Introduction

1Grammaticalisation is a given in linguistics of the 21^rst century and its basic assumption has been tested on a large scale in different languages of the world (e.g. Heine & Kuteva 2002; Smith et al. 2015; Bisang & Malchukov 2020a, 2020b). The term grammaticalisation is generally attributed to Meillet (1912) (e.g. Hopper & Traugott 1993: 18, 2003: 19; Narrog & Heine 2011: 1; Lehmann 2015 [1995, 1989]: 1), but it has not been demonstrated that his concept covers exactly the same meaning as the modern concept of grammaticalisation, nor that linguists who started talking about grammaticalisation around the 1970s were aware of his term and concept.

2In the practice of any research, the methodological starting point tends to be to review previously conducted studies with the aim of turning ‘towards the past in order to legitimate a contemporary cognitive practice’ (Auroux 1989: 13). In terms of their historical accuracy, such outlines frequently reiterate commonplaces, transmitted from one secondary source to another without the authors having investigated the primary sources in depth. This is because their objective is simply to deliver a new contribution to the discipline or to challenge established conceptions (Klippi 2010: 24). However, historical knowledge may be perceived as even more interesting when the research topic can be justified in a manner that attracts the attention of modern linguists and when modern linguistics can go as far as to determine the historian’s research agenda. The ‘meta-perspective’ of historical studies would be to arrive at knowledge that is also relevant at present, without underestimating or denying their evident sui generis value (Klippi 2010: 25).

4Nerlich and Clark (1996: 156) have stated in a seemingly paradoxical manner that numerous linguistic concepts have a ‘short past, but a long history’. This means that while certain linguistic concepts may have only recently been formally recognised, their origins can be traced back much further in history (cf. also Haßler 2025). This distinction is highly relevant here, since the integration of the concept of grammaticalisation into modern linguistics lags behind its actual historical development, as pointed out by several scholars (e.g. Lehmann 2015 [1995, 1982]; Hopper & Traugott 1993; Lindström 2004b; Itkonen 2005; McElvenny 2020; Haßler 2025). Similar ideas can be identified in grammatical and philosophical traditions of the 17^th and 18^th centuries, and similar changes were subsequently discussed within the framework of comparative grammar, primarily to explain the origin of affixes (cf. Lindström 2004b; Lindström Tiedemann 2005). Previous studies of the history of grammaticalisation have concentrated predominantly on the German tradition, which had a significant influence on Meillet. Yet, the contributions of French linguists, including Meillet’s own work, have been largely overlooked in the history of grammaticalisation. This negligence persists despite the references in 19^th century German texts to French scholars and their contributions to the study of what we now call grammaticalisation (cf. Lindström 2004b).

5In the context of the rediscovery of the concept of grammaticalisation around the 1970s (cf. e.g. Lehmann 2015 [1995, 1982]: 6–9; Hopper & Traugott 1993: 24, 2003: 25; Heine 2003: 576), linguists were confronted with the need to rethink not only their notion of language change, but also their understanding of linguistic categories. Pressure came from two directions. In the then dominant generativist framework, categories were conceived as discrete entities, whereas the notion of grammaticalisation postulates that categories are continuous by nature (Itkonen 1978: 109). In addition, in more traditionally oriented linguistics, there was also a well-established practical distinction between the grammar and the lexicon, although in linguistic reality the boundary between them is not entirely clear-cut (Anttila 1989: chapter 7.13). This was also one reason for the different understandings of the same, or similar, phenomena related to language change (cf. Joseph 2015). The question of what fits under the umbrella term of grammaticalisation has persisted.

7The structure of the present study is as follows: First, we provide an account of the history of concepts as a theoretical framework (section 1). We then review the main strands of previous research on the history of grammaticalisation and provide a brief description of some of the main aspects of 20^th century grammaticalisation (section 2). This is followed by a description of Meillet’s concept of grammaticalisation as a particular type of language change (section 3), and finally, we present our conclusions and some suggestions for future research (section 4).

1. The history of concepts as a theoretical framework

8The framework of the history of concepts can be divided into three schools of thought with slightly different emphasis: Begriffsgeschichte (Koselleck 1985 [1979], 2002), historical semantics (Skinner 1969, 1978) and analyse de discours du côté de l’histoire (Guilhaumou 2002, 2006). The main challenge of this branch of study is fundamentally semantic, as its task is to figure out how the relationship between different political, social or cultural realities and linguistic terms and concepts should be understood. According to the strict form of scientific constructivism, the phenomena studied by science may be reduced to their linguistic expressions (Boyd 1992: 132). It could thus be maintained that any scientific phenomenon exists as a linguistic product, in the form of terminology, texts and discourses, which in turn may become objects of analysis.

9One might argue that grammaticalisation as a scientific phenomenon came into existence only when the term was coined. Nevertheless, according to the uniformitarian principle (Christy 1983; Deutscher 1998) as understood since the 19^th century, linguistic mechanisms and forces are unlikely to have been completely different in the past to what they are like today (Lass 1997: chapter 1.5; Deutscher 1998: 109). Therefore, in linguistic reality the changes referred to as grammaticalisation have always existed, even if linguists were not there to conceptualise them as grammaticalisation and to lexicalise the concept as grammaticalisation.

10Consistent with semasiological and onomasiological analyses in linguistics, the research in the history of political, cultural, and intellectual concepts has shown that (1) the same term does not necessarily carry the same meaning, (2) different terms can be used for the same or similar concepts, and (3) a concept can exist even before it is given a name (Richter 1995: 9, 47; cf. Koselleck 1985 [1979], 2002; Cassin 2004). The experiences of ordinary concepts by lay people give similar results. Nevertheless, the surrounding co-text and context (i.e. the extralinguistic setting) can help identify how a concept is understood and how a term is used.

11As tested previously on the concepts of context and analogy (Klippi 2024, forthcoming; cf. also Haβler 2007, 2017, 2024, 2025), the history of concepts offers an analytical tool for the history of linguistics, because as sciences that rely (mostly) on written sources, both the history of concepts and the history of linguistics seek information about the meaning of terminology appearing in the source texts. They also seek knowledge of the terminology used to refer to the concepts under investigation. In other words, the idea underlying the history of concepts is that concepts cannot be interpreted correctly unless we know the possible means of expression during the period in question and understand what was meant by these linguistic expressions at the time. Thus, linguistic usage is a window to understanding history.

12Furthermore, we need to be careful not to assume that terminology has always meant the same. Warning of the trap of anachronism, one of the founding figures of the history of concepts, Otto Brunner (1898–1982) emphasised in the same vein that a historian should prefer the terminology of the time being studied, instead of taking recourse to modern terms (Brunner 1959; Vogelsang 2012: 9). An anachronism can be a phenomenon, thing or word that has been accidentally or intentionally placed in the wrong era as part of the reality described by the text. Furthermore, a text of the past cannot be interpreted as such from the present moment or from one’s own culture and experience; the phenomena it describes must be related to the contemporary context in which the text was produced. Consequently, strictly speaking, it would be anachronistic to use the term grammaticalisation in the present context to refer to how this phenomenon was treated in linguistic thought before the term (presumably) saw the light of day in Meillet’s article in 1912. In practice, we are nevertheless obliged to use Meillet’s term as a shortcut, not least because it has gained widespread recognition within the scientific community of modern linguists. Similarly, it is also anachronistic to assume that Meillet meant exactly the same by grammaticalisation as what was meant by the term by linguists during the 1970s revival or what is meant at present by the same term. In addition, we should remember that all linguists at present do not mean the same by this term (cf. Narrog & Heine 2011: 2).

13According to the punctual approach (Skinner 1969) to the history of concepts, the presence of a problematic term in discourse is imbued with historicity in the sense that the object of knowledge is linked to the context, i.e. to the particular conditions of production and to a specific moment in history (cf. Foucault 1969; Skinner 1969). The choice and use of a particular concept is not gratuitous or innocent, but always meaningful and intentional in relation to local contingencies (Guilhaumou 2002). The context does not exist as such to be directly accessed in space and time. The context rather emerges from the texts of the past through which the researcher constructs it by means of language, thus bringing into view the context that existed in the past. In this case, the immediate context of the concept of grammaticalisation is provided by the surrounding linguistic thought and theories, which determine how the concept should be understood.

14The concept of grammaticalisation within the punctual framework can be understood from two perspectives: a researcher can either (1) focus on the language used in relation to the concept, or (2) examine the intentions behind the use of the term by a particular writer (Bevir 2002). In this article we aim to cover both aspects.

2. The concept of grammaticalisation and previous studies of the history of grammaticalisation

15Grammaticalisation is not a ‘uniform concept’, and it has been defined in different ways (Narrog & Heine 2011: 2) and used in various ways (e.g. Giacalone Ramat 1998; Wischer 2000; Lindström 2004c). This complicates the analysis of the history of the concept, since it is not obvious exactly what is to be understood as the concept under review. Some French scholars have clarified the modern perspectives on this issue (Marchello-Nizia et al. 2003; Marchello-Nizia 2006), but there remains what Béguelin (2010: 243) has called ‘terminological inflation’ and a ‘byzantinism of discussions’, which sometimes seems more like scholarly hair-splitting.

16Hopper and Traugott (1993: 1–2) provide two definitions of grammaticalisation in their seminal textbook. (1) Grammaticalisation can be considered as a theoretical ‘framework’, based on the assumption that linguistic categories are not discrete, and that linguistic structures are subject to constant evolution. The aim of this framework is to study the use and emergence of grammatical forms and constructions. (2) On the other hand, grammaticalisation can be seen as ‘processes whereby items become more grammatical through time’ (ibid.: 2), later rephrased as ‘the steps whereby particular items become more grammatical’ (Hopper & Traugott 2003: 2). This latter definition has become (more or less) the standard definition of grammaticalisation, and most often the term refers to how a (more) lexical word develops into a (more) grammatical word and then possibly into a clitic and potentially into an affix (cf. Joseph 2011; Narrog & Heine 2011, 2021). This definition also often means that ‘Grammaticalization changes analytic to synthetic constructions’ (Lehmann 2015 [1995, 1982]: 22).

2.1. Towards the history of grammaticalisation

17Recent decades have observed an increased scholarly interest in the history of the concept of grammaticalisation. Vincent (1980) named Meillet as the coiner of the term grammaticalisation. Lehmann (2015 [1995, 1982]) presented a brief history in the first full book on grammaticalisation, after which some books (e.g. Heine et al. 1991; Hopper & Traugott 1993, 2003) included a section or chapter on its history. This was followed by a PhD thesis and some articles specifically on the history of grammaticalisation (see Lindström 2004a, 2004b, 2004c; Lindström Tiedemann 2005). More recently, grammaticalisation has attracted a growing interest among historians of linguistics. In Joseph (2015: viii, 2023: 110), Meillet is seen as the one who introduced both the term and the concept of grammaticalisation. Thereby, Meillet differs from his predecessors insofar as he perceives several diverse changes as being part of a single process. However, Joseph (ibid.: xiv) points out that Meillet’s concept of grammaticalisation is not quite the same as the post-1970s concept and explains this through a difference in the understanding of the grammar and the lexicon. McElvenny (2016, 2020) emphasises Gabelentz’s role in the history of grammaticalisation and reiterates the link between Meillet’s concept and agglutination theory, as previously pointed out by Lehmann (2015 [1995, 1982]: 3–4) among others. However, although the two overlap, the concept of agglutination has been shown to be different from grammaticalisation (Lindström 2004b; Lindström Tiedemann 2005), as we will come back to in Section 3.1.

21Even though Heine (2003: 576) recognises that ‘Indo-Europeanists’ ‘used findings on grammaticalization as part of their methodology in historical linguistics’ even during the years when this area of study declined (cf. Lehmann 2015 [1995, 1982]: 7; Hopper & Traugott 1993: 24), it has not been clear whether Givón was acquainted with Meillet’s concept of grammaticalisation, since he did not make any explicit reference to Meillet in his work (cf. Lehmann 2015 [1995, 1982]: 6). In a self-reflection on the source of his own ideas regarding the development of grammar, Givón (2018: 151) admits ‘re-inventing’ grammaticalisation ‘in blissful ignorance of Meillet’, as noted by Fagard and Fortis (2025). A few years after his presumed reinvention of grammaticalisation, Givón used the term ‘grammaticization’ (Givón 1975: 49), initially with quotation marks, just like Meillet had done (see chapter 4 below), and another couple of years later he started to use the term grammaticalisation (Givón 1977; Lindström 2004b: 355). Givón’s silence on Meillet’s theory of grammaticalisation is understandable because at the time, the history of linguistics as a field of its own right was in its infancy, and it was not commonplace to delve deeply into the history or to read the classics. Furthermore, the concept of grammaticalisation seems to have been considered part of the generally established knowledge within the field of historical linguistics, and therefore, it probably was not seen as requiring explicit reference to Meillet.

2.2. Subprocesses of grammaticalisation and the linguistic spiral or cycle

24Since Lehmann (2015 [1995, 1982]: chapter 4), grammaticalisation has often been described in terms of six ‘parameters’ or ‘criteria’ which affect the degree of autonomy of the sign both on the paradigmatic and syntagmatic axes (cf. Narrog & Heine 2021: chapters 2–3). The loss of autonomy of a linguistic sign in the ongoing process of grammaticalisation involves its weight, cohesion and variability, as explained by Lehmann (2015 [1995, 1982]: 130–131):

1. Paradigmatic weight, or phonological and semantic integrity of a sign;
2. Syntagmatic weight, or structural scope of a sign in a syntactic construction;
3. Paradigmatic cohesion of a sign with other signs in a paradigm;
4. Syntagmatic cohesion, or bondedness of a sign with other signs in a syntagm;
5. Paradigmatic variability, or possibility of using other signs in the place of a sign or omitting it;
6. Syntagmatic variability, or possibility of moving a sign in its syntactic construction.

25Some of the parameters have been criticised, such as structural scope, which Narrog and Heine (2021: 44) point out means something different to many other linguists compared to its definition as one of Lehmann’s parameters of grammaticalisation. Consequently, some scholars have worked with a slightly different set of parameters. Heine and Kuteva (2002, 2007) have chosen to focus on only four parameters or ‘mechanisms’ (Heine & Kuteva 2002: 2) which are:

1. Extension of an element into new contexts;
2. Desemanticisation, or ‘semantic bleaching’;
3. Decategorialisation, or the loss of features that typically belong to the original grammatical category of an element;
4. Erosion, or ‘phonetic reduction.’

26The loss of integrity is often called semantic bleaching and phonological attrition, and they are among the most frequently mentioned processes involved in grammaticalisation (cf. Lehmann 2015 [1995, 1982]: 134). Notably, Meillet (1912: 139) also referred to these processes by which the original word becomes ‘unrecognisable’ (Fr. ‘méconnaissable’), mentioning ‘the weakening of form’ (Fr. ‘affaiblissement de la forme’) and ‘the weakening of meaning’ (Fr. ‘affaiblissement du sens’). In Meillet’s time, these were not new ideas. The loss of integrity was expressed in some form, mainly in relation to the development of affixes, throughout the 19^th century ever since the work of the earliest comparativists, such as August Wilhelm von Schlegel (1818) and Humboldt (1822; cf. Lehmann 2015 [1995, 1982: 2]). The process of weakening destroys old structures, and at the same time it creates new structures by condensing the old. However, as Lehmann (2015 [1995, 1982]: 135) points out, phonological attrition is ‘omnipresent in linguistic change’, and thus, its occurrence alone does not lead to grammaticalisation. Similarly, semantic bleaching can also take place without resulting in grammaticalisation (ibid.: 136).

27The process of grammaticalisation has also been related to a metaphor of a cycle or spiral. Often, cycle and spiral have been used synonymously to emphasise the continuous and repeating nature of linguistic structures. Traditionally, the cycle has been associated with the typological cycle (the evolution from isolating languages to agglutinating and then to flexional languages), but it has also been used to illustrate the macro-level transformation of Indo-European languages from synthetic to analytic. The metaphor of the cycle suggests the possibility of starting the cyclical evolution anew, whereas the spiral refers merely to the progressive decline, or development, that may ultimately result in a state of zero. The so-called Givón’s cycle (cf. Givón 1971: 411, 1979, 2018: 209) offers a synthesis of these two models:

discourse > syntax > morphology > morphophonemics > zero (Narrog & Heine 2011: 7).

28We also find this in Jespersen’s cycle concerning the development of negations (see Jespersen 2025 [1917]) and it also relates closely to the cline of grammaticality, for instance, the one presented by Hopper and Traugott (2003: 7):

content item > grammatical word > clitic > inflectional affix.

29As Narrog and Heine (2011: 7) have quite rightly stated, scholars have considered certain linguistic developments to be cyclical ‘almost as long as linguistics exists as an independent discipline’. Lehmann (2015 [1995, 1982]: 8), in turn, considers that the cyclic or spiral-like conceptualisation of language change in relation to evolutive typology had a significant impact on the development of grammaticalisation theory.

Les langues suivent ainsi une sorte de développement en spirale : elles ajoutent des mots accessoires pour obtenir une expression intense ; ces mots s’affaiblissent, se dégradent et tombent au niveau de simples outils grammaticaux ; on ajoute de nouveaux mots ou des mots différents en vue de l’expression ; l’affaiblissement recommence, et ainsi sans fin (Meillet 1921 [1912]: 140).

Languages thus follow a kind of spiral development: they add accessory words to obtain stronger expressions; these words weaken, they are degraded, and fall to the level of mere grammatical tools; [subsequently,] new or other words are added for expressivity and then the weakening begins again, and so on, without end (our translation).

Avec le temps, le type j’ai dit s’est unifié, et, de bonne heure, en français, c’est purement et simplement une manière d’exprimer l’action accomplie, où l’on ne reconnaît plus la valeur ni de ai ni de dit ; […] il perd sa valeur de parfait pour devenir un simple prétérit ; et, comme il fait alors concurrence au prétérit simple, je dis, […] le prétérit simple a tendu à disparaître […]. L’affaiblissement progressif de la valeur du type j’ai dit a abouti à en faire un simple prétérit, sans aucun reste de la valeur de parfait. Le cycle est désormais parcouru, et, pour se donner un parfait, le français devra recourir à quelque tour nouveau, dont on n’entrevoit pas encore la naissance (Meillet 1921 [1912]: 142).

Over time, the j’ai dit type has become unified, and, early on, in French, it is purely and simply a way of expressing completed action, in which neither ai nor dit is recognised as having any value; […] it loses its value as a perfect to become a simple preterite; and, as it then competes with the simple preterite, je dis, […] the simple preterite has tended to disappear […]. The gradual weakening of the value of j’ai dit has turned it into a simple preterite, with no remnant of the value of the perfect. The cycle has now been completed, and French will have to resort to some new turn, the birth of which we have yet to see (translation CK).

31McElvenny (2016) states that Meillet (1921 [1912]: 133) saw innovation as ‘the key property that sets grammaticalization apart as a unique process in human language.’ Similarly, Lehmann (2004: 184–186) also mentions such language-specific innovations and claims that this ‘is peculiar to grammaticalization’. In this regard, however, it is important to bear in mind the 19^th century saying that evolution creates something new out of something old, a principle that pertains to all kinds of mechanisms of change.

32Many of the sub-processes and meanings of grammaticalisation that we know from modern theories were already in the air in the linguistic thought of the 19^th century. As McElvenny (2016, 2020) has pointed out, the key elements of weakening and a cycle or spiral which Meillet saw as part of grammaticalisation were also present in Gabelentz’ account of agglutination theory (cf. Lehmann 2015 [1995, 1982]: 3–4), and as we mentioned above, both ideas can be found early in the 19^th century. According to Gabelentz (1891: 251), the spiral development is caused by two forces that involve form and function: on the one hand, the strive for ease (Germ. Bequemlichkeitstrieb), known later as the principle of economy, ‘which leads to the attrition of sounds’, and on the other hand, the strive for clarity (Germ. Deutlichkeitstrieb), ‘which prevents this attrition from destroying the language’, and helps to maintain or reinstate the necessary distinctions. This has been interpreted as the first proper explanation of grammaticalisation (Lehmann 2015 [1995, 1982]: 4), but these are also the respective forces behind sound change and analogy.

2.3. Linguistic examples and labels: controversial cases

33Linguistic examples have played an important role in accounts of grammaticalisation. They have often been used to identify, illustrate or explain the phenomenon, especially in a historical context in which the phenomenon did not yet have a name, but they have also been used to provide evidence for further differentiation between concepts (cf. Lindström 2004a, 2004b: 127–137). In some cases, it can be unclear what the primary focus of an example is intended to be, as it can either illustrate a partial aspect of the change, for instance, phonological attrition, or it can be an example of grammaticalisation as such.

34As pointed out by Anttila (1989: 149), grammaticalisation is a process which shows that the demarcation between the grammar and the lexicon is more of a blur than a clear distinction. According to Itkonen (2001: 19), there are cases that are distinctly different from one another at the two extremes of the continuum, and the fact that these cases are linked by a continuum does not refute their difference. However, it is impossible to express with absolute precision where a lexical category ends and a grammatical category begins (Itkonen 2001: 19–20), just as it is difficult to determine when a person who loses hair becomes bald. Examples of grammaticalisation have not only raised linguist’s consciousness about the fuzzy boundary between lexical and grammatical categories, and between the lexicon and the grammar, but they have also given rise to debates concerning the fundamental nature of the process.

36During the 20^th century, the process involved in the creation of new derivational endings has also been a subject of disagreement (Lindström 2004c). Lehmann (2015 [1995, 1982]: 7–8) considers Viktor Maksimovich Žirmunskij (1891–1971; 1966: 83) as the first linguist who excluded derivation from grammaticalisation, while Kuryłowicz is noted to have included it. This debate has continued: Lehmann (1989: 12) uses Gothic haidus (Eng. ‘personality, character, way, manner’) > German heit as an example of lexicalisation, whereas Ramat (1992: 558) cannot understand why this is used as an example of lexicalisation when it is such a ‘fine example of grammaticalization’ (Lindström 2004c: 846). Hopper and Traugott (1993) consider new derivational affixes as the end point on the cline of lexicality, that is, not the cline of grammaticality which instead ends with inflectional affixes (Hopper & Traugott 2003: 7). But the French derivational morpheme -ment was first seen as a ‘straightforward instance of grammaticalization’ (Hopper & Traugott 1993: 131), only to be described more vaguely as ‘a new lexical formative coming into existence out of a formerly autonomous word’ and labelled as ‘morphologization’ in the second edition (Hopper & Traugott 2003: 141). Morphologisation is, however, discussed as a process which is part of grammaticalisation (Hopper & Traugott 2003: 141–142).

37The reason that Hopper and Traugott (2003) toned down the presentation of this as grammaticalisation is probably because, as Joseph (2011: 200) suggests, derivational endings are seen as ‘less grammatical’ than inflections. However, the fact that Romance -ment(e) is a derivational suffix used to create new adverbs, and not nouns, as is the case with -heit, may have a bearing on the perception of the process as either lexicalisation or grammaticalisation, since there is a tendency to regard adverbs as more grammatical than nouns, as we saw with the example of German heute above.

39The predominant 20^th century perspective on grammaticalisation seems to exclude cases which could just as easily be counted among clear cases of grammaticalisation. Among them are the functional reuse of umlauts or word order as a grammatical device, the latter being one of the examples used by Meillet to illustrate grammaticalisation. It seems that a linguist has the option of either being satisfied with the standard definition, such as provided by e.g. Hopper and Traugott (2003), or considering grammaticalisation in an unusually broad sense as ‘any coding of a function within the grammatical system of a language’ (Frajzyngier 2011; cf. supergrammaticalisation in Lindström 2004b, 2005). The discussion is not closed. The concept of grammaticalisation has continued to be refined and developed since its revival in the 1970s, and currently, a number of parallel concepts intended to clarify the complex issues surrounding the phenomenon are in use by some linguists, e.g. constructionalisation (e.g. Hoffman & Trousdale 2013; Traugott & Trousdale 2013), degrammaticalisation (e.g. Lehmann 2015 [1995, 1982]; Norde 2009; Narrog & Heine 2021), pragmaticalisation (e.g. Erman & Kotsinas 1993; Erman 2000; Claridge & Arnovick 2010).

3. Meillet’s concept of grammaticalisation

40Antoine Meillet (1866–1936) used the term ‘grammaticalisation’ and the verb ‘grammaticaliser’ in his frequently cited article (Meillet 1982 [1921, 1912]: 133, 148) in quotation marks (Lehmann 2015 [1995, 1982]: 4; Joseph 2023: 111), possibly as a means of indicating that the term was new. Some years later, the words grammaticalisation and grammaticaliser are still found within quotation marks in his work (Meillet 1982 [1921, 1915–1916]: 169), suggesting that the terminology had not yet been adopted by the rest of the scientific community, or that Meillet was not satisfied with these terms. It seems as if Meillet was trying out a new term and/or searched for confirmation for the crystallisation of a well-known phenomenon as well as its various manifestations and labels by using a single transparent term. Meillet’s term thus fulfils the criterion for something to be a concept: ‘for a word to become a concept, a plurality of meanings and experiences must enter into this single word’ (Prost 1996: 127). However, after 1915, Meillet no longer uses the term grammaticalisation (nor the verb grammaticaliser), even though he does continue to write about changes which could be referred to by this term.

41Meillet was a legitimate heir of the ‘cognitive goods’ (Bod et al. 2019) of the historical-comparative paradigm, including its methods, concepts, and models. He not only had the retrospective advantage of looking back on the development of comparative grammar and shaking its commonplaces, but he was also a representative of a new era that put emphasis on the sociological aspects of language and language use (cf. Bergounioux & Lamberterie 2006; Savatovsky 2006; Joseph 2020; Meyer & Moret 2023). This dual positioning is also reflected throughout in his explanations of language change which cover the language-based macrolevel, as well as the user-based microlevel. In fact, Joseph (2023, cf. 2020) discusses the links between Meillet and some of Saussure’s ideas, for instance, he relates grammaticalisation to the process of adopting changes in parole into langue, but he also sees possible links between grammaticalisation and stylistics.

42In his inaugural lecture as chair of comparative grammar at the Collège de France, Meillet pointed out that the linguistic thought of the 19^th century had recognised three types of language change: sound laws, analogy and borrowing (1982 [1906b]: 4). This shows that the Neogrammarian doctrine still had a strong influence on the understanding of language change at the beginning of the 20^th century (see Meillet 1982 [1921, 1912]: 133). Meillet himself distinguished the universals of change that can be found in any language, that is, regular sound changes to which human anatomy and physiology set conditional boundaries, and value changes based on the isomorphic ideal of one meaning for one form and analogy. Meillet further distinguished three factors that condition language change: (1) physical, anatomical and physiological conditions for the production and reception of sounds; (2) psychological conditions of language perception; (3) sociological conditions which involve the accommodation of one individual to another in a spatiotemporal context of communication (see Meillet 1982 [1900]: 36, 1982 [1908]: 48, 59, 60).

43In recent historical linguistics grammaticalisation has been described as a process of change that can be seen as a potential ‘parallel to sound change, analogy, borrowing, reanalysis, and metaphorical extension’ (Joseph 2011: 194; cf. Anttila 1989: chapter 9.1). In the next section we will take a closer look at the relation between grammaticalisation and other mechanisms of change according to Meillet.

3.1. Agglutination, grammaticalisation and analogy

45Nonetheless, agglutination was included in Saussure’s lectures, likely directly from Bopp, as some of his examples suggest, and it was treated in his first Genevan course on general linguistics (1907–1908) along with sound change and analogy (on Saussure’s diachronic linguistics, see Klippi 2025). The term agglutination, as employed by Saussure, simply reflects its primary lexicographic sense, that is, ‘the gluing together of contiguous parts’ (Littré 1863–1872, s.v. agglutination), and this idea was transferred to its linguistic meaning to refer to the process whereby words in a dependent or contiguous relationship merge to form a single word.

46In agglutination in the traditional sense, according to Saussure, ‘there is not any creation but an acceptance’ by the speaker ‘that the two words form a unity’ (CLG/E [I], I R 2695), such as in tous jours ‘all days’ > toujours ‘always’; au jour d’hui ‘on the day of today’ > aujourd’hui ‘today’; dès ‘from, since’ ja ‘already’ > déjà ‘already’ (CLG/E [I], I R 2680) or the French future je choisir ai > je choisirai ‘I will choose’ (CLG/E [I], I R 2704). Hence, whereas analogy creates a new form by taking as its model pre-existing elements that are not directly observable, since they reside in the unconscious (CLG/E [I] I R 2462), agglutination requires a specific syntactic context. The speaker’s unconscious task is to interpret the linear flow of speech and to assign values to the linguistic segments in relation to the other segments.

47Setting values is normally arrived at through analogical reasoning based on a regular model. However, in the case of agglutination language has made a poor analysis (‘la langue a mal analysé’), as Saussure puts it, in reference to the actual agency of the speaker (CLG/E [I], I R 2592; see Klippi 2024: 134). For instance, the German lexical element lich ‘outer appearance’ loses its independent lexical value (manno lich ‘who has the appearance of a man’, redo lich ‘who has the appearance of reason’) and is reinterpreted by the speaker as a grammatical element, a suffix -lich, as in redlich, which then serves to create new words by adding it to a root (CLG/E [I], I R 2596):

48As it involves an unconscious reanalysis by the speaker, ‘a change of point of view’, to use Saussure’s own formulation (CLG/E [I], I R 2597), agglutination, just like any other grammatical change, obscures the strict division between synchrony and diachrony. This is because the emerging new values manifest themselves first in speech (parole), and thus, have as their starting point the productions of individuals before they are integrated in langue (see CLG/E [I], II R 115, 2215–2216). Furthermore, Saussure (CLG/E [I], I R, 2706–2707) claims that it is often difficult to tell whether a phenomenon is due to agglutination or analogy. Regardless of the process by which the grammatical change occurs, analogy is always an underlying factor for Saussure, required in the processes of reanalysis and extension (see Itkonen 2005: 109–113). To compare his position to the modern standpoint, Hopper and Traugott (2003: 32) claim that analogy constitutes a significant element in the ‘rule spread’ ‘within the linguistic system or community’ involved in grammaticalisation. However, in contrast to Saussure, they differentiate between analogy and reanalysis, considering the latter to be the most significant component in the process of grammaticalisation.

49According to Meillet, the fundamental distinction between analogy and grammaticalisation as mechanisms of change is attributed to the presence of a regular model in analogy, as opposed to the absence of such a model in grammaticalisation. Analogy always requires a model to follow, as emphasised by Meillet:

Le procédé par lequel les formes grammaticales se transforment ne diffère de celui par lequel elles se reproduisent que dans l’application : il s’agit toujours de l’analogie. L’analogie qui fait faire la forme régulière vous téléphonez sur téléphoner, nous téléphonons, etc., d’après le modèle de laisser, nous laissons, vous laissez ne diffère pas de celle qui conduit à une forme contraire à l’usage ancien dans vous disez (au lieu de vous dites), sur dire, nous disons, d’après le modèle de lire, nous lisons, vous lisez, et des verbes pareils (Meillet 1982 [1908]: 47).

The process by which grammatical forms are transformed does not differ from the one by which they are reproduced except in the way it is applied: it is always a matter of analogy. The analogy that creates the regular form vous téléphonez from téléphoner, nous téléphonons, etc., following the model of laisser, nous laissons, vous laissez does not differ from the analogy that leads to a form which is against the ancient usage as in vous disez (instead of vous dites), from dire, nous disons, based on the model of lire, nous lisons, vous lisez, and similar verbs (translation CK).

50However, according to Meillet, grammaticalisation exerts a more profound influence on language than analogy, since it possesses the capacity to transform the entire linguistic system, while analogy may or may not do so. The forms generated through analogy usually pertain to specific linguistic elements and can thus be understood and explained only within a limited portion of a linguistic system (see Itkonen 2005: 13). In his article, ‘Comment les mots changent de sens’, Meillet provides examples which demonstrate the importance of the local context in comprehending analogical creations. For instance, the word chèvre, meaning ‘goat’, is employed in a specific sense to refer to a device used to lift heavy loads (Eng. ‘scaffolding goat’) (Meillet 1982 [1906a]: 246). The extension in meaning can be attributed to the more or less physical resemblance of the object to a goat, based on analogy, as observed in specific circumstances. This is to say that the scope of analogy varies, while grammaticalisation invariably results in a change in the whole linguistic system.

52Meillet criticised Franz Bopp (1791–1867) for having imposed the search for the mythical origins of Indo-European (IE) languages on linguistic studies during the first half of the 19^th century (Meillet 1982 [1921, 1912]: 131–132). Bopp had a firm faith in the power of the comparative method in reaching the primitive forms, and that made him formulate root theory. Influenced by Indian scholars (Lindström 2004b: 211, based on Bopp 1974 [1820]), Bopp believed that languages could be reduced to some one hundred primitive roots that represented the very origins of the IE languages. Based on this, he is seen by Lehmann (2015 [1995, 1982]: 3) as the first to ‘apply’ agglutination theory. However, Lehmann also notes that Bopp was engaged in correspondence with Humboldt at the time, and that Humboldt (1822: 54f; see Lehmann 2015 [1995, 1982]: 2) presented the development of grammatical forms as four stages, of which the third was focused on agglutination. Bopp’s root theory was supposed to prove how the most reduced autonomous forms of the IE protolanguage progressively gave rise to grammatical forms, an idea parallel to certain aspects of grammaticalisation.

53By contrast, Meillet (1982 [1921, 1912]: 131]) considered that Bopp was wrong to claim that with ‘the study of the most ancient types of every language’ (that is, the comparison of Sanskrit, Greek and Latin) a linguist could reach the ‘primitive forms’ that in turn could be ‘analysed in their composite elements’ (roots). For Meillet’s generation of comparativists it was clear that the system of correspondences between historically attested languages, on which the comparative method was based, constituted the only positive reality (Meillet 1964 [1903]: 47; cf. Klippi 2010: 332). After vain and useless effort, as Meillet (1982 [1921, 1912]: 131–132) points out, linguists finally understood that all known languages already represented an advanced state of evolution—even the languages of the so-called primitive peoples and civilisations. Hence, it was completely illusory to try to attain the primitive origin.

54The study of language origins had been forbidden by ‘mainstream’ linguistics in France in 1866, even though the idea to reach the very first origins remained the Leitmotiv in several other disciplines (cosmology, natural history, geology, anthropology) (Naumann et al. 1992; Desmet 1996; Auroux 1998, 2007; Blanckaert 2009; Klippi 2010). By 1912, it was generally accepted that the results of the comparative method, based on the correspondence of attested features in related languages, constituted an absolute historical limit beyond which science could not extend. In this context, any speculation concerning the origin of language was considered to lead to an impasse (Meillet 1964 [1903]: 44, 46, 49–50; see Klippi 2010: 304). Time had superseded the ‘romantic part of Bopp’s work’ and the ultimate goal of his agglutination theory was hopelessly out of date (Meillet 1936 [1923]: 154). As Meillet aptly describes, it had become ‘obsolete’, ‘chimerical’ and ‘ruinous’ (Meillet 1936 [1923]: 154; for more detail, see Fagard & Fortis 2025: §4).

55However, when it comes to the mechanisms of change in a hypothetic primitive state of language, analogy would be excluded as a procedure for creating new grammatical forms, since it requires a model to follow in a pre-existing system, as was suggested above (Meillet 1982 [1921, 1912]: 132). Therefore, the basic idea of Bopp’s root theory was correct. Furthermore, Meillet did not deny the value of other achievements of the first comparativism based on positive evidence. He agreed with the validity of the great transformation of IE languages—their general tendency towards simplification of morphosyntactic structure as a result of the isomorphic ideal. The eradication of redundancy in conjugations and declinations leads to fixed word order and a new concept of word, as Meillet describes:

1. Some of the branches of the IE language family had started to indicate grammatical notions by autonomous words, where the old IE grammar had inflectional forms (Meillet 1982 [1918a]: 69).
2. The characteristic feature of IE grammar is that the syntactic role of the words in a sentence is expressed by the form of these words. Latin, for instance, preserves the IE state, whereas French, English or Persian differ from it in two essential ways: a) fixed word order, and b) that grammatical function is expressed with an additional word, i.e., periphrastically (Meillet 1982 [1918a: 62–63]; 1982 [1909]: 156–157).
3. The essential feature of the morphological structure of IE is that a word does not exist independently of an inflected grammatical form: there is no word meaning ‘cheval’, there is a singular nominative equus, a singular genitive equi, a plural accusative equos etc., whereas in English, and to a lesser extent, in French, words exist independently of inflectional ‘morphemes’, whatever their role is in a sentence (dog, chien) (Meillet 1982 [1909]: 158).

3.2. The importance of syntax

56Meillet’s article on the evolution of grammatical forms is largely about syntax in its etymological meaning in Greek which means that elements are arranged together in a certain order (σύν Eng. ‘together’ + τάξις Eng. ‘arrangement’). This definition of syntax covers both the linear sequence of subunits of the word (cf. morphosyntax), and the sequence of words in a sentence. Some basic concepts of comparative grammar, juxtaposition, composition and derivation, are fundamentally syntactic. Hermann Paul (1846–1921; 1890 [1880]: 88) observed that the principal unit of language is not the word, but the sentence, and that all languages proceed by sentences. According to this position, syntax is the end to which all language tends, since the sentence is considered the most adequate means of expressing thought (Klippi 2010: 275).

Toutefois ces combinaisons originales existent à des degrés divers suivant les individus. Un sujet de culture médiocre parle surtout par formules qui ne varient guère ; chez la plupart des gens les associations de mots ne sont ni libres ni personnelles. Le mot n’est qu’une partie de combinaisons pratiquement constantes ; la valeur du mot dans un pareil ensemble ne s’explique pas par le sens universel et général de ce mot, mais par l’habitude que l’on a de le voir dans certaines combinaisons. Le mot « pied », par exemple, représente quelque chose de bien déterminé lorsqu’il est lié à un adjectif : « un pied large », mais tous les autres emplois ont une valeur différente de celle qu’il a lorsqu’il désigne l’organe : dans le « pied » d’une montagne, d’un meuble, rien n’éveille l’idée de quelque chose qui ressemble à l’organe, de même lorsqu’on dit « pied de porc farci » ou « pied de mouton » ; cela est si vrai que si on avait à parler de l’organe en question on dirait : patte (Meillet 1936 [1924]: 10).

• 18 It is interesting to note that this is one of the familiar names for the mushroom (Hydnum repandum) (...)

However, these original combinations exist to varying degrees in different individuals. A person of mediocre culture speaks mainly in formulas that hardly vary; in most people the associations of words are neither free nor personal. The value of a word in such a set is not explained by its universal and general meaning, but by the habit we have of seeing it in certain combinations. The word ‘foot’, for example, represents something very specific when it is linked to an adjective: ‘a wide foot’, but all the other uses have a different value to that which it has when it designates the body part: in the ‘foot’ of a mountain or a piece of furniture, nothing gives rise to the idea of something resembling the body part, just as when we say ‘pied de porc farci’ [lit. ‘stuffed pig’s foot’ = stuffed pig’s trotter] or ‘pied de mouton’ [lit. ‘sheep’s foot’ = hedgehog mushroom or pig’s trotter18]; this is so true that if we had to talk about the body part in question we would say: paw (translation CK).

58The example of the metaphorisation of ‘foot’, based on analogy, has a long tradition in French lexicography (cf. Klippi forthcoming). Meillet implements it to show that new interpretations could not take place without the syntagmatic axis in speech. At the same time, to be recognised by interlocutors, the innovative ways of using language must be motivated either in relation to the associative axis of the linguistic system or in relation to extralinguistic reality (as the English translations of the word pied [lit. ‘foot’] show), rather than being entirely arbitrary.

59In present-day grammaticalisation, reanalysis equally shows the importance of (morpho-)syntax (cf. e.g. Hopper & Traugott 2003). It can involve reanalysing the boundaries between two words or determining what belongs to a phrase and what does not. It has often been recognised that, at an initial stage, a construction becomes ambiguous as to its structure and interpretation; over time the reanalysis becomes more common, possibly replacing the old analysis or developing as an alternative of its own (cf. Hopper & Traugott 2003).

3.3. Grammaticalisation as creation, innovation and renewal

In verbis etiam tenuis cautusque serendis dixeris egregie, notum si callida verbum reddiderit iunctura novum (Horace, 1926: 454, 46–48, our emphasis).

Moreover, with a nice taste and care in weaving words together, you will express yourself most happily, if a skilful setting makes a familiar word new (Horace 1926: 455, our emphasis).

62As Horace already pointed out, any callida iunctura must not overstep the limits of the acceptance of usus, the social norm, and therefore, any individual innovation must be approved by the rest of the language users. So, the need to be expressive is limited by the need to be understood. However, should the entire community begin to employ the surprising new way of using these words, this would result in a change in the system. This process is particularly salient in the well-known case of the development of French negation, as evidenced by the numerous nouns employed by language users in the post-verbal position to reinforce the negation: pas ‘step’, ‘pace’, point ‘point’, mie ‘crumb’, gote (goutte) ‘drop’, amende ‘almond’, areste ‘fish-bone’, beloce ‘sloe’, eschalope ‘pea-pod’ (Gamillscheg 1957: 753; Hopper & Traugott 1993: 115; see Meillet 1982 [1905–1906]: 248).

64The question is ‘to what extent collective actions manifest themselves spontaneously in individuals’ (Meillet 1982 [1918a]: 73). What can actually be observed is the language change in its state of completion and the individual language users in action, but the change in progress has never been empirically observed according to Meillet, and therefore, for a change to take place there must be a convergence of collective and individual actions (Meillet 1982 [1918a]: 73). When a linguist observes language in action, he is always confronted with specific facts, without being able to provide an explanation of its overall development (other than post hoc), and therefore, any form of historicism would not have a place in the explanation of language change according to Meillet:

les principes posés n’expliquent jamais que des faits particuliers, et ne fournissent que des conclusions particulières ; on aboutit à une poussière d’explications, dont chacune est juste peut-être, mais qui ne constituent pas un système, et qui ne sont pas susceptibles d’en constituer jamais un (Meillet 1982 [1906b]: 7).

the principles laid down always explain only particular facts and lead only to specific conclusions; we end up with a smattering of explanations, each of which may be correct, but which do not constitute a system, and are unlikely ever to do so (translation CK).

65Meillet describes grammaticalisation in terms of creation, innovation and renewal. His use of these terms is not entirely straightforward or transparent, but they were tied to his theory of language change. Meillet seems to use the term innovation in reference to any independent change that deviates from the established norms or grammatical rules defined as ‘the set of procedures by which words are grouped together’ in a language (Meillet 1982 [1918]: 66a). Referring to the creation of ‘new phonological and morphological processes’, he uses the term ‘innovation’ in connection with ‘sounds hitherto unused, new grammatical forms, and even new categories’ (Meillet 1970 [1903]: 117). In his 1912 article, he also refers to the ‘creation’ of new grammatical categories, ‘innovation’ in modern terminology, as grammaticalisation, leading to a change in the system (Lehmann (2015 [1995, 1982]):

La ‘grammaticalisation’ crée des formes neuves, introduit des catégories qui n’avaient pas d’expression linguistique, transforme l’ensemble du système (Meillet 1982 [1921, 1912]: 133).

‘[G]rammaticalisation’ creates new forms, introduces categories that had no linguistic expression, transforms the whole system (translation CK).

66There are also innovations that follow general and universal ‘lines of development’ (Meillet 1982 [1918a]: 61, 66) that can be observed, for instance, in the tendency of the IE languages to develop in the same direction. Traditional comparative grammar expressed this in terms of the shift from analytic to synthetic and from synthetic to analytic languages. Such long-term tendencies, further conceptualised by the term drift by Edward Sapir (1921), could take cyclical or ‘spiral’ form (Meillet 1982 [1921, 1912]: 140; cf. Swiggers 1988: 158). These law-like tendencies could apply because IE languages were thought to share similar conditions. Therefore, this development occurred in a parallel but independent manner across different IE languages. These languages all went through a simplification of the inflectional system which, according to Meillet, shows not only in the general plan of grammar but also in many details (Meillet 1982 [1921, 1900]: 41). However, these tendencies should not be viewed as organic, and thus, seen as an inherent result of the dynamics of the language system, but rather tied to something in the extralinguistic context. Meillet demonstrated a keen awareness of social factors influencing language change, and therefore he emphasised the role of external factors in shaping linguistic evolution even in the most distant historical periods.

67Furthermore, there are what Meillet calls ‘accidental innovations’ which are specific to a particular language and result in an independent development, even if they often intersect with more general tendencies (Meillet 1982 [1918a]: 70). The most important grammatical innovation that distinguishes the Romance languages from Latin is, according to Meillet (1982 [1916]: 317), the destruction of the nominal declension, since it resulted in a ‘complete renewal of the way in which words were grouped’ by replacing the missing case endings with a fixed word order. Another important case that Meillet (1936: 32) cites is the ‘creation’ of periphrastic verb forms in Romance languages from a Latin participle and an auxiliary. Consequently, nothing could be any farther from the Latin form videram ‘I had seen’ than the French sentences ‘j’avais vu’/‘I had seen’ ou ‘je ne l’avais encore jamais si bien vu’/‘I had never seen him/it so well before’ (Meillet 1936: 80). In addition, by changing the mutual values in the past tenses, the introduction of the periphrastic forms into the system led to a ‘renewal’ of the aspectual as well as the modal system of verbs. However, this development is not specific to Romance languages but joins the general history of IE languages which provides a striking illustration of the ‘erosion’ or ‘attrition’ (Fr. ‘usure’) that concern the preterite forms in general (Meillet 1921: 187). Later, as noted in section 2 above, Meillet’s student Émile Benveniste similarly writes about two types of ‘mutations’ (Fr. ‘transformation’, see Benveniste 1968): innovating mutations (cf. innovation) and conservative mutations (cf. renewal; Benveniste 1968: 85–86). He classifies many of the prototypical examples of grammaticalisation as ‘cases of transformation’, e.g. Romance future as a conservative ‘transformation’ (and ‘auxiliation’, Fr. ‘auxiliation’) and the development of a definite article as an innovating ‘transformation’ (cf. Lindström 2004b: 339–340).

68In addition to the general and accidental innovations, there is ‘an emotional tendency towards more intense expression’ which takes place in the ordinary interaction between speakers, and this tendency involves individual choices and needs (Meillet 1982 [1918a]: 69). As Charles Bally (1865–1947) and Leo Spitzer (1887–1960), to whom Meillet refers in a footnote in the 1921 edition of his article on grammaticalisation (cf. Joseph 2020, 2023; Fagard & Fortis 2025), pointed out in their work, not only great writers but also ordinary language users are sensitive to pragmatic and stylistic matters. A speaker uses marked and surprising ways of expressing themself to attract the listener’s attention, as indicated above in reference to Ars Poetica. For instance, in Latin the use of a personal pronoun in front of the verb, as in ego dico, was a way to put emphasis on the speaker (Meillet 1982 [1918a]: 69). In French, the personal pronoun became an obligatory element and the default form, it lost its tonic accent but also its autonomy. To find a more expressive form, a French speaker therefore uses an emphatic pronoun in front of the clitic one: ‘moi je dis’/‘me I say’ (Meillet 1982 [1918a]: 71).

70In summary, Meillet systematically employs the terms création, renouvellement, and innovation in his early articles from 1921 in relation to this type of language change, but less frequently in his 1936 articles. However, the terms are also often used synonymously, and in certain contexts, it can be challenging to differentiate between them. Following the French stylistic conventions, writers are expected to employ synonyms rather than relying on a single term consistently. This becomes particularly problematic in scientific texts, where the use of theoretical terms should be precise and unambiguous.

71As any process of change, grammaticalisation is rooted in the communicative needs of speakers, that is, in their intentions and goals within a particular context of interaction. Although the terms creation, innovation, and renewal primarily refer to individual actions, for change to occur, it requires the acceptance of the broader speech community, which determines the future direction of the language system. This applies to both the observable present and the unobservable past.

72When it comes to the mechanism of change, similarly to grammaticalisation studies in present-day linguistics (cf. section 2.2), Meillet recognised both semantic bleaching and phonological attrition as important features in the process of grammaticalisation (Meillet 1982 [1921, 1912]: 139, cf. Lehmann [2015 (1995, 1982)] integrity]), using the French affaiblissement (Eng. ‘fading, weakening’) for both. According to Meillet (1982 [1915–1916]: 164, 170), the role of the sentence or utterance (French phrase) and the attrition of phonetic volume were of utmost importance in grammaticalisation. He addresses the process of grammaticalisation in many of his articles, but it is surprising that he does not seem to return to the term after the first two articles (1912, 1915–1916).

3.4. The term as a hapax

73The term ‘grammaticalisation’ could be considered as a hapax with enormous longevity and considerable destiny. But how new was the term and the concept to Meillet, and is it his term and concept which we use today? The French terms grammaticalisation, grammaticaliser and se grammaticaliser were included in a French dictionary, Le Grand Larousse de la langue française (Guilbert et al. 1973: 2285, s.v. grammaticalisation, grammaticaliser, se grammaticaliser) in 1973:

grammaticalisation [gramatikalizasjɔ̃] n. f. (de grammaticaliser ; milieu du xx^e s.). Phénomène selon lequel en élément lexical, doté d’un sens défini, devient un élément purement grammatical : « Pendant » est un participe devenu préposition par grammaticalisation.

grammaticaliser [gramatikalize] v. tr. (dér. savant de grammatical ; 1962, Larousse [1^er ex. au part. passé, en 1845, J.-B. Richard de Radonvilliers]). Donner à un élément lexical le statut d’élément grammatical.

se grammaticaliser v. pr. (milieu du xx^e s.). Passer au rang d’élément grammatical.

74The same year, this term appears in the Dictionnaire de linguistique of Dubois et al. (1973):

grammaticalisation

En linguistique diachronique, on parle de grammaticalisation quand un morphème lexical, au cours de l’évolution d’une langue, ou dans la transformation d’une langue en une autre, est devenu un morphème grammatical. Ainsi, le mot latin mens, mentis (à l’ablatif mente) est devenu en français un suffixe d’adverbe de manière dans doucement, violemment, bêtement, etc.

75However, as noted by Marchello-Nizia (2006: 18), the term had had ‘a discreet career in French’ even before that, since it was used in a study on the French interrogatives by C. de Boer as early as in 1926, as well as in a study on the French subordinate conjunctions by J. Herman (1963). As one of the aims of lexicography is to make an inventory of the neologisms related to encyclopaedic knowledge, on the one hand, it is surprising, based on this background, that these technical terms appear in this genre so late. On the other hand, the introduction of this technical term suggests that in French linguistics grammaticalisation was regarded as common knowledge. Indeed, this knowledge had persisted in circulation in between Meillet’s days and its supposed rediscovery in the 1970s, particularly within Romance linguistics, as Fagard and Fortis (2025) have statistically demonstrated.

76The dictionary of Larousse is of the utmost importance since it provides a valuable means of tracing the source for the history of this term, but contrary to what this dictionary claims, it had not only been used in the past participle. Jean-Baptiste Richard de Radonvilliers (1788–1850), a French lexicographer, offers a whole range of terms related to grammaticalisation with their definitions in his Enrichissement de la langue française, dictionnaire de mots nouveaux (1845) (Fagard 2023; Fagard & Fortis 2025):

Grammairisme, s.m. ; système de grammaire, des grammairiens : le grammairisme français.

Grammaticalisable, adj. des 2 g. ; qui peut être, doit être grammaticalisé, qui est susceptible de se grammaticaliser, d’ordre, de classement grammatical.

Grammaticalisation, s. f. ; action de grammaticaliser ; état grammaticalisé.

Grammaticalisé, e, part. pas. et adj. ; qui est rendu grammatical. qui est dans un ordre, dans un classement grammatical, qui a de la grammaticalité, qui est selon les règles de la grammaire.

Grammaticaliser, v. act. et pr., se‒ ; rendre, devenir grammatical, donner de la grammaticalité : faire entrer, entrer naturellement dans un ordre, dans 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.

Grammaticalisme, s.m. ; système de grammaticalisation.

Grammaticalité, s.f.; état, qualité de ce qui est grammatical, science grammaticale.

77This suggests that there is a distinction between the terms grammaticalité (what is grammatical) and grammaticalisation (what is about to become or what has already become grammatical) and grammaticalisable (what can become grammatical). All terms which make it vital to figure out what was considered as grammatical, especially if we want to understand which kind of changes were conceived as grammaticalisation in the eyes of French linguists at this time and by Meillet in 1912.

78Furthermore, it seems this very early occurrence of the term grammaticalisation in a general sense of becoming grammatical is an early hapax since we find no references to grammaticalisation in the Dictionnaire de l’Académie française, nor in the great dictionaries (‘dictionnaires de langue’) of the 19^th century, nor in the Encyclopédie of Diderot and d’Alembert from the 18^th century.

79For comparison, in the Oxford English Dictionary (OED) (2024) ‘grammaticalization’ in the linguistic sense of ‘[t]he process by which a lexical item, construction, etc., develops or acquires a grammatical function’ is listed with examples (in the form of citations from publications) from 1918 with quotation marks. Subsequently, there are examples from 1936 and 1965, and an explicit etymological reference to an influence from the French grammaticalisation, mentioning Meillet (1912), is included in the current edition. The verb grammaticalize has examples of usage from 1941 in the current sense, and from 1810 in an earlier obsolete sense of ‘[t]o write (a grammar).’ Additionally, the OED (2024) lists ‘grammaticize’ as first occurring in 1673 in the meaning of ‘[t]o discuss grammar’, from 1791 in the sense of ‘[t]o make (something) grammatical’, and finally from 1969 as ‘[t]o express (a concept) by grammatical rather than lexical means’ or ‘to develop or acquire a grammatical function’. Considering the derivation of the English term, it is possible that more than one person coined the same term independently, and hence a possible coinage taking place also during the revival. This is supported by the existence of similar derivations such as grammaticisation, morphologisation, pragmaticalisation and others (cf. section 2.3).

4. Final remarks

80Sometimes it is difficult to keep terms and concepts apart. A concept may be considered as a representation created by thought, whereas a term is its linguistic expression. According to the history of concepts, we should not assume that concepts do not exist before there are terms for them. Nor should we assume that a term always means the same thing. We should also be especially careful not to interpret terminology and concepts of the past as the same as today, but also in using modern terminology to describe the past. However, without terminology, we lack the means to refer to concepts. Simultaneously, it is equally true that discerning which concept a term refers to is not always a trivial task.

81Meillet’s concept of grammaticalisation bears a resemblance to the past concept of agglutination but also to the present-day concept of grammaticalisation, without being identical to either (Lindström 2004b; Lindström Tiedemann 2005). Agglutination included the development of both lexical and grammatical affixes, and it focused on the process of univerbation, the merger of two or more autonomous linguistic units, and did not include the development of new grammatical words. Meillet’s grammaticalisation is broader than agglutination, and even broader than today’s concept. It is the meaning attached to grammar and grammatical (in comparison to lexicon, lexical) that makes Meillet’s concept different from today’s concept (cf. Joseph 2015).

82There are clear indications that Meillet proposed grammaticalisation as a superordinate term to bring together the different types of change that had already been presented by then (see also Joseph 2023: 110). The concept, including the coinage of the term grammaticalisation, was an excellent innovation by Meillet. He recognised it as ‘a single process’ encompassing a series of ‘diverse changes’ (Joseph 2023: 110), whereas other scholars focused on the composite changes involved. This is also why he compared this process to analogy, according to him, one of the three forms of language change, as identified by the Neogrammarians alongside sound changes and borrowing (Meillet 1982 [1906b]: 4). Nevertheless, it should be noted that even today, linguists sometimes question whether grammaticalisation is one process, many processes, or maybe rather the result of a variety of processes (Joseph 2011).

83Meillet’s description of grammaticalisation as ‘le passage d’un mot autonome au rôle d’élément grammatical’ (‘the transition of an autonomous word into the role of a grammatical element’) (Meillet 1982 [1921, 1912]: 131) is similar to Hopper and Traugott’s (1993, 2003) definition of grammaticalisation as a process, but their description of the framework of grammaticalisation also bears parallels to Meillet’s understanding of grammaticalisation. Indeed, Meillet himself seems to have been interested more broadly in how the speech community comes to agree (unconsciously) on new constructions, which can lead both to new affixes and to the fixation of word order to convey grammatical meaning.

84Strictly speaking, Meillet should not be considered as the coiner of the term since Richard de Radonvilliers had already reinvested the linguistic potential of the French language in the neologisms grammaticalisation, grammaticaliser, etc. in 1845 in a general sense of something becoming grammatical (see Fagard & Fortis 2025). However, Meillet should still be considered to have created the theoretical term of grammaticalisation in an abstract meaning requiring a precise definition within linguistic theory. Unlike his contemporaries, Meillet was able to raise the level of abstraction and discern the varied expressions of similar phenomena. It is this sense that sets Meillet’s term apart from the existent labels. As is the case of the Greco-Latin grammatical terminology which is still used to describe the languages of the world (see Auroux 1994), the concept of grammaticalisation by Meillet is flexible and open enough to be able to gather a plurality of meanings. In this sense, some might say that grammaticalisation ante verbum could be seen as a ‘key notion’ during the 19^th century, as claimed by Heine (2003) (cf. also Lindström 2004b; Lindström Tiedemann 2005).

85As is also the case with analogy (see Klippi forthcoming), grammaticalisation as a concept combines the characteristics of Ideal-typ and transversality. On the one hand, the term Ideal-typ, as used by Weber (1949 [1904]), refers to the notion that all concepts possess an ideal core, enabling their use in a variety of contexts. On the other hand, concepts are also transversal, in so far as they may occur in different contexts carrying specific meanings, while still retaining their own recognisable features. Similarly, discourses related to a concept are also transversal—that is to say, they contain or can have several different, context-dependent transversal meanings of a concept. Because of the Ideal-typ, it is possible to understand and compare diverse discourses on grammaticalisation despite time and space, even when there is no explicit term or when the term is different. But because of the transversality, i.e., its use in different contexts, the concept of grammaticalisation gives rise to controversies as to its precise meaning. In contemporary linguistics, the heterogeneous nature of grammaticalisation has frequently been emphasised, and discussions regarding the distinction between grammaticalisation and lexicalisation have been commonplace.

86It would be unfounded to dismiss Meillet’s work on grammaticalisation as a mere parenthesis in the history of grammaticalisation, as Heine (2003) appears to do, since it played a pivotal role at the beginning of the 20^th century and had a significant influence on historical linguistics after its rediscovery. Meillet’s description of a change from an autonomous word to a grammatical form helped shape what we see as grammaticalisation today, even though it might not have been his own understanding of it, but an anachronistic interpretation of his views. However, Meillet’s non-use of the term grammaticalisation after the 1915–1916 publication still requires explanation, especially as he continues to discuss this type of change in his later work. One hypothesis that needs further clarification could be that Meillet’s silence coincided with the publication of the Cours de linguistique générale (CLG) in 1916. The concept of value, which is largely developed in the CLG, might be considered to make the concept of grammaticalisation a redundant term and concept, since it can help us explain how ‘cutting the linear chain of speech to arrive at units can also be a source of error, to the point of causing a change in the system’ (Klippi 2024: 134). The repeated appearance of certain linguistic elements on an identical linear syntagmatic axis may result in a reinterpretation of their values. This may involve the reanalysis of a lexical element to become a grammatical element, or the reanalysis of a grammatical element to become another grammatical element (Klippi 2024: 134). In accordance with Béguelin’s (2010) proposal of an ‘anticipated criticism of the concept of grammaticalisation’ by Saussure, it is possible that Meillet may have perceived a posthumous critical voice from his esteemed teacher and colleague. Consequently, he chose to abandon the term ‘grammaticalisation’ while nevertheless retaining the underlying concept.

87Grammaticalisation is a multifaceted phenomenon. Studying the history of the concept of grammaticalisation is complicated by the fact that the same term may be used for different concepts, and that concepts may be referred to by different terms. Moreover, examples can illustrate either changes that are related only to grammaticalisation, or changes that can be involved in grammaticalisation, but that can also appear independently without grammaticalisation (e.g. semantic bleaching, phonological attrition, reanalysis, lexicalisation). We still know little of how the changes which we now view as grammaticalisation were understood in France in the latter half of the 19^th and the beginning of the 20^th century. There is clearly a need for more careful study. Future research should look more closely at the work of other French linguists at the turn of the 20^th century and examine how they understood changes which we now consider as grammaticalisation, but also at their understanding of grammar and grammatical, possibly in comparison with their notion of what is lexical.