1Our starting point for this paper consists of the following commonly accepted facts, to be qualified below: (1) Antoine Meillet is believed to have coined the term “grammaticalisation” in 1912. (2) He thus referred to the phenomenon which converts lexical elements (items or constructions) into (more) grammatical ones. This phenomenon had been described before, of course: the idea that grammatical elements were originally lexical was not new. Indeed, it had been known and described for decades, if not centuries, as shown for instance in other articles of this special issue (see namely Haβler 2025; Pagani-Naudet 2025; McElvenny 2025). But (3) Meillet’s 1912 paper was the starting point for a wealth of papers, starting mostly in the 1980s.
- 1 In the “grammaticalization revival” of the late 1970s (see e.g. Butler 1977, Traugott 1980, Vincent (...)
2Our research question is the following: how pivotal was Meillet’s paper in the history of grammaticalization? In other terms, how much does Meillet’s work coincide with, or depart from, that of his predecessors and contemporaries? And how influential was his 1912 paper before the end of the century, at a time when it was generally agreed that Meillet had coined the term (at least since Lehmann 1982),1 and in doing so potentially opened up this new field of investigation?
3In order to answer this question, we delve further both into the background of Meillet’s work, and into the story of grammaticalization. We retrace the story of both the word and the concept of grammaticalization, with a focus on Meillet, who has undoubtedly had a great impact on its development, to measure this impact more precisely. In order to do so, we combine two different approaches and perspectives on linguistics and language change, along with their respective methodologies: a metalinguistic—and mostly semasiological—perspective, and a historical—and mostly onomasiological—perspective. Both approaches build on a large corpus of papers in linguistics bearing at least partly on phenomena of grammaticalization, regardless of the term they use to describe them. In a semasiological perspective, part of this corpus is used to track occurrences of the term “grammaticalization” and its cognates. The method, therefore, consisted in gathering and analyzing occurrences of the term “grammaticalization” (and its cognates) in an international, multi-lingual corpus.
4Before exploring this corpus, our first concern will be to describe the historical context which prevailed when Meillet introduced his concept of grammaticalisation. Since Meillet both advocates a return to Bopp and refrains from mentioning the latter’s Agglutinationstheorie, we need to understand what separates and what unites the two scholars, and what had changed when Meillet’s grammaticalisation entered the stage. This leads us to present the background to Bopp’s theory, which will hopefully enable us to see in what ways it was contingent on a theoretical environment that was definitely dated at the time of Meillet. Next, we present Meillet’s grammaticalisation and discuss some important facets of this concept, such as the puzzling notion of mot accessoire “accessory word,” and the role of force expressive “expressive force” in diachrony. The latter tenet is in line with Bally’s stylistic approach.
- 2 The query “grammaticalization,” with or without caps, run on November 9, 2024, yielded 90,000 hits; (...)
5In order to better understand Meillet’s impact on the field, we further explore the fortunes of the term grammaticalization, before and especially after Meillet. Our results not only confirm that grammaticalization studies predated Meillet, but also show that the term itself actually predated Meillet. Beyond this, our results show that the term grammaticalization had been used widely by linguists well before it became more widely popular, in the 1980s. Among the early borrowings of Meillet’s grammaticalisation by other scholars, some figure prominently in our corpus of occurrences. We shall pay special attention to two of them: Vossler and Spitzer, whose analyses have a connection to Meillet and Bally’s emphasis on expressivity, and have the additional benefit of discussing an aspect neglected by contemporary linguistics, namely the importance of stylistic matters. Our data enable us to follow the peculiar journey of grammaticalization as both term and concept, from past centuries all the way to the eve of the 1960s, i.e. when it became truly commonplace in linguistics. Indeed, grammaticalization has been the object of countless studies in the past decades. For instance, a query for the English term grammaticalization on the Google Scholar website returns 90,000 hits; even restricted to papers published in the last 4 years, it yields well over 15,000 hits.2
6In our paper, we first describe Meillet’s definition of the term grammaticalisation. We go on to investigate specific aspects of Meillet’s research, namely his focus on the notion of force expressive “expressive force,” the Boppian context and the importance of the notion of force expressive. We then describe the historical backdrop to Meillet’s research, especially the importance of Bopp, and the use of grammaticalization after Meillet, with a special emphasis on the aesthetic perspective and on Spitzer in particular. We then turn to the role of Meillet and Meillet’s 1912 paper as a turning point in grammaticalization studies. In order to do this, we report on the more semasiological aspects of our corpus study, aiming to analyze how the word grammaticalization has been used in the past centuries and to see whether or not its modern use is compatible with, and indeed goes back to, Meillet’s definition. We describe the methodology of this corpus study, and report on the results, highlighting the importance of Romance linguistics—and of a number of Romance linguists—in the emergence of grammaticalization as one of the main tenets of modern linguistics, before summing up.
- 3 The first coinage is due to Richard de Radonvilliers, who lists in his dictionary cognate words rel (...)
- 4 In Meillet’s 1906 paper on semantic change, instances of what would later be described as grammatic (...)
7The term grammaticalisation, though not first coined by Meillet, appears to enter into usage with his 1912 paper, entitled “L’évolution des formes grammaticales,” i.e. “The evolution of grammatical forms.”3 The term, however, is infrequent in Meillet’s 577-page long collection of papers (2015 [1921–1926]). Grammaticalisation occurs three times, and se grammaticaliser once, in only two papers: in the just mentioned “inaugural” text, and in a study on the renewal of conjunctions in Indo-European languages (1915). The words are conspicuously absent in texts where they would be expected, e.g. in a paper where Meillet describes a line of evolution that we may describe as a trend toward analyticity (not Meillet’s wording), a tendency that is, according to him, attested in Indo-European languages and beyond, but that is also reversible (Meillet 2015 [1918]). In Meillet’s parlance, inflections tend to be replaced by “mots accessoires,” for reasons having to do with the phonetic weakness of inflections, their lack of uniformity, the quest for more emphatic or expressive manners of speaking, and a striving toward greater semantic generality.4
8This brings us to the way Meillet defines and describes grammaticalisation in the 1912 paper. Meillet first states that grammaticalization consists in the transition transforming an autonomous word into a “grammatical element” (“passage d’un mot autonome au rôle d’élément grammatical,” Meillet 2015 [1912]: 209), or as imparting grammatical significance to the grouping of words (cf. the syntactic relevance of word order in English vs Latin, for example). The constitution of grammatical forms, says Meillet elsewhere (Meillet 2015 [1912]: 217), is made possible by the weakening of the pronunciation, the loss of the concrete meaning and that of the expressive value of autonomous words.
- 5 See Pagani-Naudet 2025 for a critical genealogy of this archetypal example of grammaticalization.
9Readers who would be offered the latter definition out of context would probably think of prime examples such as the formation of the Romance future.5 However, to this definition is added another consideration which pertains to the distinction between mots principaux and mots accessoires. Examples of mots accessoires include, unsurprisingly, auxiliaries, including venir in Il vient me dire cela ! (lit. “He comes to tell me that,” i.e. “He has the nerve to tell me that;” Meillet 2015 [1912]: 213, where venir does not have its spatial meaning). But examples of mots accessoires also include petits in les petits enfants, vs apportez le petit paquet, in which petit is in focus and contrasts with gros (Meillet 2015 [1912]: 212). We may note that for Meillet differential focus is at play in the differential attrition of the elements of hiu tagu “this day” in modern German heute (Meillet 2015 [1912]: 217): the deictic component has kept its phonetic load because, according to Meillet, it was in focus.
- 6 In chronological order, Condillac takes the first place in Lehmann’s short historical account of gr (...)
- 7 “Analytic” is a term that Meillet finds ill-advised, on the ground that it would suggest that speak (...)
10The large extension given to the notion of mot accessoire, and the very choice of the word accessoire in this context, cannot but evoke the notion of idée accessoire, whose meanings, especially in the tradition of grammaire générale, have ranged from “idea associated with an inflection or with complementation” (the base expressing an idée principale) to that of a distinctive feature separating synonyms (their idée principale consisting in their semantic overlap) and to that of an idea evoked by a lexical item, though not designated by it (Delesalle 2005). The morphological perspective illustrated by Girard, Beauzée and Condillac is of particular relevance for us, since inflections occupy center stage, with word order and conjunctions, in Meillet’s discussion and in studies on grammaticalization at large. Of the three, Condillac comes closest to Meillet’s diachronic perspective on the evolution of mots accessoires, if by this term we understand words expressing idées accessoires (e.g. Condillac 1775, chap. VIII: 72, chap. XIII: 79–80).6 The emphasis, however, is no longer on a speculative account of the origin of language, but on the evolution of reconstructed and attested forms. The sense of idée accessoire as a source for the formation of grammatical forms has its counterpart in Bopp, under the term of Nebenbegriff. Thus, for example, cases are said to express syntactic relations via the original Nebenbegriff of a relation in space (Bopp 1833–1852, §115: 245). In his translation of Bopp’s Vergleichende Grammatik, Bréal renders Nebenbegriff as idée accessoire (Bopp 1866, §115: 275). What may underlie this use of the concept of mot accessoire is also the observation that, according to Meillet, Indo-European languages tend toward greater analyticity, for example adopting “compound” preterites in place of simple ones (cf. French j’ai aimé vs j’aimai; Meillet 2015 [1909]).7 This trend will be described in all its generality in the 1918 paper on converging lines of evolution among these languages. Finally, in texts devoted to specific languages or language groups, the expression mots accessoires is used to designate, e.g. in Old Persian, clitics and conjunctions (Meillet 1931) or, in Germanic languages, prepositions, preverbs and articles (Meillet 1917). In the latter study, the “autonomous value” (“valeur autonome”) of a form seems to be inversely correlated with semantic bleaching and coalescence, or in other words, a form is accessoire in proportion to its assumed degree of abstraction (e.g. in the passage from a demonstrative to a determiner) and its dependence on a mot autonome (as in Gothic ga “with” progressively used as an aspectual marker).
11Given Meillet’s very extensive notion of grammaticalisation, it could be equated with what we may characterize (admittedly in vague terms) as a loss of semantic and pragmatic weight (in the case of “defocused” words). Semantic weight appears to be proportional to concreteness, as would be suggested by the just cited example with venir and the above definition of grammaticalization as involving loss of concrete meaning. Indeed, Meillet declares that the constitution of grammatical forms is, among other factors, made possible by a loss of “concreteness” (Meillet 2015 [1912]: 217; cf. 2015 [1918]: 144). However, it is difficult to establish a gradient of concreteness in the very first example of bleaching cited by Meillet: “Je suis celui qui suis > Je suis chez moi > Je suis malade/maudit > Je suis parti.” Most linguists, given the prevalence of localist ideas, would surely characterize existence as having a lesser degree of concreteness than location (cf. e.g. Lyons 1967; Fortis 2020). Meillet’s intention may have been to point to an interplay between concreteness and the capacity of suis to effect a predication in full autonomy, that is, without a complement and without functioning as a copula or an auxiliary. Semantic weight would thus interact with autonomy in defining the degree of grammaticalization.
- 8 At this point, we may have expected Meillet to mention Bréal’s account by contagion, that is, the f (...)
12In the 1912 paper, the driving force behind the evolution of grammatical forms is the wish of speakers to find “expressive” ways of conveying their point. The quest for expressivity would underlie, for instance, the reinforcement of the negation ne by pas, mie or point in French. This split negation, once overused, would in turn become bland and cease to satisfy the need for expressivity, hence the use of doubly reinforced negations like pas du tout (Meillet 2015 [1912]: 218).8 The whole process is cyclic, or, as Gabelentz had said before Meillet, spiralling into the recreation of autonomous forms (Gabelentz 2016 [1891]: 269; cf. Heine & Reh 1982: 59, McElvenny 2025).
13In Meillet’s 1909 paper on the evolution of “simple” (i.e. synthetic) preterites toward compound forms, there is no hint that this change would be driven by, for instance, the greater expressivity of habeo dictum with respect to dixi (Meillet 2015 [1912]: 220). This means that the notion of expressive force is enlisted in the set of causes of grammatical change between 1909 and 1912. We would argue that Bally played a decisive role in convincing Meillet of the relevance of expressivity. While Meillet does acknowledge his proximity to Bally (and Spitzer) in a 1921 postscript to his 1912 paper, he makes no mention of the Swiss scholar in the body text, which is unexpected, for it is the purpose of Bally’s stylistique to explore the ways languages find means to be expressive, first and foremost, of affects and subjectivity.
- 9 E.g. in Bally (1909: 68). A gradient is posited between avoir “have” as “autonomous” word, e.g. exp (...)
- 10 As observed by Hopper and Traugott (2003: 26), the Greek future is summoned again by Benveniste who (...)
14Although in his Traité de stylistique (1909) Bally extensively discusses phraseology and the variations in autonomy exhibited by forms as a function of their context, he does not, however, touch upon the relevance of this autonomy for linguistic change.9 This he does in a later text, Le langage et la vie (1913), in which the role of affectivity and the loss of autonomy are broached together for the purpose of explaining the renewal of forms. For example, the periphrastic future, he argues (Bally 1913: 88), is meant to give an expression to various subjective conceptions of futurity, involving in particular the subject’s will, as in the Greek θέλω ἵνα, which underwent attrition and bleaching in the shape of the prefix θα, and became a pure marker of tense (the very same example is marshalled by Meillet 2015 [1912]: 223).10 Agglutination and attrition reflect a propensity to establish a univocal correlation between form and meaning (e.g. a stable marking of the future) and is associated with a loss of expressive force. This propensity is called “tendance analytique” by Bally, and is posited in contradistinction to a “tendance expressive,” which promotes variation (Bally 1913: 89). The opposition between the two tendencies runs parallel to that between the “intellectual” and the “affective.” As we shall see, this double dichotomy will be adopted by Spitzer.
- 11 Joseph observes that the success of the notion of grammaticalization is consonant with the fact tha (...)
15There are points on which Meillet, Bally, Gabelentz and Spitzer do diverge: Bally’s “tendance analytique” is not endorsed by Meillet; the driving force behind grammaticalization is, for Gabelentz, the interplay between ease of formulation (Bequemlichkeit) and semantic explicitness (Deutlichkeit), between an aesthetic drive to self-expression and trite use that degenerates into an invariant rule (McElvenny 2016, 2025). In Spitzer (as we will see in section 5.2), the frontier between stylistic features, textual patterns and grammaticalization is at issue. But the wish to make room for speakers’ expression, and thereby, for the aesthetic dimension in the wide sense given to it by Croce (i.e. the dimension of individual expression), is a feature common to Meillet, Bally, Gabelentz and Spitzer.11
- 12 See Lindström 2004, Haßler 2025, Klippi & Lindström 2025, Pagani-Naudet 2025.
16Our purpose here is not to engage in a history of the notion of grammaticalization or other kindred concepts.12 It is rather to contextualize the return to Bopp advocated by Meillet in his 1912 paper. By doing so, we hope to gain an appreciation of what unites Bopp’s and Meillet’s views, and what separates them, given the different intellectual contexts in which they were working, about a hundred years apart. Bopp’s own intellectual context might explain why Meillet distances himself from him, and at the same time, aspects of Bopp’s project might justify why Meillet makes a plea for revitalizing part of it.
17As we just said, grammaticalisation is presented by Meillet as a return to Bopp’s fundamental idea that inflections are essentially (though not exclusively) formed by agglutinating formerly autonomous roots and stems to the word base. Grammaticalisation, as the creation of forms, is set in opposition to analogy, which requires that the forms serving, as it were, as templates, be already in existence. Analogy, as emphasized by Meillet, had been an all-important concept in the linguistics of the preceding forty years or so (that is, since the Morphologische Untersuchungen of Osthoff and Brugman 1878), but it is neither suited to account for innovation nor, we may add, to finding a motivation for the renewal of forms that would go beyond purely formal relations.
- 13 Arnauld and Lancelot say that a form like vit “lives” is an abbreviation of est vivant “is living” (...)
18In Bopp’s account, agglutination involves two primitive (and broad) classes of monosyllabic roots which Bopp calls Verbalwurzeln (“verbal roots”) and Pronominalwurzeln (“pronominal roots”). The first class furnishes the stock of nominal, adjectival and verbal roots, while the second class is the source of pronouns, primitive prepositions, conjunctions and “particles” (Bopp 1835, §105: 194). The Latin verb legebam provides an example of a double agglutination: it is analyzed by Bopp into leg-ē-ba-m, with leg as the lexical base, ē, a “Bindevocal,” ba ultimately deriving from a “substantive verb” bū and m being originally a pronominal root present in Sanskrit mām “me” (Bopp 1842, §527: 768). It has been noted that Bopp’s ubiquitous use of the “substantive verbs” in analyzing verbal inflections owes much to the practice of glossing a verb as a copula followed by an attributive part rendered as a participle (so, lives = is living).13 This practice, which can be traced back to Aristole, was well established in grammaire générale but other influences may have been at play (see Swiggers 2016, for a discussion and references).
- 14 See, in this respect, Bopp’s letter to Humboldt translated and cited by Swiggers, dated 1820 (Swigg (...)
19Some aspects of the historical context that prevail when Bopp publishes his Conjugationssytem (Bopp 1816) either were antagonistic to agglutination theory, or, on the contrary, serve as a background to some of its main tenets, sometimes furnishing arguments that may seem outlandish to us. Among the obstacles is F. Schlegel’s claim that Sanskrit is an “organic” language in which inflections are expressed by a modification of the root, not by the “mechanical” aggregation of words and particles whose semantic import is independent of the root and transparent. Such a claim is not incompatible with the possibility of agglutination in Indo-European languages, and in fact Schlegel argues that modern languages have partly lost the “organic” character of the classical languages. In stating that agglutination is, more than Ablaut and reduplication, the dominant way of forming inflections in Indo-European, Bopp opens up the possibility of increasing grammatical complexity that contrasts with Schlegel’s “decadentism.”14
- 15 Allusion is made to Becker’s position by Delbrück (1880: 62), who lumps together, under the name of (...)
20Another obstacle to the idea that language may complexify through agglutination was the claim that it had to be complex from the beginning, that is, necessarily expressed not only concepts but also “relations” (through inflections).15 This claim was voiced in particular by Becker, as a consequence of his view that language, being an organic embodiment of thought, must by its own nature possess what caters to the needs of human thinking (Becker 1841: 19–23). Chinese presents itself as a counter-example, but Becker relegates it to a linguistic monstrosity. By contrast, for Condillac there is a stage at which signs serve to identify speakers and time reference, and a relation to the speaker’s appreciation of reality (moods) is not linguistic but gestural, hence inherently situational. The lexicalization of these “situational” signs is a historical process which finally issues into their conversion to endings (Condillac 2014 [1746], chap. IX, especially §85–90 for the evolution of the verb). This process is partly cyclic. French periphrastic verbal forms (like j’ai fait or je suis aimé) partly fall back on the method that primitive languages must have used, and which separated the expression of tense, mood and person from the bearer of lexical meaning (that is, a particular action or “passion;” “lexical” is not a term of Condillac’s). Some aspects of French, however, are not primitive since it has inflectional endings and since its word order is a historical legacy which is not motivated by the conditions of communication prevailing in the primitive context.
- 16 Pagani-Naudet (2025) demonstrates that a diachronic perspective on Romance futures does not arise b (...)
21The way Condillac historicizes grammar does not seem to be a practice that was well-established before his time. Indeed, there are indications that the very idea of distinguishing synchronic relations between forms and establishing their historical genealogy is a relatively late development.16 Bourquin (1980) documents this fact for French morphology and its terminology and shows in particular that in the notion of derivation have long been enmeshed synchronic and diachronic points of view, before etymology and synchronic derivation came to be distinguished clearly.
22An aspect of the German context that is most relevant for us here is the spate of more or less speculative studies on roots and their role in elucidating the birth and evolution of languages. The concept of root was borrowed from Hebrew grammar in the early 16th century, the term itself being a loan translation of one of the words for root in Hebrew (šoreš; Kessler-Mesguish 1996).
23That roots may serve as bases for derivation is a modern innovation. Ancient grammars, observes Jellinek, derived a full form from another full form. For example, Priscian derived Latin genitives in -orum from ablatives in -o, with the addition of -rum. Of special interest for us are early analyses of words into components classifiable as roots, derivational affixes (another notion of Hebrew origin), and inflections. In the history of these analyses, the grammar of Schottelius appears to be a landmark, for the clarity with which the three classes of components are distinguished and put to use (Jellinek 1913). After Schottelius, various authors attempt to offer a diachronic account of the formation of words by isolating a sort of “germinative” root reduced to a single syllable. Both Jellinek (1913: 154s) and Bréal, in his introduction to his translation of Bopp’s Grammaire comparée (1866: xxii) single out Adelung as a scholar whose perspective on the complexification of words and on the origin of inflections is clearly historical and anticipates the theory of agglutination, though in Adelung’s account inflections serve, right from the beginning, to express Nebenbegriffe (e.g. plurality) and relations between concepts (Adelung 1782).
- 17 In Max Müller’s account, Indo-European languages, like all languages, are the product of a grammati (...)
- 18 According to A. W. Schlegel and Delbrück, Bopp had a very limited knowledge of Indian grammarians, (...)
24In Bopp’s case, the decomposition of words into roots, affixes and inflections is not only an offshoot of this tradition. It is also firmly grounded in the Indian grammatical tradition, in which the identification of roots (especially verbal) goes back to pre-Pāṇinian times, and forms the backbone of a genre ancillary to grammar proper, the dhāthupāṭhas, or lists of verbal bases.17 It may be noted that in both traditions, Hebrew and Indian, roots do not coincide with words minus inflectional and derivational morphemes, that is, they are relatively “abstract,” and therefore better suited for cross-linguistic comparison. On the other hand, the fact that this concept of root diverged from anything known to the Western tradition did not facilitate its adoption (Alfieri 2014). Bopp does acknowledge his debt to the “Indian grammarians” in some passages, for example in his introduction to the chapter devoted to cases in the Vergleichende Grammatik (1833). However, in this occurrence and elsewhere as well, Bopp tends to underrate the merits of his Indian predecessors, noting for example that they were led to posit basic forms (Grundformen) or stems not thanks to their own scientific pursuit but simply because Sanskrit compounds laid these forms bare before their eyes. Yet, his practice shows his indebtedness to the tradition he does not seem to hold in very high esteem.18
- 19 Elsewhere, Delbrück attributes this trichotomy to the direct influence of Gottfried Hermann, and in (...)
25As emphasized by Delbrück, the main tenets of Bopp’s agglutination theory are not only empirical consequences of his description of roots. They are grounded in ideas inherited from the “scholarly tradition.” These tenets are three: the monosyllabicity of the roots; the trichotomy of parts of speech into noun, substantive verb and attributive, a principle of which we saw an application above with leg-ē-ba-m;19 finally, the claim, inspired by Hebrew grammar, that the personal endings of verbs are personal pronouns joined to the verbal stem (Delbrück 1919: 71).
- 20 De Brosses described this sound as a “un coup d’organe vocal, qui tout de suite a servi de germe d’ (...)
- 21 Significantly, Delbrück voices his qualms about “glottogonic issues” in a later edition of his Einl (...)
26That primitive roots must be monosyllabic is a claim that is sometimes defended with an argument that may seem highly speculative to us, yet is not left out of consideration by a scholar as level-headed as Delbrück. This is the claim that in primitive man an impression triggered a sound “in a lightning-like” way, hence that this sound was necessarily a single syllable (blitzartig, says Curtius, cited by Delbrück 1880: 80).20 Implicit too is the idea that primitive roots characterize a language state, and that they can therefore be assigned a uniform shape. In posterior research, variations in the form of roots were relativized to the language sample, as observed by Delbrück, and dialectal differences had to be recognized at a very early stage (Delbrück 1919: 158–159). This diachronization of the stage corresponding to what was envisaged as an Indo-European language preceding the division into its offsprings (the “glottogonic” stage) does not necessarily challenge the idea that primitive roots of an Ursprache really existed, as pointed out by Delbrück and Curtius (Curtius 1869 [1867]). But it makes the picture considerably more complex. Thus, Curtius divides the process of grammaticalization (which he does not designate that way, of course) as it was envisaged by Bopp into seven distinct periods. As a result of this progressive complexification, compound verbal forms combining with a substantive verb belong to a relatively late stage (the fifth). In Curtius’ Chronologie, one of the main motivations for splitting the primitive stage into different periods is that one and the same element cannot be resorted to time and again for the purpose of analyzing various forms with very different functions. For example, the Boppian claim that suffixing sa was instrumental in forming both a nominative and a genitive during the selfsame period is unlikely, given the distance that sets apart the functions of these cases (Curtius 1869 [1867]: 109). In addition, adherence to the Neogrammarian doctrine turned out to have adverse consequences for some central Boppian tenets. The inflexibility of sound laws entailed that some reconstructions had to be abandoned, while some discrepancies between cognate forms of Indo-European languages invited scholars to push back the existence of analogical processes to a very early date (Delbrück 1919: 127–128; p. 128 about analogy already in the Sanskrit middle voice). Regarding roots and their role in the formation of languages or “glottogony,” greater agnosticism prevailed in Brugmann, who put in doubt that roots may have existed at all in a separate state before coalescing. On the whole, then, Neogrammarians could be said to engage in an “anti-glottogonic reaction” to agglutination theory (Morpurgo-Davies 1998: 266).21 Finally, had Meillet mentioned agglutination, it might have been expected that he would take a stance on alternative theories, or at least allude to them (on these theories, Delbrück 1880, 1919). Perhaps he was not willing to enter this discussion.
- 22 In Bopp (1820), there is no implication that the development of the Romance future illustrates the (...)
- 23 The explanation could also lie partly in the vulgarizing nature of the paper (see Fagard 2025).
27We may surmise that Meillet’s caution in reintroducing the question of the origin of grammatical forms is better understandable in this context: he is not willing to return to the idea of a glottogonic stage during which complexification concerns primitive roots. Correlatively, he is no doubt aware of the just mentioned diachronization and complexification of this stage, as well as of alternatives to agglutination theory. Note too that his view entails that the creation and renewal of forms, being cyclic, does not have to be situated with respect to a glottogonic stage.22 Finally, as regards agglutination, it is but a facet of grammaticalisation, since the latter includes for example the fixity of word order or words being used out of focus. All these aspects of the historical context may go some way toward explaining the prima facie surprising fact that Meillet refrains from mentioning agglutination in his 1912 paper.23
- 24 The translation of Vossler’s paper was published in the Italian version of the internationally-orie (...)
28After Meillet, a survey of our corpus shows that a significant number of occurrences of grammaticalization, of words of the same family and equivalents in other languages are found in texts by Romanists sharing a concern for stylistic, aesthetic and cultural matters, especially Karl Vossler (1872–1949) and Leo Spitzer (1887–1960). In a 1913 paper referencing Meillet’s, and translated into Italian one year later (Vossler 1914), Vossler provides a long and well-illustrated discussion of the phenomenon of grammaticalization.24 Vossler also uses the term in at least two other references (Vossler 1919a, 1919b), for a total of 67 occurrences. Leo Spitzer uses the term in 26 different references, written in French, German and English, between 1914 and 1950 (cf. our section 8 below). Spitzer also references Meillet, in a 1919 paper. While Vossler situates the notion of grammaticalization in a more general framework which he views as inadequate, Spitzer does adopt it as a descriptive tool, but gives it a different and larger embrace than Meillet. Let us first turn to Vossler.
29For us to understand his critique of grammaticalization, a few words about his general perspective on linguistics are in order. This perspective Vossler characterizes as “idealism,” in opposition to “positivism” (Vossler 1904). Idealism is a search for the deepest causes of the creation and evolution of linguistic forms, and these causes are of a “spiritual” (geistig) nature: grammar must be the history of forms insofar as they are created or reshaped for aesthetic or stylistic purposes. The motto is “Stylistics first, then syntax!” (Vossler 1904: 16).
30For example, according to Vossler, a general drive to individualization and concretization is characteristic of Italian and manifested in its stylistic use of the definite article (mi dia il burro! “give me the butter”), and this drive resonates with the concern for plastic and intuitive representation (Vossler 1904: 20–22). Note that the definition of a national spirit is predicated on the social environment. For example, Vossler explains the extension of the partitive article in French as emanating from a new mercantile, quantity-oriented mentality (Vossler 1913: 191; Haßler 2014). The spiritual inclinations of speakers comprise aesthetic proclivities which govern phonetic structure as well: for example, the rhythmic pattern involved in the use of geminate consonants in Italian (fatto “done” from fare, conobbi “I knew,” from conoscere) reflects an aesthetic preference which operates at the expense of the root’s integrity. Preserving the integrity of stems, qua bearers of meaning, is by contrast a prime concern of German (Vossler 1904: 75–76).
- 25 This conception of change through small deviations is characteristic of Hermann Paul (1880–1920).
31In opposition to idealism, positivism focuses on the forms themselves, and preferably conventionalized and regular forms, that is, forms stabilized after individual speakers’ creations have converged on common structures. The crucial step of expressive agency thus falls out of the purview of positivist grammar. From this it follows that positivist grammar tends to obliterate the spiritual activity of the conscious and attentive speaker and submits a “mechanical” account of linguistic evolution: changes proceed in steps so small that they are incapable of provoking counter-measures on the part of speakers, and these changes impose themselves by dint of frequency (Vossler 1923: 80–81).25 Grammaticalization is a case in point: in the positivist spirit, it is the culmination of semantic change, the latter consisting in a series of hardly detectable semantic divergences between speaker and addressee (Vossler 1923: 81–82). There is no evidence, objects Vossler, for the relevance of frequency in this process, and a host of questions are left unanswered anyway: how has a change spread? In what areas and among which social groups? At what rate? Research on analogy, contamination, grammaticalization, etc. should therefore capitalize on the resources of statistics, linguistic geography and sociology (Vossler 1904: 93). In all these regards, idealist philology shares some concerns of Bréal and Meillet: reinstating human agency (in the case of Bréal, see De Palo 2001, chap. 3) and factoring in social considerations in accounting for linguistic change (Meillet 2015 [1906]).
- 26 Obviously modelled after the Peripatetic formula first coined by Thomas Aquinas: “nihil est in inte (...)
32In Spitzer’s general perspective on linguistics, and in consonance with Vossler, stylistics has pride of place and should be accorded the status of basic discipline in the organization of language studies, with syntax and phonetics being subsidiary (e.g. Spitzer 1943). Hence his adage, quoted in the title of this section.26
33Spitzer’s production is huge and his papers scattered in many diverse publications. We shall confine ourselves to a mere sample illustrating the various ways in which he uses the term grammaticalization. In the studies cited below, grammaticalization encompasses phenomena modern linguistics would classify as bona fide instances of what it recognizes under this term, but also cases we would be tempted to categorize as instances of reanalysis and of constructionalization (that is, the pairing of a form with a specific semantic and pragmatic import), and the term even refers to formal structures having a symbolic function. Spitzer’s descriptive practice, then, invites us to take a broader view of what is understood today by grammaticalization. His importance also lies in the fact that grammaticalization is for him a crucial concern, one that is especially visible in the Aufsätze zur romanischen Syntax und Stilistik (1918), as rightly noted by Vossler in his review (1919b). In fact, the number of places in which the notion seems to lurk in the background far exceeds the occurrences of the term grammaticalization.
- 27 Spitzer does not comment on the affective aspect of the fictive motion forms. He might mean that th (...)
- 28 These views will not be to Vossler’s taste, who is reluctant to recognize the validity of the affec (...)
34Let us begin with the most familiar, Spitzer’s treatment of one arch-example of grammaticalization, the Romance future (Spitzer 1918 [1916]). His analysis evinces the influence of Bally (1913), who is duly cited: the key factor in the renewal of the inflectional future through periphrastic forms is the need of expressing the affective dimension of speakers’ relation to future time, whether this involves obligation, free will or (in modern parlance) fictive motion (Fr. Je vais aimer). To these forms with an affective load is opposed the “logical” simple future, confined to the expression of tense.27 Note that grammaticalization is not merely the loss of affectivity, since Spitzer speaks as if “logic” were a driving force in itself (1918 [1916]: 178–179).28
35The opposition owes much to Bally’s distinction between the intellectual and the affective. In another text Spitzer appears to establish a parallelism between, on one hand affectivity and style, and, on the other hand, absence of affect and “grammaticalness:”
Die Bally, Meillet, Voßler haben gezeigt, daß es etwas wie ‘Grammatikalisation’ gibt, d.h. daß die ursprünglich affektischen Formen ihres Gefühlswertes sich entkleiden, sich scheidemünzengleich abnutzen, schal und gefühlsleer werden (genau wie die Metaphern zu gewöhnlichen Vokabeln abblassen), kurz daß das ursprünglich Stilistische zum Syntaktischen wird. Man kann mit Variation eines berühmten Satzes sagen: Nihil est in grammatica quod non fuerit in affecto und daraus folgern: Multa sunt in grammatica quae fuerunt in stilo (Spitzer 1919: 334).
Bally, Meillet, Voßler have shown that there is something like “grammaticalization,” i.e. that the originally affective forms lose their affective value, they are worn out like coins, become stale and devoid of affect (just like metaphors fade into ordinary words), in a word that what is originally stylistic becomes syntactical. One can say, with a variation on a famous sentence: nothing is in grammar which was not in affect and deduce from this that: many things are in grammar which were in style (our translation).
36The term of grammaticalization is also used to refer to reanalysis, that is, to the fact that a form is no longer analyzed in conformity with its origin. This can be illustrated by Spitzer’s study on the Romanian proclitic article al used in possessive constructions, e.g. un prieten al mieu “a friend of mine” (Spitzer 1950). Following a lead suggested by Meyer-Lübke, Spitzer makes the hypothesis that al originates from Latin ad illud, with ad taking on a partitive value. Three facts indicate, according to Spitzer, that the form has undergone grammaticalization. First, the Latin neuter form has become variable, under the attraction of the demonstrative (a)cel, so that the present form now carries markers of gender and number. Secondly, the partitive force has been lost, which occasions a reinforcement through de, as in un prieten de-ai miei “a friend of mine.” Finally, Spitzer regards as an instance of grammaticalization the fact that al mieu may be moved to a position preceding the head noun, as in un al mieu prieten. In his own terms, the transition from a “double-beat” construction (like un prieten al mieu) to a “single-beat” one signals an evolution which, in this occurrence, has a functional correlate (namely, focusing on the possessive). We may doubt that contemporary linguists would subsume a case of increased variation in syntactic behavior under the heading of grammaticalization.
37If the line between style and syntax may be difficult to draw, and if style comes before syntax, the possibility arises that what is viewed prematurely as a syntactic fact, or as grammaticalization at a more or less advanced stage, may be better analyzed as a stylistic device. Such is the gist of Spitzer’s brilliant paper on the style of the Peregrinatio Aetheriae: the frequent use of Latin his, iste and ipse as determiners is not a harbinger of the Romance article, contrary to what has been asserted before. These items function as stylistic devices meant to convey a sense of “pious astonishment” before the very things (“very” = ipse) that Aetheria is witness to, whose present reality can be attested (deictic hic, iste), and which the pilgrim thinks are imbued with a divine meaning. The intention is to inspire in the reader the awe which impressed itself on Aetheria when she was confronted with “these visible things which contain such portentous indications of transcendent happenings” (Spitzer 1949: 231).
38In another sense still, Spitzerian grammaticalization appears to involve what in modern parlance is akin to constructionalization, that is, the pairing of a form with a specific semantic and pragmatic import, associated with a semantic drift when this form comes to be used increasingly. Again, the line between the stylistic effect and its grammaticalized version may be difficult to draw. In a paper bearing in particular on the use of the French indefinite third person on, and French nous (“we”), Spitzer attributes to both forms various stylistic values which are not on an equal footing with respect to grammaticalization (Spitzer 1918). In the case of on, the fundamental value is the subsuming of an act under a norm or a general principle, but its extension to instances in which this value is no longer salient (e.g. on s’aimait, lit. “one loved each other”) shows that on + verb is a structure that is becoming grammaticalized. Further, on, by placing an act under a general norm or principle, acquires an “impersonalizing” aspect which motivates its uses in some particular circumstances, for example in non-intrusive questions (on sort ? vs sortons-nous ?). In short, grammaticalization is the name for the semantic extension of a construction beyond its original stylistic value.
39Finally, a word should be said about grammaticalization as the recurrent use of a form having an originally symbolic value. This sense stretches the notion to a point that it seems difficult to lump it with all the ways it has been envisaged heretofore, except perhaps in the case we have just seen, and described as “constructionalization.”
40In his paper on circular style (“style circulaire”), Spitzer identifies a structure which is characterized by the fact that some of the initial words (or of formally related words) of a text are repeated at the end of this text, typically in a reverse order, thus forming a chiasmus (Spitzer 1940). This is illustrated, among other examples, by the following excerpt from Dante, in a passage alluding to the divine Trinity:
Quell’ uno e due e tre che sempre vive,
E regna sempre in tre e due ed uno,
Non circonscritto, e tutto circonscrive (Dante, Paradiso, Canto XIV, v. 28–29, 13th c.)
That One and Two and Three who ever lives
and ever reigns in Three and Two and One
Not circumscribed and all things circumscribing (our translation)
41Remarkably, God is represented as an all-encompassing circle. The circular structure of the text is therefore iconic with respect to this representation. This form of iconicity is not the exclusive property of this text, argues Spitzer, for it is in fact a hallmark of the genre, that is, of writings which allude to a form of natural or supernatural ontological circle. Now, in cases where this chiasmic structure is no longer iconic and has become a simple stylistic device, Spitzer says we are dealing with a grammaticalisation (Spitzer 1940: 499). This semantic (or symbolic) weakening may be likened to the one observed with on.
42To wrap up this discussion, we could distinguish two broad meanings of grammaticalization in Spitzer’s writings: one is close to the contemporary perspective, which focuses on the co-evolution of morphosyntactic form and semantic import, as is the case of the Romance future and the Romanian article. Admittedly, in the eyes of today’s linguists, stylistic considerations in dealing with this evolution are unusual. But more quixotic for them might be the second broad notion of grammaticalization encountered in Spitzer. This notion involves a semantic degeneration for structures initially endowed with a stylistic value that is meant to symbolize a certain content in the text. Illustrations of this latter meaning of grammaticalization have been provided by the case-studies dealing with on/nous and with the “circular style.”
43In order to understand the emergence of both word and concept, and the possible interplay between these two perspectives, we chose to combine the methodologies traditionally associated with epistemology (sections 2–5) and historical (corpus) linguistics (sections 6–8). As far as the historical linguistics perspective is concerned, the methodology adopted is that of a corpus study, with an approach that is mainly semasiological: the aim is to study the emergence and evolution of the concept of grammaticalization, by focusing on its occurrences in a very specific corpus, i.e. the literature in linguistics at large, from the 19th century onward. It is thus a corpus study—keeping in mind, of course, the underlying epistemological question, i.e. we are trying to bring to light what motivated the use of this terminology, and how it gradually came to the fore as an important theoretical component of general linguistics.
44In order to analyze the evolution of the term grammaticalization, we set out to map its occurrences in European languages. We thus ran queries on all nominal and verbal forms of European equivalents of the term in scientific publications spanning more than a century (from the 1840s to the end of the 1950s). We included case-marked, conjugated and derived forms. Our corpus thus contains occurrences of e.g. Dutch grammaticalisatie, Serbian Граматикализација, French grammaticalisation; English grammaticalized, German grammatikalisiert, Italian grammaticalizzato, etc.
- 29 All hits must indeed be individually checked. This implies finding the mention of the term, its con (...)
- 30 There are two outliers in our corpus: Vossler, with 67 occurrences in 4 different papers, and Spitz (...)
45We combined a traditional approach—searching for mentions of earlier works in the reference sections of papers on grammaticalization phenomena, such as Lindström (2004), McElvenny (2016)—with a corpus approach. For the corpus approach, we used the large online databases Scopus, Jstor, ProQuest, Google Scholar, and the Google Books corpus series (for Italian, Russian, English, German, Spanish, French). We also included national diachronic corpora (Frantext for French, Corde for Spanish, Corpus do Português for Portuguese, Deutschen Textarchiv for German, MiDia for Italian). This yielded a total of 557 verified29 occurrences of the term, in 248 references by 186 different authors, with a mean of 2.2 occurrences per reference and 3 occurrences per author.30 Our corpus also includes a number of publications with no mention of the word grammaticalization or its equivalents, but with descriptions of similar phenomena.
46We have seen that the process of grammaticalization had been recognized long before the term was coined, especially in analyses resorting to agglutination. However, its status as a universal phenomenon results from a specific set of subprocesses (semantic weakening, decategorization, etc.), which seems to be a modern understanding, and the outcome of several decades of studies in the field of what is now known as grammaticalization.
47A modern definition of the phenomenon—which will suffice for our purpose here—is conveniently outlined e.g. in Hopper & Traugott (2003 [1993]: 2) as “that subset of linguistic changes whereby a lexical item or construction in certain uses takes on grammatical characteristics, or through which a grammatical item becomes more grammatical.” As described in e.g. Lehmann (2015 [1982]), Hopper & Traugott (2003 [1993]) or Heine, Claudi & Hünnemeyer (1991), this “subset” includes a series of phenomena which exist outside grammaticalization but, when concurrent, contribute to it. Within this subset, the main traits of grammaticalization relate to different facets of the linguistic sign (semantics, morphology, syntax, phonetics, etc.). One such feature is semantic bleaching, i.e. loss of meaning—though it is now commonly seen as a change rather than a loss; it can also be described as a dereferentialization process (Fagard 2024). Another is morphological decategorization, i.e. loss of morphological features. Another feature is the loss of syntactic autonomy, and yet another phonetic reduction. We have already mentioned some prototypical examples such as the Romance future, in which the Latin verb habeo “to have” has been gradually reduced to a mere inflection, and the French negation, in which the Old French noun pas “step” was gradually reduced to a negation marker.
48And we have seen that, while related to Meillet’s emphasis on expressivity, Vossler’s and Spitzer’s stylistic approach significantly departs from Meillet’s conception and, we may now add, from the just cited modern definition. Among our questions in this paper is to what extent this observation holds in our corpus. To this end, we examine how the term has been used after Meillet.
49Our analysis bears on the occurrences of the term grammaticalization, as illustrated in (1–3).
| (1) |
Tandis que l’analogie peut renouveler le détail des formes, mais laisse le plus souvent intact le plan d’ensemble du système existant, la « grammaticalisation » de certains mots crée des formes neuves, introduit des catégories qui n’avaient pas d’expression linguistique, transforme l’ensemble du système (Meillet 1912: 387). |
| Whereas analogy may renew forms in detail, usually leaving the overall plan of the system untouched, the ‘grammaticalization’ of certain words creates new forms and introduces categories which had no linguistic expression. It changes the system as a whole (translation Hopper & Traugott 2003: 23). |
| (2) |
A. Meillet nennt diesen Vorgang der Bedeutungsentleerung Grammatikalisation (Vossler 1913: 207–208). |
| A. Meillet calls this process of semantic depletion grammaticalization” (our translation). |
| (3) |
Cette construction […] a perdu toute sa valeur stylistique : c’est en français seul qu’elle a été « grammaticalisée » (M. Meillet), « cristallisée » (M. Foulet), en perdant ainsi toute valeur affective et prenant le nom d’« interrogation complexe » (De Boer 1926: 309). |
| This construction has lost all its stylistic value: in French alone was it ‘grammaticalized’ (Mr Meillet), ‘crystallized’ (Mr Foulet), thus losing all its affective value and taking the name of ‘complex interrogation’ (our translation). |
50We code for the type of construction the context refers to: the general concept of grammaticalization as in (1–2), a specific type of construction as in (3), or a specific construction as in (4).
| (4) |
Damit trat es aus dem Bereich des Bedeutungswandels hinaus und wurde zu einem Formwort der Steigerung grammatikalisiert. Einen ähnlichen Weg ist das Wörtchen arg im Begriff zu durchlaufen. Man vergleiche: ein arger Fehler und es hat mich arg gefreut (Vossler 1913: 214). |
| It thus came out of the realm of semantic change and was grammaticalized as a function word of intensification. The word arg ‘bad’ is following a similar semantic path. Compare: ein arger Fehler ‘a bad mistake’ and es hat mich arg gefreut ‘it really pleased me’ (our translation). |
51Our coding scheme includes the analysis of linguistic attitudes with respect to the use of the concept of grammaticalization: we code all types of hedging, such as Italian per così dire “so to speak” in (5).
| (5) |
Nel campo delle interiezioni una vecchia interiezione eminentemente onomatopeica come sí compare attenuata, e per così dire grammaticalizzata, secondo il procedimento di cui aveva dato prova Terenzio […] (Devoto 1940: 170). |
| In the field of interjections, an old, eminently onomatopoeic interjection such as sí ‘yes’ appears to be attenuated, and so to speak grammaticalized, following the process demonstrated by Terence […] (our translation). |
52In all such uses, we code not only for the presence of hedgers, but also for the presence of scare quotes (present as late as 1980, e.g. Ramat 1980: 3) as in (1) (distinct from actual quotation marks as in [3]), and for the presence of equivalents (cristallisée “crystallized” in [3]), explanations and glosses. In some cases, several hedging devices can cooccur, as in (6).
| (6) |
l’élément qui figure à la jonction de deux phrases tend à devenir un simple outil grammatical : il se “grammaticalise” pour ainsi dire (Meillet 1915: 21). |
| the element standing at the junction between two sentences tends to become a simple grammatical tool: it ‘grammaticalizes,’ so to speak (our translation) |
53We also code for the presence of a defining sequence. For instance, the beginning of the sentence in (6) actually provides a definition of grammaticalization, before introducing the term, as shown in (7), which displays additional context for example (6).
| (7) |
Quel que soit le point de départ, le trait commun à tous ces développements consiste en ce que, par l’effet de la répétition qui en a atténué progressivement la valeur expressive et en a fait oublier la signification propre, l’élément qui figure à la jonction de deux phrases tend à devenir un simple outil grammatical: il se “grammaticalise” pour ainsi dire (Meillet 1915: 21). |
| Whatever the starting point, the common feature in all these developments is the fact that, as a result of the repetition which progressively diminished its expressive value and caused its original meaning to be forgotten, the element which appears at the junction between two sentences tends to become a simple grammatical tool: it becomes ‘grammaticalized,” so to speak.” (our translation) |
54An essential but very time-consuming part of the coding scheme was the identification of the phenomenon described by the author, for each and every context. To this aim, we try to identify various elements: (a) the specific phenomenon (e.g. a shift from verb to auxiliary, from noun to adposition), (b) the language(s) described (e.g. French, Romance, Indo-European), (c) the essential traits associated with the change (e.g. morphological attrition, syntactic fusion, phonetic reduction, semantic bleaching), (d) the general characterization of the change described (e.g. as a case of grammaticalization and nothing else; as a case of grammaticalization, “formalization,” “intellectualization,” etc.). For instance, example (8) (which provides further context for example (2)) is coded as follows: (a) affix formation, (b) German, (c) semantic loss, (d) grammaticalization.
| (8) |
Die deutschen -heit, -schaft, -tum, -bar, -lich, -sam, -haft waren, bevor sie zu Suffixen wurden, sinnstarke Eigenwörter. A. Meillet nennt diesen Vorgang der Bedeutungsentleerung Grammatikalisation (Vossler 1913: 207–208). |
| The German -heit ‘-ness’, -schaft ‘-ship,’ -tum ‘-dom,’ -bar ‘-able,’ -lich ‘-ly,’ -sam ‘-able,’ -haft ‘-ly’ were, before they became suffixes, independent words with a strong meaning. A. Meillet names this process of semantic depletion grammaticalization” (our translation) |
55Finally, we code the metadata, including the following elements: word form, author, venue, language, country and scientific domain, in order to understand how this concept gradually spread in Europe and in the rest of the world.
56We also include in our corpus cases in which we can surmise that the word is used, but find no direct evidence, e.g. indirect references, or some occurrences in the Google Books corpus. In some of these cases, the metadata is thus incomplete (e.g. missing page number, partial context, missing author).
- 31 We did look for and systematically code occurrences beyond 1960, especially for languages for which (...)
57All in all, the coded corpus we use for this paper consists of 557 occurrences of the word grammaticalization or some equivalent in another European language, in 134 distinct references, from 1842 to 1959, and 11 languages: Catalan, English, Esperanto, French, German, Hungarian, Italian, Latin, Romanian, Russian, and Spanish.31 We analyze the results in the following section.
58We do not claim to have identified all occurrences of the word grammaticalization, its equivalents and cognates in the scientific literature from 1840 to 1959. On the contrary, there are probably many more occurrences which further digitalization of this literature could make accessible. What our results do show, undeniably, is that the word grammaticalization was in use in that period of time. Beyond this, our study yields the following results:
- Meillet did not coin the word grammaticalization: we found 15 occurrences before 1912. However, with 542 occurrences and 185 different authors for the 1912–1959 period (not counting multiple authorship), and at least eleven explicit references to Meillet,32 our investigation confirms that Meillet’s 1912 paper is indeed the starting point for the expansion of the concept across the range of linguistic domains (Romance linguistics, general linguistics, typological linguistics, corpus linguistics, etc., and even, albeit on the late side, formal linguistics).
- Between 1912 and 1959, the term is almost exclusively used to describe phenomena which “would today be described as cases of grammaticalization” (Vossler and Spitzer being notable exceptions). In some cases, especially in Spitzer’s work, the phenomena are only marginal cases of grammaticalization, corresponding more closely for instance to a case of lexicalization (note, however, that the two phenomena are closely related, see e.g. Wischer 2000, Lehmann 2002, Mihatsch 2025).
- It seems to be mostly associated with semantic change, but the literature also clearly identifies other hallmarks of grammaticalization, viz. syntactic, morphological and phonetic features, or the fact that grammaticalization is a gradual, stepwise phenomenon.
- Another important hallmark in modern descriptions of the phenomenon is entirely absent in our corpus: the notion of unidirectionality. Though it might underlie some of the descriptions, it remains implicit.
- Finally, the term is used to describe a wide variety of languages, language families, periods, phenomena. However, the range of languages—and thus the main subarea in which this terminology is uses—seems to shift gradually from Romance linguistics to general linguistics.
59In what follows, we discuss and illustrate each one of these results.
- 33 Referring back to a chapter of his 1979 book, in the new and considerably revised edition (Givón 20 (...)
60The results of our analysis show that the term “grammaticalization,” along with translations and variations thereof, had been used in (at least) three hundred papers, by over a hundred linguists and in over ten languages by the time Givón claims he “reinvented” it, in 1979.33 Surprisingly, our research also shows that Meillet did not invent the term: in fact, we found several occurrences in French dating back to the late 19th century, with the oldest attestations dating back to 1842 (Fagard 2023), as illustrated in (9) (note that De Radonvilliers also lists other nominal and verbal forms of the term).
| (9) |
Grammaticalisation, subst. fém. ; action de grammaticaliser, état grammaticalisé. […]
Grammaticaliser, v. act. et pron. (se –) ; rendre, devenir grammatical, donner de la grammaticalité; faire entrer, entrer naturellement dans un ordre, un classement grammatical, rattacher, se rattacher aux règles grammaticales, entrer, rester dans les principes de la grammaire, définir selon l’esprit et l’exigence de la grammaticalité (De Radonvilliers 1842: 186–187).
|
|
Grammaticalization, noun, fem.; action of grammaticalizing, state of being grammaticalized. […]
Grammaticalize, trans. and pron. verb (-oneself); make, become grammatical, give grammaticality to; help enter, enter naturally into a grammatical order or classification, attach, be attached to grammatical rules, pertain to, stay within the principles of grammar, define according to the spirit and the requisites of grammaticality (our translation).
|
61The term resurfaces once in 1879, under the form “grammaticaliser” (10); this is the last occurrence we found before Meillet’s 1912 paper.
| (10) |
Si l’on me parlait de l’unité de langue pour tout le globe, alors je me tranquilliserais et j’enverrais de suite ces utopistes tenir leurs écoles sur les ruines de la Tour de Babel, refaçonner la langue chinoise, grammaticaliser les barbares avant de venir me troubler dans l’usage de mon dialecte, si rapproché déjà de nos langues savantes […]a |
| If one were to talk to me of unifying all the world’s languages, then I would stay calm and send these utopians right away to build their schools on the ruins of the Babel Tower, bring Chinese back to its original state, grammaticalize barbarians before coming here to trouble me with the way I use my dialect, so close already to our learned languages (our translation). |
| a. Excursion au Mont Fallère, le 21 Août 1879. Lettre à M. le chevalier abbé P. Chanoux, Recteur de l’Hospice du Petit St.-Bernard, 1880: 406. |
62In these occurrences, the meaning is not yet the modern one. It takes on the meaning of “making or becoming grammatical” (9) or “teaching grammar to someone” (10)—both quite far from the modern meaning, and much closer to that of English grammaticize “make more grammatical,” already in use in the early 19th c. (11).
| (11) |
At first,’ replied Mr. Stanley, ‘I only meant to give Lucilla as much Latin as would teach her to grammaticize her English, but her quickness in acquiring led me on, and I think I did right; for it is superficial knowledge that excites vanity.a
Grammaticize had made its way into French by the mid 19th c., with occurrences both in French-English dictionaries as in (12) and in translations from English (13), in linguistics.
|
| (12) |
GRAMMATICIZE [gråmmåt’isiz] v. a. rendre grammatical ; écrire grammaticalement (Spiers 1846: 261). |
| Grammaticize [gråmmåt’isiz] trans. v. make grammatical; write grammatically (our traduction). |
| (13) |
J’ai toujours cru, – disait l’oracle, – que Shakespeare avait eu assez de latin pour « grammaticiser » son anglais (1780 ; Johnson avait alors 71 ans) (Sayce 1884 [1874, 1875]: 65). |
| ‘I always said,’ quoth the oracle, ‘Shakespeare had Latin enough to grammaticise his English’ (anno 1780, æt. 71) (translation Jovy 1884: 58). |
| a. Cœlebs in search of a wife, vol. II. New York: T. and J. Swords. From the 2nd London edition. 1809: 138. |
63This seems to confirm that Meillet’s use of grammaticalisation is independent from these earlier occurrences.
64Confirming preliminary findings in Fagard (2019), there is a wealth of papers in linguistics between 1912 and the end of the 1950s which mention grammaticalization, or even focus on the phenomenon: it was by no means “forgotten” between Meillet and, say, Givón (see also Lindström 2004, which lists a number of references to Meillet in that period, albeit in a different perspective).
65But beyond this, in our corpus, 3 references contain the term grammaticalization before 1912. Radonvilliers seems to have coined the term, which would be fitting, given the title of his book: Enrichissement de la langue francaise. Dictionnaire de mots nouveaux, which roughly translates to Enriching the French Language, Dictionary of New Words. What Radonvilliers aims at is the copia verborum: there is no intention to reflect current or past usage, only the wish to introduce neologisms which speakers and writers may seize upon, or not (cf. de Radonvilliers 1845: I; see also Klippi and Lindström Tiedemann, 2025). The term is also used in 1880, as an equivalent of the (much earlier) term grammaticize (cf. section 7.3).
66Most of the occurrences we found, however, are indeed posterior to Meillet’s 1912 paper. Besides Vossler and Spitzer, two other authors reference Meillet early on: József Balassa (1864–1945), in his review of Meillet’s paper; and Louis Couturat (1868–1914), with a brief account—in Esperanto—of Meillet’s paper.
67Table 1 below illustrates the growing number of authors and publications containing the term grammaticalization.
Table 1. Occurrences of some form of grammaticalize or grammaticalization, in any language, in our corpus (including 6 occurrences with prefixation).
| Decade |
Occurrences |
Authors (& number of new authors, if any) |
Papers |
Languages (& number of new languages, if any) |
| 1840s |
14 |
1 (1) |
2 |
1 (1): French |
| 1880s |
1 |
1 (1) |
1 |
1: French |
| 1910s |
89 |
9 (9) |
16 |
8 (7): English, Esperanto, French, German, Hungarian, Italian, Latin, Russian |
| 1920s |
46 |
12 (10) |
14 |
4: French, German, Hungarian, Russian |
| 1930s |
34 |
17 (15) |
22 |
4: English, French, German, Russian |
| 1940s |
55 |
27 (22) |
34 |
5 (1): English, French, German, Italian, Russian, Spanish |
| 1950s |
318 |
135 (128) |
159 |
8 (2): Catalan, English, French, German, Italian, Romanian, Russian, Spanish |
| Total |
557 |
186 |
248 |
11 |
68The term is thus almost absent before Meillet’s 1912 paper. The number of authors using it, and of references including it, gradually picks up thereafter—with an isolated dip in the 1920s—as illustrated in Graph 1.
Graph 1. Occurrences of grammaticalization and other equivalents in our corpus (number of authors and references).
- 34 The query was run on November 18th, 2024, for the terms grammaticalization, grammaticalisation and (...)
69After 1960, the number of studies on grammaticalization increases considerably. For instance, a query for papers on grammaticalization in Google Scholar yields a mean of over 10,000 hits per decade after this date, as illustrated in table 2:34
Table 2. Occurrences of grammaticalization and equivalents in Google Scholar (number of hits, period 1960–1999, per decade).
| Decade |
Number of hits |
| 1960–1969 |
154 |
| 1970–1979 |
281 |
| 1980–1989 |
1,290 |
| 1990–1999 |
6,350 |
- 35 Including debates on the very existence of grammaticalization, which are of no relevance here (cf. (...)
- 36 A phenomenon which has received various analyses: as cases of grammaticalization (e.g. Traugott 199 (...)
70A side effect has been the publication of a number of studies discussing not only typical grammaticalization phenomena, but also more marginal ones (often explicitly so, cf. Bisang et al.’s 2004 volume title: What makes grammaticalization? A look from its fringes and its components). This prompted some debate in the past decades concerning the range of phenomena that should count as cases of grammaticalization,35 for instance concerning the emergence of discourse markers.36
71These debates also yielded a clear picture of—and a broad consensus on—what is at the core of grammaticalization, and what its main traits are (see section 1. above).
72One of our goals in this study was to understand if what scientists in the 19th and early 20th centuries understood by grammaticalization is partly, completely or not at all in line with this modern understanding of the phenomenon. From this perspective, the results of our corpus study seem to indicate that most occurrences from 1912 on indeed correspond to phenomena which we would now describe as cases of grammaticalization, either in its narrowest sense or in a broader sense including, for instance, the emergence of discourse markers.
73Indeed, in our corpus, the analysis of grammaticalization phenomena seems mostly associated with semantic loss: in our 1912–1959 corpus, 80 authors (43%) mention semantic loss at least once in their discussion of cases of grammaticalization, as illustrated in (15–18) and shown in table 3.
| (14) |
A. Meillet chiama grammaticalizzazione questo procedimento di perdite del significato (Vossler 1914: 212, Italian translation of Vossler 1913). |
| A. Meillet calls grammaticalization this process of meaning loss (our translation). |
| (15) |
l’inversion de c’est que, avec perte de son sens causal à cause de sa grammaticalisation (De Boer 1926: 314). |
| the inversion of c’est que ‘it is that’, with loss of its causal meaning on account of its grammaticalization (our traduction). |
| (16) |
a szavak ilyen tartalombeli átalakulásának (Balassa 1913: 536). |
| the [process of] transformation of the words’ contents (our translation). |
| (17) |
The demonstrative there retains its primitive semasiological value; the expletive there has become grammaticalized, that is, it has lost its semasiological value and has value only as a device to obtain syntactical or stylistic variation in sentence structure (Bull 1943: 119). |
Table 3. Features explicitly mentioned by authors in their description of grammaticalization phenomena (i.e. number of authors mentioning these features; the explicit nature of this mention does not imply that the author used the same term, e.g. semantic loss can be expressed as a loss in [semasiological] value, as in [18]).
| Decade |
Semantic loss |
Syntactic autonomy |
Shift from lexicon to grammar |
Stepwise, gradual process |
| 1910s |
7 authors (54 mentions) |
5 (11) |
5 (8) |
3 (16) |
| 1920s |
5 (13) |
0 |
0 |
1 (3) |
| 1930s |
10 (13) |
4 (6) |
2 (2) |
5 (8) |
| 1940s |
16 (25) |
4 (7) |
1 (1) |
10 (12) |
| 1950s |
55 (116) |
10 (16) |
33 (43) |
43 (81) |
| Total |
80 (221) |
21 (40) |
41 (54) |
60 (121) |
74As a whole, the literature also identifies other hallmarks of grammaticalization, viz. syntactic, morphological or phonetic features. However, these are mentioned both less explicitly, as in (18–20), and by much fewer authors.
| (18) |
der Ausdruck ist ‘grammatikalisiert’ und damit der Syntax anheimgefallen (Lerch 1919: 287). |
| “the expression is ‘grammaticalized’ and thereby becomes a part of syntax (our translation) |
| (19) |
Merece mención el desgaste fonético de mire usted > misté, ya que éste evidencia la gramaticalización del verbo usado para llamar la atención del interlocutor (Krüger 1952: 382). |
| the phonetic erosion of mire usted > misté ‘look.imp you.nom > look now, look here’ is worthy of mention, since it demonstrates the grammaticalization of the verb used to call on the hearer’s attention (our translation). |
75Interestingly, the fact that grammaticalization implies a shift from lexicon to grammar, though it must be obvious to all linguists using the term, is seldom made explicit, as in (20).
| (20) |
lo svolgimento stesso dei fenomeni costringe a passare dal campo lessicale a quello morfologico. La «grammaticalizzazione» di termini, costruzioni, suffissi, è infatti uno degli aspetti di quella crescente importanza dell’espressione astratta (Schick 1956: 294). |
| the unfolding of these phenomena brings about a shift from the lexical domain to the morphological domain. The ‘grammaticalization’ of words, constructions, suffixes, is indeed one of the aspects of that increasing importance of abstract expression (our translation). |
76Conversely, the fact that it is a gradual, stepwise phenomenon, is frequently mentioned (21–23).
| (21) |
en muchos casos no se ha operado una gramaticalización completa y continúa vivo, en mayor o menor medida, un elemento enfatizador (Monge 1954: 15). |
| in many cases the grammaticalization process has not gone to completion and an emphatic element remains more or less alive (our translation). |
| (22) |
il tend à se grammaticaliser en auxiliaire (Demiéville 1950: 42). |
| it tends to grammaticalize into an auxiliary (our translation). |
| (23) |
una gramaticalización –en trance de cumplirse– que acabará el día en que los pronombres, perdido el asidero, tan débil, de su presencia enfática, sean mero utensilio vacío de significado (Alvar 1955: 311–312). |
| a grammaticalization process—in the making—which will be over the day the pronouns, having lost the oh so feeble grasp of their emphatic presence, become simple tools, devoid of meaning (our translation). |
77Another important hallmark in modern descriptions of the phenomenon is entirely absent in our corpus: the notion of unidirectionality. Though it might underlie some of the descriptions, it remains implicit, as far as we know.
- 37 In a few cases, as noted above, the term is used with a very different meaning (something more akin (...)
78While the term grammaticalization is mostly used, between 1912 and 1959, to describe phenomena which would today be described as cases of grammaticalization,37 in some cases the analysis or the case study itself seem to indicate that the phenomenon under study is only a marginal case of grammaticalization, as in (24).
| (24) |
This is a second Arabic etymological-semantic story. Almost all languages call the apple of the eye “pupil,” pupilla, i.e. “little girl,” “sweetheart,” because the primitive lover, seeing his own picture reflected in the eye of his beloved, fancied he would carry hers about in his. Now, whereas all the other Romance languages have grammaticalized the Latin pupilla and deprived it of its original sense, Spanish has retranslated and vivified the picture by way of the Arabic pattern insan ul ‘ain by saying niña del ojo (Hatzfeld 1946: 351). |
79This discrepancy is sometimes remarked on by contemporaries, as is the case for Spitzer in (25) (about the use of the term grammatikalisiert “grammaticalized” to describe what seems closer to a lexicalization phenomenon).
| (25) |
Auſserdem verstehe ich das Wort „grammatikalisiert“ in obigem Satze nicht (Spitzer 1928: 106). |
| Besides, I do not understand the word ‘grammaticalized’ in the above sentence (our translation). |
80Another discrepancy has already been mentioned: the pervasive focus on style, which is often mentioned in our corpus, especially in Spitzer’s work. Our corpus provides, as it were, a forerunner to Givón’s famous saying “Today’s morphology is yesterday’s syntax” (Givón 1971: 413), in the form of today’s style is tomorrow’s syntax. To quote another Romanist:
| (26) |
Der andere Leitgedanke aber ist der, daß alles, was jetzt Syntax ist, einst Stilistik gewesen sei, daß die Syntaktika erstarrte, grammatikalisierte Stilistika darstellen (Lerch 1919: 287). |
| The other leading thought is, however, the fact that everything that is now syntax was earlier stylistics, that syntactic elements are fixed, grammaticalized stylistic elements (our translation). |
81In our corpus, the term grammaticalization is used to describe a wide variety of languages, language families, periods, phenomena. However, for the 1912–1959 period, the range of languages—and thus the main subarea in which this terminology is used—seems heavily dominated by Romance linguistics; only very gradually does it shift from Romance linguistics to general and/or typological linguistics, as we show below.
Graph 2. The importance of Romance linguistics in our corpus. For each decade, the graph indicates the number of authors providing examples of Romance languages (with or without other languages), or exclusively other languages (i.e. excluding Romance languages and only generic references).
82This tendency holds even for the Russian subcorpus, in which the tendency is, however, less obvious. That the term was used in Russian should come as no surprise: despite a mixed reception owing to the peculiarities of linguistics in Russia under the Stalinist regime, Meillet remained an important figure in Russian linguistics (Moret 2024: 117).
- 38 A survey of the literature in the 1960s, with over 146 verified references, suggests that the impor (...)
Graph 3. Proportion of papers focusing on Romance languages, in our corpus.38
83The expansion of grammaticalization throughout linguistics seems to have stemmed mainly from one subdomain, Romance linguistics. And one author in particular seems to have had an outsized role in this worldwide expansion: Leo Spitzer, with 27 occurrences appearing in French, German and English, in 15 different references, ranging from 1914 to 1950—including 10 different journals (Comparative Literature, Die Neueren Sprachen, the Hispanic Review, Modern Language Notes, Modern Language Quarterly, Neuphilologische Mitteilungen, Revista de Filología Española, Word, Zeitschrift für französische Sprache und Literatur, Zeitschrift für romanische Philologie).
84The increase in frequency is such that it is even possible to retrace it in large, generic corpora, e.g. in French (graph 4) or English (graph 5), though this comes about much later.
Graph 4. Frequency of the term grammaticalisation in the Frantext database (x-axis: date of publication; y-axis: relative frequency, per million words).
Graph 5. Frequency of the term grammaticalisation in the Google Ngram database (x-axis: date of publication; y-axis: relative frequency, per million words).
85An S-curve analysis of these data (i.e. an analysis in terms of sigmoid patterns, evidencing language change, cf. Feltgen 2017, 2025) show that the important phases of the frequency increase do not start before the 1980s (graph 6).
Graph 6. Identification of an S-curve in the frequency graph (graph 4) (x-axis: date of publication; y-axis: relative frequency, per million words).
86Authors seem to be conscious of this evolution. At the very least, many uses of the term seem to indicate that they are aware of the fact that the concept may be present even though the term is not, as in (28).
| (27) |
Estamos pues, para WISTRAND, aunque él no utilice este concepto, ante un proceso de gramaticalización de un elemento significativo (Monge 1954: 14). |
| We are thus, for Wistrand, even though he himself does not use this concept, dealing with a process of grammaticalization of a meaningful element (our translation). |
- 39 It is hard to provide an estimate of the number of references for the intervening period (1960–1979 (...)
87As mentioned above, the results of our analysis show that the term grammaticalization, along with translations and variations thereof, had been used in (at least) 245 papers, by over 200 linguists (counting coauthors) and in over ten languages, two decades before it was “reinvented” in 1979.39 Surprisingly, our research also shows that Meillet did not invent the term: in fact, we found various occurrences dating back to the late 19th century. The notion itself did not come out of the blue. Meillet’s grammaticalisation certainly owes much to a century of research on agglutination, and to his illustrious initiator, Bopp. However, Meillet’s plea for a return to Bopp is cautious, for reasons to be sought in the new “anti-glottogonic” context. It has been noted too that Meillet’s view of grammaticalization, which brings into play the puzzling notion of mot accessoire, is rather idiosyncratic. On the other hand, his emphasis on expressivity in the renewal of forms links him to Bally’s views on the affective dimension of language and, more distantly, to scholars who give pride of place to stylistics, such as Vossler and Spitzer. Among the scholars of the period considered here, Spitzer is probably the one who made the widest use of the term of grammaticalization.
88The survey of our corpus confirmed two things. First, in accordance with what has been said in the literature on grammaticalization since the 1970s, Meillet’s 1912 paper is indeed the starting point for the expansion of the concept across a range of linguistic domains, starting with Romance linguistics. The importance of Meillet’s impact, despite the limited use he himself makes of the word (cf. also Klippi & Lindström 2025), can be attributed to Meillet’s great influence in that period (see e.g. Bat-Zeev Shyldkrot & Moret 2024: 10). Most papers in the 1910s point back to Meillet, and he is mentioned at least once in each of the following decades.
89Second, confirming the findings in Fagard (2019), there is a wealth of papers in linguistics between 1912 and the 1970s which mention grammaticalization, or focus on the phenomenon: neither the concept nor the word had been “forgotten” since Meillet. The gap which Hopper and Traugott see between Meillet and important work done by Indo-Europeanists in the 1960s (Hopper & Traugott 2003: 25) does not really exist, or, rather, it is not as wide as might be thought. Our data, however, do show that from the 1950s on grammaticalization attracted a growing interest which ushered in a “grammaticalization boom” in the 1980s.
Flux RSS
