form only and have no readily identifiable meaning.13 Alternatively, one could say that -i- finds its meaning in what it contributes to the combination of hand + work, perhaps what allows for the difference between ‘work by hand’ in the abstract and the more concrete ‘product of work by hand’. Admittedly, the modern English -i- in this word does not do the same work, so to speak, as the Old English -ge- did, but it is a piece of the word that is identifiable, once one segments off hand and work, and therefore needs to be included in an exhaustive analysis.
Pushing the grammaticalization-based account here, and in particular, looking to characterize the historical changes in terms of the movement in a particular direction along a developmental cline of grammaticalization from morpheme to phonological bulk, thus leads to an unworkable analysis of handiwork.
2.2. Old English eom ‘am’
As a similar case involving details of morphological analysis with grammatical implications, some developments with the Old English first person singular (1SG) present of the verb ‘be’ are of interest. The facts in question have been presented by Hogg 1980, whose interest was in their relevance for the issue of grammar complication, and Joseph 2017 follows up in a similar vein. However, they have not been talked about specifically within a grammaticalization framework, so that there is no overt contrast to be made here as there is handiwork; nonetheless, there are some observations to be made that are relevant to the present discussion.
The form in question is the indicative eom ‘I am’. At one (diachronic or diatopic) stratum of Old English, this form was unanalyzable, inasmuch as there were no other first person singular forms in -m nor any other paradigmatically related forms of ‘be’ that would warrant segmenting eom as eo-m;14 in that way, eom was in the same situation morphologically as its modern counterpart, am. What Hogg draws attention to is the fact that nonetheless, eom was the source for an innovative 1SG subjunctive form, bēom, replacing earlier bēo, and possibly even from there to an innovative form of another partially rhyming word flēo ‘I flee’, thus giving flēom. Quite reasonably, Hogg treats the spread of m innovatively into bēom and flēom as a case of analogy, perhaps aided in part by it being a near-rhyming partner to the forms influenced by eom.
A question to be asked here now is what the status of eom is in terms of its morphological structure at the point at which the innovative subjunctive form bēom (and later flēom) arose analogically. The answer is that clearly, once there is beom, and especially when there is also fleom, the principles of morphological analysis described in the previous section (§3.1) would require that the –m be segmented off as a variant 1SG marker, given that it is a recurring element of form in these three verbs that correlates with a recurring element of meaning, namely ‘first person singular’. Moreover, forms such as bēon for the subjunctive plural (all persons) show that the segmentation of the innovative bēom, once it has come into existence analogically, is bēo-m, so that, morphologically speaking, eom would likewise be segmented as eo-m, The relevance of this analysis for grammaticalization is taken up in §4.
2.3. Innovative verbal inflection in Modern Greek
As a final case-study, I review the relatively well-known facts of an innovative verbal paradigm in Modern Greek, discussed in Thumb 1912:§239 and Householder & Nagy 1972:44, and specifically tested against claims made within a grammaticalization framework in Joseph 2014. The facts in question concern the present-tense paradigm of verbs that are end-stressed, i.e., accented on the final syllable.
The older present tense paradigm for an end-stressed verb with a-vocalism in the endings, such as ‘ask’, is as follows:
1SG rotó
2 rotás
3 rotá
These forms can be segmented in the following way:
1SG rot-ó
2 rot-ás
3 rot-á
The innovative paradigm is:
1SG rotáo
2 rotáis
3 rotái
which, as Thumb (op. cit.) notes, is based on the addition of the endings seen in root-stressed verbs, e.g. kan- ‘do’:
1SG káno
2 kánis
3 káni
to the 3SG form of the older paradigm, treated as if it were a stem with a zero ending (i.e. rota-ø). Given the co-existence of the root-stressed and the innovative end-stressed paradigms, the segmentation of the innovative paradigm would be:
1SG rotá-o
2 rotá-is
3 rotá-i
Thus –a, originally the 3SG ending, has become part of the stem, and the paradigm has been reconstituted with the productive endings from a related verb class.
As Householder & Nagy (op. cit.) note, this development can be subsumed under the phenomenon known in the literature as “Watkins’ Law”, after the work of Calvert Watkins and especially his 1962 analysis of the origins of some Old Irish past tense formations.15
These developments give rise to another question, namely what is the status of –a- in the older paradigm and in the innovative one? As indicated above, in the older paradigm, the segmentation is:
1SG rot-ó
2 rot-ás
3 rot-á
with –a as the 3SG ending; this segmentation is indicated by the singular paradigms for other end-stressed verbs (e.g., boró / borís / borí ‘I/you/(s)he can’),16 the root-stressed verbs like káno (given above), and past tense forms (e.g., ékana / ékanes / ékane ‘I/you/(s)he did’). Since the elements that spread are the root-stressed endings -o / -is /-i, the -a- of 3SG rotái is not part of the ending in the innovative paradigm; rather, -o / -is / -i are the endings and rota- is the stem.
This a of the stem rota-, in one interpretation, could be taken to be just a part of the stem, mere phonological material, perhaps even “phonological bulk”, to use Hopper’s term. However, upon consideration of more data, it turns out that this a has morphemic status, though different from its earlier status as an inflectional ending. In particular, the perfective stem of ‘ask’ (e.g. in the aorist past tense) is rot-is- (e.g. 1SG rót-is-a ‘I asked’); thus –a- in the new stem rotá- can be segmented off as the (present) imperfective stem marker. That is, morphologically there is a contrast between imperfective stem rot-a- and perfective rot-is-. In this view, -a- is a derivational suffix, deriving the present imperfective stem, and is parallel to other suffixes deriving imperfective stems, e.g. -íz- as in arxízo ‘I begin’ (vs. perfective -is- as in árxisa ‘I began’) or -n- as in línume ‘we solve’ (vs. perfective -s- as in lísame ‘we solved’).17
3. Consequences for claims within grammaticalization and otherwise
These three mini-case studies provide different points that are relevant to concerns about grammatical change more generally and about grammaticalization more specifically. First, though, to sum up what these case-studies show:
a) the example of handgeweorc/handiwork in §3.1 shows that there has actually been little morphological change if any in that word; the meaning of the -i- has changed from what it was earlier but the status of the piece between hand and weorc/work as a morpheme remains the same;
b) the example of eom in §3.2 shows that a word which is unanalyzable at the stage where it is isolated, with no support for -m as a segmentable element and thus presumably stored in the lexicon as a whole, can gain internal segmentation; once bēom/flēom arose, eom was parsable into eo-m, with -m as a separate, even if restricted, morpheme, so that a nonmorpheme has become a morpheme;
and,
c) the example of rotó/rotáo, etc., in §3.3 shows that the -a- in the innovative paradigm is part of the present stem so that –a has gone from being an inflectional ending to, in one interpretation, being just a part of the stem, mere phonological material, while, in another interpretation — that favored here — it is actually a stem-deriving, i.e. derivational, suffix (see footnote 17).
In the subsections that follow, the consequences of these developments are explored for understanding the way languages change over time. Individually and collectively they bear on the question, discussed further in §5, of the extent to which grammaticalization studies offer a full accounting of all that is relevant to the study of language change.
3.1. The source of grammatical morphemes
Hopper & Traugott 1993/2003 claim that there is a lexical source for all grammatical morphemes. The case of eom shows that claim to be too strong, since a grammatical morpheme, here -m, has a nonlexical source; in this case, the spread of -m by analogy, together with the principles of morphological analysis, creates a situation in which -m becomes a grammatical morpheme where previously it was just phonological material (“phonological bulk”, in Hopper’s 1994 terms) prior to the analogies (see also Lass 2000 and Fischer & Rosenbach (2000: 21)).
3.2. Unidirectionality
A cornerstone of grammaticalization is the claim that it is unidirectional, always moving in the direction of greater grammaticality.18 The Greek case of rotó/rotáo (etc.) offers a direct challenge this claim. In particular, the older paradigm has -a- in rota- as inflectional, as it represents the 3SG ending on its own and is part of the 2SG ending, while the innovative paradigm has -a- in rotá- as a derivational element, a suffix that creates imperfective stems. In most formulations of the degree of grammaticality, inflection is considered to be more grammatical, more embedded in the grammar, than derivation, which typically is more lexical in nature and therefore less grammatical. Kurylowicz 1965, for instance, explicitly holds to this view (and see also footnote 17). This means that the innovation leading to forms like rotáo is a counterexample to the principle of unidirectionality since the movement has been up the cline of grammaticality, going from inflection to derivation, rather than down the cline.
However, the principle of unidirectionality suffers more here than just the one counter-example. In particular, it has been said that some apparent counterexamples do not involve grammaticalization but rather are instances of lexicalization, the creation of new lexical items; the extraction of a suffix -ism from words like capitalism and communism and its upgrading to independent word status, as in a phrase like socialism and other isms, or –stan, as in the stans of central Asia, or -ologist, for medical specialist, based on cardiologist, gynecologist, etc., would be instances of that sort.19 One might wonder then if the development of a stem rota- would not fall within the realm of grammaticalization but would rather be a matter of lexicalization, and as such would not be a counterexample to any claims about grammaticalization proper. That view, however, cannot be maintained because the stem rota- is not a new independent lexical item but rather a representative of the lexeme for ‘ask’, and moreover it does have some nonlexical value, being involved in the formation of a present/imperfective stem; it just is not as grammatical as an inflectional ending is.
Therefore, this example can be added to the list of true counter-examples to claims about unidirectionality in grammatical change, at least the eight that Haspelmath 2004 recognizes and the several that Norde 2009 argues for in her book-length study of degrammaticalization, i.e. of counter-directional developments, as well as others that have been offered in the literature. Furthermore, this rota- example from Greek poses an even more significant problem since it is often claimed that counter-examples such as those recognized by Haspelmath exhibit no pattern and are sporadic and unsystematic; however, there are numerous Watkins’ Law examples that have been discussed in the literature (see footnote 15) and each one could lend itself to the same sort of morphological analysis as with rotá-, so that this particular instance of degrammaticalization is thus one token of a more general — and thus systematic — type of counter-example to unidirectionality.
3.3. More on unidirectionality: Something out of nothing
With regard to eom, it is observed above in §4.1 that the m in eom at one stage in Old English was non-morphemic, thus something that Hopper might well call just “phonological bulk”. In such a view, it would reflect, as noted in §3.1, an “advanced stage of grammaticalization”, having devolved from a fully morphemic element in Proto-Indo-European (see footnote 14) to a non-meaning-bearing piece of the phonological make-up of a word. That means, however, that in then moving to morphemic status, supported by the innovative bēom and flēom, it would have moved counter-directionally, counter to the usual direction associated with changes in grammaticality. Moreover, to the extent that it is mere phonology (Hopper’s “bulk”) at the point prior to becoming a morpheme, it would be an instance of the development of something, i.e. an element of form with meaning, out of essentially nothing, i.e. something that is just an element of sound. It would thus be the kind of development that Joseph & Ralli 2022 characterize as a curious combination, namely degrammaticalizing grammaticalization — grammaticalization because a new grammatical element, a person/number ending on a verb, is created, but degrammaticalizing because in coming into existence as a grammatical marker, it moves counter to unidirectionality, going from Hopper’s “advanced stage of grammaticalization” (see §3.1) to a more weakly grammatical derivational morpheme.
3.4. A bonus for language change in general
As a bonus not specifically connected to grammaticalization but relevant for language change in general, the case of eom allows for an insight into the issue of whether language change is gradual. Change is often said to take place gradually, devolving over long periods of time, and that may certainly be true for spread of a change either through the lexicon, as in the case of analogically driven changes, or through a speech community, in the classic Labovian paradigm for change. However, the emergence of a new inflectional morpheme -m in eo-m / bēo-m / flēo-m comes about abruptly, as is generally the case, it must be admitted, with instances of reanalysis, without passing step-wise through a stage of being lexical or derivational or the like; that is, once there is bēom, to be segmented bēo-m as argued above, the analysis of once non-morphemic eom into eo-m follows automatically, given the principles of morphological analysis discussed above in §3.1. This example, as well quite possibly as the rota- case, therefore, demonstrates that language change need not be gradual.20
4. More on what grammaticalization is not
If grammaticalization practice is best thought of as a type of internal reconstruction then it is not really a theory of grammar in any full sense. Admittedly the issue of what a “theory” is and the complaints made, e.g. by Newmeyer 2001, about “grammaticalization ‘theory’” are maybe ill-aimed, in that it is not clear that anyone has made representations that it is a synchronic theory of grammar. Nonetheless, one can note links between the framework of grammaticalization and notions like “emergent grammar”.
More to the point for language change, I argue that we should not look to grammaticalization for a total theory, or even a total account, of all that is relevant to the study of language change; thus, to sound a terminological note, using “grammaticalization” as a near-synonym for “change”, something I have observed on occasion,21 seems to me to be a dangerous terminological step to take. That is, there is so much more to language change and historical linguistics than what the grammaticalization framework offers insights into. In particular, language can change at all levels, in all components, and given enough time, perhaps there is nothing in language that cannot change, except for the most foundational aspects of language, the universal properties that make human language what it is. As we all know, language, as traditionally viewed, encompasses phonology, morphology, syntax, semantics, and lexicon, as well as pragmatic interpretations given to the use of these elements in discourse. This means there are sounds, patterns to the sound combinations, prosody, combinations of sounds into meaningful units we call morphemes, patterns for the combination of morphemes (word-formation rules), and so on. All of the elements, all of the abstractions that govern the patterning of these elements, all of the combinatoric principles, all of the processes of interpretation, and so on, are subject to change.
How much of all of this do grammaticalization and grammaticalization studies cover? In a narrow sense, grammaticalization deals with just those aspects that constitute change in the direction of creating “grammar”, by which is meant, to judge from what is focused on in the literature, the creation of function morphemes, though derivational elements and discourse markers are routinely treated in grammaticalization-based discussions. This is an interesting, indeed fascinating, area of investigation, but it is not all there is to language change or to the study of language in its historical dimension.
One area especially overlooked in grammaticalization studies is sound change, except via reference to phonological reduction or erosion of grammatical markers. But sound change in its classical sense is not really covered by grammaticalization. That is, within many grammaticalization studies one can see claims that particular grammatical morphemes are subject to changes in their phonological realization that are not regular sound changes in the Neogrammarian sense of being purely phonetically determined and exceptionlessly applying to all instances that meet a particular phonetic context. I single out sound change here because it is the historical foundation of the field, both the field of historical linguistics narrowly and the modern field of linguistics more broadly. But, there is even more in historical linguistics that is not covered within grammaticalization studies, as the field takes in such sub-areas as language relatedness, linguistic contributions to human prehistory (e.g., via a Wörter-und-Sachen approach), philology (text interpretation), and etymology, among other topics. Admittedly those working within the framework of grammaticalization often employ the results and occasionally the methods of these areas of investigation, but not to the particular end that these sub-areas point towards. This situation has led me to wonder if the intense interest that grammaticalization studies show in one type of historical grammatical development might perhaps be skewing the field of historical linguistics overall, leading me to ask in Joseph 2004 whether historical linguistics in general can “survive” grammaticalization. I for one feel that we cannot really afford to give up on the foundational notion of regularity (and phonetic-only conditioning) of sound change, for if we fail to adhere to Neogrammarian doctrine and adopt the grammaticalizationist view of construction- or lexeme-specific sound change, it is not clear what we would be left with. Maintaining the Neogrammarian view of the “exceptionlessness” (Ausnahmslosigkeit) of sound change provides a principled basis for making decisions about possible analyses, and that is too valuable a tool to give up.
5. Conclusion
By way of a conclusion, for all the criticism I have offered here, I readily admit that grammaticalization, when viewed properly and when suitably enhanced with notions from traditional historical linguistics, enriches our field. Nonetheless, there is an important correlate to the position taken at the outset, in §1, of grammaticalization as a result, not a process. That is, one can develop an Occam’s Razor-style argument, one based on parsimony, for treating grammaticalization as the result of independently needed processes of change, rather than taking it to be a process on its own. That is, the traditional triad of regular (Neogrammarian) sound change, analogy, and borrowing, using this last simply as a cover term for a much more complex “universe” of contact-based effects22 and taking semantic change to be essentially metaphor and thus essentially analogical in nature, gives us all we need in the way of mechanisms of change. Adding grammaticalization to these as a further mechanism adds something extraneous, something that duplicates the effects of these mechanisms. Thus it is not needed as a separate and distinct mechanism, or process, of change and is best taken to be a result of the workings of these independently needed processes. Moreover, this triad is satisfying since it corresponds precisely to the different dimensions of language: language is physiological in nature, realized through our speech organs, and Neogrammarian phonetically determined sound change corresponds to that dimension; language is cognitive in nature, residing in and emerging out of our brains, and analogy, as a basic cognitive tool for understanding the world, corresponds to that dimension; and language is social in nature, existing in the interactions we have with other speakers, and borrowing, representing contact effects more generally, corresponds to that dimension. Adding grammaticalization to this roster would not, it seems to me, correspond to any particular dimension of language and would duplicate what the traditional triad offers.
References
Anderson, Stephen R. 1992. A-Morphous Morphology. Cambridge: Cambridge University Press.
Arlotto, Anthony. 1972. Introduction to historical linguistics. Boston: Houghton Mifflin.
Borer, Hagit. 1984. The projection principle and rules of morphology. Proceedings of the North Eastern Linguistics Society 14.16-33.
Bowden, John. 1992. Behind the Preposition: Grammaticalisation of Locatives in Oceanic Languages. Canberra: Dept. of Linguistics, Research School of Pacific Studies, Australian National University.
Campbell, Lyle, ed. 2001. Grammaticalization: A Critical Assessment (special issue of Language Sciences 23.2–3).
Collinge, Neville E. 1985. The laws of Indo-European. Amsterdam: John Benjamins.
De Smet, Hendrik & Olga Fischer. 2017. The role of analogy in language change: Supporting constructions. The Changing English Language: Psycholinguistic Perspectives, ed. by Marianne Hundt, Sandra Mollin, & Simone E. Pfenninger, 240-268. Cambridge: Cambridge University Press.
Diewald, Gabrielle. 1997. Grammatikalisierung: Eine Einführung in Sein und Werden grammatischer Formen. Tübingen: Niemeyer.
Fischer, Olga. 2000. Grammaticalisation: Unidirectional, non-reversable? The case of to before the infinitive in English. In Fischer et al., 149-170.
Fischer, Olga 2009. Grammaticalization as analogically driven change? VIEW[Z]. Vienna English Working Papers 18, 2: 3-23.
Fischer, Olga. 2011. Grammaticalization as analogically driven change? Handbook of Grammaticalization, ed. by Heiko Narrog & Bernd Heine, 31-42. Oxford: Oxford University Press.
Fischer, Olga & Anette Rosenbach. 2000. Introduction. In Fischer et al, 1-38.
Fischer, Olga, Anette Rosenbach, & Dieter Stein, eds. 2000. Pathways of Change: Grammaticalization in English. Amsterdam: John Benjamins.
Givón, Talmy 1971. Historical syntax and synchronic morphology: An archaeologist's field trip. Papers from the Regional Meeting of the Chicago Linguistic Society 7.394-415.
Haspelmath, Martin. 2004. Directionality in language change, with special reference to unidirectionality in grammaticalization. Up and down the Cline, ed. by Olga Fischer, Muriel Norde, & Harry Perridon, 17-44. Amsterdam: Benjamins.
Heine, Bernd. 1993. Auxiliaries: Cognitive Forces and Grammaticalization. New York: Oxford University Press.
Heine, Bernd. 1997. Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge: Cambridge University Press.
Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer. 1991. Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press.
Heine, Bernd & Tania Kuteva. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press [1st edition].
Hogg, Richard. 1980. Analogy as a Source of Morphological Complexity. Folia Linguistica Historica 2: 277-284.
Hopper, Paul J. 1994. Phonogenesis. In Pagliuca, pp. 29–45.
Hopper, Paul J. & Elizabeth C. Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press.
Hopper, Paul J. & Elizabeth C. Traugott. 2003. Grammaticalization. 2nd edition. Cambridge: Cambridge University Press.
Householder, Fred & Gregory Nagy. 1972. Greek: A survey of recent work. The Hague: Mouton.
Janda, Richard D. 2001. Beyond “pathways” and “unidirectionality”: on the discontinuity of language transmission and the counterability of grammaticalization. In Campbell, pp. 265–340).
Janda, Richard D. & Brian D. Joseph. 2003. On Language, Change, and Language Change — Or, Of History, Linguistics, and Historical Linguistics. In Joseph & Janda, pp. 3-180.
Janse, Mark. 2009. Watkins’ Law and the development of agglutinative inflections in Asia Minor Greek. Journal of Greek Linguistics 9: 32-48.
Jasanoff, Jay & Brian D. Joseph 2015. Calvert Ward Watkins. Obituary. Language 91.1:245-252.
Joseph, Brian D. 1980. Watkins’ Law and the Modern Greek Preterite. Die Sprache 26: 179-184.
Joseph, Brian D. 2001. Is There Such a Thing as ‘Grammaticalization’? Language Sciences 23: 163–186.
Joseph, Brian D. 2003. Morphologization from Syntax. In Joseph & Janda, pp. 472-492.
Joseph, Brian D. 2004. Rescuing Traditional (Historical) Linguistics from Grammaticalization "Theory". Up and Down the Cline – The Nature of Grammaticalization, ed. by Olga Fischer, Muriel Norde, & Harry Perridon, 44-71. Amsterdam: John Benjamins.
Joseph, Brian D. 2006. How accommodating of change is grammaticalization? The case of "lateral shifts". Logos and Language. Journal of General Linguistics and Language Theory 6.2: 1-7.
Joseph, Brian D. 2010. Internal reconstruction. The Continuum Companion to Historical Linguistics, ed. by Silvia Luraghi & Vit Bubenik, 52-58. London: Continuum Publishers.
Joseph, Brian D. 2011a. Grammaticalization: A General Critique. Handbook of Grammaticalization, ed. by Heiko Narrog & Bernd Heine, 193-205. Oxford: Oxford University Press.
Joseph, Brian D. 2011b. Historical Linguistics and Sociolinguistics – Strange bedfellows or natural friends? Language and History, Linguistics and Historiography, ed. by Nils Langer, Steffan Davies, & Wim Vandenbussche, 67-88. Bern: Peter Lang.
Joseph, Brian D. 2014. What counts as (an instance of) Grammaticalization? Refining Grammaticalization ed. by Ferdinand von Mengden & Horst Simon. Special issue of Folia Linguistica 48.2: 1-23.
Joseph, Brian D. 2017. Morphological Change. The Cambridge Handbook of Morphology, ed. by Andrew Hippisley & Gregory Stump, 743-764. Cambridge: Cambridge University Press.
Joseph, Brian D. 2021. On the Relationship between Grammaticalization and Historical Comparative Linguistics. To appear in Grammaticalization and Studies of Grammar, Vol. X (The Commercial Press, Yichang, China)
Joseph, Brian D. & Richard D. Janda, eds. 2003. Handbook of Historical Linguistics. Oxford: Blackwell.
Joseph, Brian D. & Angela Ralli. 2022. Something out of nothing: Degrammaticalizing grammaticalization. To appear in Papers from the 24th International Conference on Historical Linguistics (ICHL24), Canberra. Amsterdam: John Benjamins.
Karlsson, Keith. 1981. Syntax and Affixation: The Evolution of MENTE in Latin and Romance. Tübingen: Niemeyer. (Beihefte zur Zeitschrift für romanische Philologie, 182.)
Kurylowicz, Jerzy. 1965. The evolution of grammatical categories, Diogenes 51: 55-71.
Kuteva, Tania, Bernd Heine, Bo Hong, Haiping Long, Heiko Narrog, & Seongha Rhee. 2019. World Lexicon of Grammaticalization 2nd edition. Cambridge: Cambridge University Press [Revision of Heine & Kuteva 2002].
Kuteva, Tania. 2001. Auxiliation: An Enquiry into the Nature of Grammaticalization. Oxford: Oxford University Press.
Labov, William. 1994. Principles of Linguistic Change. Volume I: Internal Factors. Oxford: Blackwell.
Lass, Roger. 2000. Remarks on (uni)directionality. In Fischer et al., 207-228.
Lehmann, Christian. 1982/1995. Thoughts on Grammaticalization. München: LINCOM Europa. [Originally published as Thoughts on grammaticalization: a programmatic sketch. V. 1. Köln: Institut für Sprachwissenschaft der Universität, 1982. (Arbeiten des Kölner Universalienprojekts; 48).]
Lehmann, Christian. 2015. Thoughts on Grammaticalization. 3rd edition. Berlin: Language Science Press [Revision of Lehmann 1982/1995].
Lessau, Donald A. 1994. A Dictionary of Grammaticalization (3 volumes). Bochum: Universitätsverlag Dr. N. Brockmeyer.
Murtoff, Jennifer. 1999. Spanish adverbs in -mente. Paper presented in Diachronic Morphology Workshop, The Ohio State University, 1 June 1999.
Newmeyer, Frederick J. 2001. Deconstructing grammaticalization. In Campbell, pp. 187–229.
Norde, Muriel. 2009. Degrammaticalization. Oxford: Oxford University Press.
Pagliuca, William, ed. 1994. Perspectives on Grammaticalization. Amsterdam: John Benjamins.
Ramat, Anna Giacalone & Paul J. Hopper, eds. 1998. The Limits of Grammaticalization. Amsterdam: John Benjamins.
Ringe, Don. 2003. Internal reconstruction. In Joseph & Janda 2003, pp. 244-261.
Thumb, Albert. 1912. Handbook of the Modern Greek vernacular. Grammar, texts, glossary. (Translated from the second improved and enlarged German edition (1910) by S. Angus.) Edinburgh: T. & T. Clark.
Traugott, Elizabeth C. & Bernd Heine, eds. 1991. Approaches to Grammaticalization (2 Volumes). Amsterdam: John Benjamins.
Watkins, Calvert W. 1962. Indo-European origins of the Celtic verb. vol. I. The sigmatic aorist. Dublin: Institute for Advanced Studies.
Wischer, Ilse & Gabriele Diewald, eds. 2002. New Reflections on Grammaticalization. Amsterdam: John Benjamins.
Ziegeler, Debra. 2000. Hypothetical Modality: Grammaticalisation in an L2 Dialect. Amsterdam: John Benjamins.
XML
