Usage-Based Motivations for Diachronic Language Change

Definition

With the understanding that language variation, whether synchronic or diachronic, is, for the most part, not random but systematic and predictable, linguists and philologists have long engaged with the task of providing accounts and explanations for how a phenomenon in a given language moves from stage a to stage b, with a temporally preceding b. A usage-based approach to diachronic change takes as its basis a fundamental link between usage and the evolution of linguistic structure through time. More specifically, motivations for language change are sought in areas that involve speakers using language and interacting with one another, such as discourse and communication, as well as cognitive processes involved in producing and understanding language. Mechanisms that contribute to language change include pragmatic inferences, frequency, (inter)subjectification and discourse management (turn-taking).

1. Historical Introduction

In the early 19th century, a period that saw a renewed interest in the diachronic development of languages, both as individual entities and in relation to each other, von Humboldt [1] (p. 57) recognized that ‘Language itself is not a product (ergon) but an activity (energeia)’. This acknowledgment of the dynamic nature of language, which had its earliest echoes in antiquity in the works of Plato, Aristotle and Varro, was complemented by the identification of correspondences between ancestor languages and their offsprings by several of von Humboldt’s contemporaries, who were focusing on sound change. Jacob Grimm and August Schleicher were among the first to formulate the idea that sound change follows regular patterns, hence explicitly rejecting randomness. Their observations were later reformulated in stronger terms by the so-called Neogrammarians, a group of German scholars who focused on sound change: Leskien [2] was the first to claim that sound laws have no exceptions, a statement later mirrored in what has become known as the Neogrammarian’s manifesto, ‘Every sound change, insofar as it occurs mechanically, takes place according to laws that admit no exception’ ([3] (p. xv)). The scene was set for the recognition of the general systematic nature of language change.

By the 20th century, the focus of investigation for language change had shifted to syntactic structures: identifying recurring patterns in the diachronic change of lexical items, Meillet [4] recognized the existence of a process by which an autonomous word evolves into a grammatical element. Implicit in this is the generalization that language change is directional, one of the core tenets of grammaticalization theory, which took form in the second half of the 20th century. The predominantly observed path sees a lexical item (a free word) gradually weakening on the phonetic (erosion), semantic (bleaching) and syntactic (loss of independence) levels, and becoming an inflectional affix (a morpheme). This generalization is captured by Givón’s [5] (p. 413) famous observation that ‘today’s morphology is yesterday’s syntax’.

Much work in grammaticalization theory in the second half of the 20th century focused on identifying recurring directional paths of change, so-called ‘grammaticalization clines’, sequences of sequential and ordered steps that captured the gradual nature of change and were found to hold across languages. Underpinning the validity of such clines was the recognition that ‘forms tend to undergo the same kinds of changes’ [6] (p. 6).

The reasons adduced for such systematic and predictable changes vary depending on the theoretical perspective. Two of the main approaches to language change are the usage-based, or cognitive, and the formal mentalist, or generative. Their diverging views stem from different understandings of what language is, where change occurs and what drives it. In the rest of this article, we explore these two main positions and, focusing on the usage-based approach, we explore four main motivations for language change (and grammaticalization): pragmatic inference, frequency, (inter)subjectification and discourse management in the form of turn-taking.

2. The Nature of Language, Language Change and Motivations: The View from the Formal and Usage-Based Perspectives

Given the remits of this contribution, the aim of this section is to offer a brief overview of the main points about the nature of language, language change and its motivations that set the formal and the functional approaches apart, rather than a thorough and in-depth presentation of the two theoretical models in general. It is hoped that readers already well versed in these matters will not find the approach overly reductive, and that those less familiar will not be discouraged.

Casting this discussion in terms of a ‘formalist versus functionalist’ dichotomy is somewhat simplistic, glossing over the internal subtle yet significant distinctions that exist both within the functionalist (e.g., constructional, variationist, typological, socio-cognitive, etc.) and the generative (e.g., Cartographic approach, Minimalism, Principles and Pa-rameters, etc.) models. However, these two ‘orientations’, the formal and the functional, exist [7], and the distinction between them is real both in terms of classifying a linguist’s theoretical identity and in terms of the fundamentals of the two theoretical orientations. Therefore, there is value in pursuing a discussion of diachronic language change comparing their respective positions that can be of interest to researchers and, more widely, the general public.

There are a few generative models that fall under the label ‘formal’; in this brief overview, we focus on two, the earlier Principles and Parameters model of the 1970s and 1980s, and the later Minimalist Program of the 1990s and subsequent years. For a fuller discussion of the relationship between generativism and grammaticalization, the reader is referred to ref. [8]. Similarly, there are a number of theoretical approaches that are compatible with the label ‘usage-based’; here, we broadly interpret the label as corresponding to the general area of Cognitive Linguistics, without necessarily identifying it with any specific subtypes. However, given the focus of this entry on diachronic language change, reference is made to Diachronic Construction Grammar [9,10]. For exhaustive yet succinct state-of-the-art overviews on the meaning of constructions and their role in diachronic language change, the reader is referred to refs. [11,12]. In contrast to formal approaches, as discussed below, a general theory of grammaticalization is highly compatible with usage-based theories, especially Diachronic Construction Grammar [13]; we delve a little deeper into this below in Section 2.2.

2.1. Formal Perspective

The formal approach to language emerged in the middle of the 20th century through the works of Noam Chomsky, and it represented a clear theoretical rupture with earlier structuralist and behaviorist approaches to language. Chomsky’s 1957 book Syntactic Structures [14] is widely regarded as the founding document of generative syntax, since it introduced several of its most fundamental tenets, the main ones being to claim that grammar was an autonomous system (hence separating it from semantics) and to define it as a formal object, not as a simple descriptive taxonomy. Grammars are seen as formal, explicit generative systems that should generate all and only the grammatical sentences of a language; native speakers of a language possess tacit knowledge of its abstract rules. A crucial aspect of generativism is the distinction between competence (mental knowledge of language) and performance (actual language use); this distinction was fully articulated in Chomsky’s (1965) Aspects of the Theory of Syntax [15], and the move legitimized abstraction away from usage and variation, marking a sharp break from functionalist and sociolinguistic traditions of the time. This distinction, which was later recast as the opposition between I-language (the internal, innate linguistic knowledge in the mind of a native speaker) and E-language (the external, observable language that we hear and see in the world), is at the core of generativism. In the same volume, Chomsky also laid out the new goals of syntactic theory, which should not just adequately describe a grammatical system, but also account for how language is acquired by children, reaching explanatory adequacy: this led to the postulation of Universal Grammar (UG), the genetically endowed blueprint of human language responsible for acquisition.

Later developments saw a shift in the role assumed to be played by UG. In some cases, refs. [16,17,18], the shift involved a move away from the centrality of UG and towards factors that are not specific to the Faculty of Language, the so-called ‘third-factor principles’, ‘general properties of organic systems’ [19] (p. 105) that include principles of ‘efficient computation’ [16] (p. 3), such as ‘Minimal Search, Determinacy and Economy’ [18] (p. xi). In others, ref. [20], the shift saw a much-reduced UG. To a richly structured UG, a slimmer system was opposed, seeing as possibly only including invariant operations such as Merge, Agree and Copy, an underspecified template for formal features (as opposed to a pre-set list of substantive features such as Gender or Number), and a distinction between optional and non-optional formal features.

The relationship between generativism and grammaticalization has been fraught since the beginning due to fundamental differences in conceiving of language and the locus of change [8]. The main areas of conflict include the generativists’ focus on I-language and their belief in the autonomy of syntax from other modules of language: ‘[n]ative speakers of extinct languages are unavailable, and that makes it hard for generativists to study their I-language’ [8], and the emphasis placed on semantic and pragmatic factors in grammaticalization processes is problematic when syntax is seen as an independent module of language.

A defining feature of the generative approach is that grammaticalization is not treated as a primitive mechanism of change [21]; rather, it is reinterpreted as the epiphenomenal outcome of independently motivated syntactic reanalyses that take place during the process of language acquisition. Lightfoot [22,23] is the main exponent of this view, asserting that there are no principles of grammatical change (in the sense of predictive laws), nor is there a process of grammaticalization as such: what we observe are changes in grammars, not changes in forms. Therefore, the locus of language change is the process of language acquisition, and the mechanism responsible for it is reanalysis, a discrete shift in underlying syntactic representation. While adults’ language can only change in minor ways, children acquiring their native language reorganize and reanalyze the system by restructuring it; the outcome of this is not gradual but an abrupt change. Grammar shifts from structure A to structure B in acquisition [24], and any observable gradualness in the shift from A to B is an illusion due to the coexistence of grammars with structure A and grammars with structure B, and to any sociolinguistic competition that may arise between the two. Under this view, grammaticalization clines are nothing more than descriptive generalizations with no explicative power.

To find an explanation for language change, linguists need to turn to the system that generates grammars, the linguistic genotype, more specifically, to changes in parameter setting that happen during language acquisition and that are triggered by a reanalysis of the primary linguistic data. Language change results from parameter resetting that reanalyzes a lexical item as a functional head [25] in the earlier formal tradition, or from small changes in features of lexical items [26] in later developments. Specifically, lexical items are reanalyzed as being specified for uninterpretable features that continue to drive the derivation, resulting in their further upward movement. Grammaticalization is the outcome of upward and leftward movement (i.e., movement to a higher position within the syntactic tree) of lexical items to functional heads: a lexical head that moves to a higher functional projection is reanalyzed as a functional head that is merged in that position [25]. As a consequence of this, elements that undergo this type of change appear to follow a specific, directional and cross-linguistic trajectory: the grammaticalization clines mentioned above are nothing more than the surface outcome of a process of reanalysis of lexical elements as higher functional elements along a highly structured, allegedly universal clausal structure [27,28].

The notion of parameters, understood as the locus of language variation, too, has shifted. Their initial formulation in the classical Principles and Parameters theory of the 1970s and 1980s saw them as a finite, pre-programmed set of values provided by UG, located in the brain, that had to be fixed by experience of/exposure to a given syntactic construction. The grammar of any specific language was seen as nothing more than a specification of these innate values. Both principles and parameters were considered part of the human genetic endowment. A fundamental revision [20] envisages them as emergent properties resulting from the interaction of an underspecified UG, the primary linguistic data a child is exposed to, and ‘third-factor’ principles of computational efficiency. Instead of a list of innate ‘switches’, they are understood as a four-way taxonomy (Macro, Meso, Micro, and Nano) that reflects how widely a formal feature is distributed across functional heads in a language. Parameters are formed as the learner navigates a ‘NO > ALL > SOME’ learning path, attempting to organize linguistic data with the greatest possible efficiency. The locus of diachronic language change is still seen as the result of parameter resetting during language acquisition: change occurs when a new generation of learners reanalyzes the linguistic input and converges on a system of parameter choices that differs from the previous generation’s in crucial ways. The degree to which a parameter can be reset is, under this revised view, dependent on the type of parameter itself: while macroparameters are diachronically stable as they are salient and widely expressed in the input, nanoparameters, which affect only individual lexical items, are the most diachronically unstable and prone to change or loss. Grammaticalization is specifically viewed as a microparametric change in which lexical material is reanalyzed as functional material due to the loss of head-movement [20].

By way of a summary, let us highlight the most salient points of the generativist approach to diachronic language change. The Chomskian model takes I-language as the formal object of investigation and asserts the autonomy (and primacy) of syntax. Language change is the result of abrupt changes that occur in the process of first language acquisition as a result of reanalysis. Reanalysis is, in turn, the result of a process of parameter resetting. Grammaticalization is not recognized as an independent and primitive mechanism of change; rather, it is seen as the result of a process of syntactic reanalysis, involving the upward movement of a lexical item to a higher functional position over time. The driving mechanism behind it is purely structural, and it has been variously interpreted as the result of a parameter resetting or the reanalysis of the features associated with a lexical item.

Let us now turn to the usage-based/cognitive perspective.

2.2. Usage-Based Perspective

Usage-based theories emerged from functional linguistics in the 1970s and 1980s, and aligned with its rejection of autonomous, purely formal models of grammar, placing themselves in opposition to the Chomskian model. Seminal contributions were Givón’s On Understanding Grammar [29], whose core aim is to redefine how grammar must be understood, that is, in relation to discourse, communication, cognition, and diachrony, rather than as an autonomous formal system, and Hopper’s Emergent Grammar [30], whose central goal is to challenge static, autonomous models of grammar, arguing that grammar is not a fixed system, but it emerges from language use, and it is changed by it. Both publications established the idea that structure reflects patterns of usage rather than abstract rules, and both underlined the essential role played by diachronic change in understanding synchronic structure, anticipating later developments in grammaticalization. The usage-based theories see language as ‘an embodied and social human behavior’ [31] (p. 953), shaped by the ongoing experience of communication. The idea of embodiment, that is, we experience and understand the world through the body, and that this embodied experience shapes language, is central to Cognitive Linguistics: because we access reality through embodied cognitive systems, concepts are partially metaphorical (‘The essence of metaphor is understanding and experiencing one kind of thing in terms of another’ [32] (p. 5)). From recurrent patterns of sensori-motor experience, which are pre-linguistic and provide sources for schemas (primitives that structure rich images), we abstract to concepts. For example, from bodily interaction such as being inside rooms, moving through space, etc., we derive image schemas such as states conceptualized as a CONTAINER (in/out; to be in love, to be out of trouble, to be slowly getting into shape, etc.), or in terms of ORIENTATION (to have high hopes, to be in low spirits), etc. Within usage-based theories, explanations for linguistic structure and change are sought in the context of actual language use and are therefore impacted by the communicative dynamics between speaker and hearer, social factors and cognitive processes. Unlike paradigms that focus on abstract competence (such as generativism’s focus on I-language), usage-based theories use evidence from usage patterns, frequency of occurrence, variation and change to gain direct clues about the cognitive organization of language. Because grammatical knowledge is inseparable from the experience of using language, ‘[n]o relevant methods for gaining evidence about language are excluded; studies of corpora, large and small, diachronic data, psycholinguistic experiments, cross-linguistic comparison and child language development all provide essential data for constructing a comprehensive theory of language.’ [31] (p. 953).

Linguistic structure is not an innate, static blueprint but an emergent phenomenon resulting from the interaction between cognitive processes and social usage events. The theory posits that units and structures, which range from phonemes to complex syntax, arise from the brain’s automatic tracking of transitional probabilities, transitional patterns, and categorization of communicative events. Crucially, a usage-based model views language as an extension of other cognitive domains [33] (p. 1180), and ‘seeks to derive the mechanisms of language from more general and basic capacities of the human brain’ [31] (p. 955).

In the first decade of the 2000s, usage-based researchers developed detailed mechanisms for how constructions arise from experience. Three fundamental elements were included in the formal apparatus: exemplars, networks and constructions. An exemplar, a notion imported from cognitive psychology’s exemplar theory of categorization [34], is a richly specified, stored memory representation of every token of language use, preserving fine-grained phonetic, semantic, structural, and contextual detail, which is continuously updated through experience. Linguistic categories and generalizations emerge from the distribution, frequency, and similarity relations among large sets of such exemplars. First introduced in Bybee’s work on dynamic network models of morphology [35], and later explicitly conceptualized by Diessel [36], networks represent the way a language user’s knowledge is organized: a network of interrelated elements, rather than as a set of abstract rules. Links are established by similarity, phonological or semantic, during categorization, facilitating analogy, by which speakers extend patterns to new items. Constructions are the basic units of grammar, ranging from morphemes to large syntactic strings. They are conventionalized, non-compositional form-meaning pairings, and they were explicitly introduced as a structural pattern that pairs syntactic form with semantic and pragmatic interpretation [37]. In later work, they were further formalized as primitives of grammar [38] and became a core notion of usage-based theories. More recent work has both consolidated the core idea and formalized how constructions relate [36,39]: families of related constructions, inheritance relations, probabilistic weighting, and dynamic network models that tie cognitive processes (memory, categorization, sequence learning) to constructional structure. Diessel’s Grammar Network explicitly models grammar as a layered network shaped by usage; Goldberg’s work continued to develop how constructions license argument structure and generalization. At the same time, computational and corpus methods (probabilistic models, distributional analyses) have provided quantitative tests of constructional hypotheses.

The importance of a diachronic perspective in understanding synchronic structure had already been brought to the fore by Givón and Hopper; it was further expanded by Bybee, Perkins and Pagliuca’s (1994) The Evolution of Grammar [40]. A particularly influential work, Bybee et al. [40] showed that grammatical categories and constructions develop diachronically in systematic, cross-linguistically recurrent ways, and that these developments are driven by language use, meaning, and frequency, rather than by abstract syntactic parameters, providing strong support for a meaning-driven model of grammatical change. The recurrent grammaticalization paths witnessed across the 76 languages surveyed support the view that they are, in fact, universal (typological regularities are expected because the same cognitive processes operate across languages), and movement along these paths proceeds in one direction only. Some of the highly recurrent paths are the development of a motion verb into a future marker, of a demonstrative into a determiner, and of a lexical verb into an auxiliary and then into an affix; Bybee et al. [40] argue that these recur because the same usage-based cognitive and pragmatic mechanisms operate across languages. Frequency, too, is identified as essential for grammatical evolution, since high-frequency forms become more grammatical, frequent constructions undergo phonetic reduction, and repetition leads to entrenchment. Later on, the role of frequency was to take center stage in usage-based theory [41,42]. Heine, Claudi and Hünnemeyer [43] recognize the fundamental role of context, explicitly discussing context types enabling reinterpretation, and develop context-based models of grammaticalization.

With its roots in the 1980s and 1990s, at the intersection of Construction Grammar [37,38] and grammaticalization theory [6,41], a usage-based, constructionist approach to language change emerged that modeled diachrony as the dynamic restructuring of a network of constructions: Diachronic Construction Grammar. It was consolidated as a theoretical framework by Traugott and Trousdale’s Constructionalization and Constructional Changes [9], who introduced the notion of ‘constructionalization’, the emergence of a new construction/node in the grammar, and envisaged grammar as a network whose nodes and links change over time. Further work in the first decade of the 2000s contributed an explicit formulation of usage-based theory as a coherent framework: particularly pivotal were Bybee [42,44] and Tomasello [45], whose work emphasized learning, frequency, and entrenchment as central explanatory factors, underlining the strong link between diachrony and synchrony again.

The three fundamental elements of usage-based theories mentioned above are central to how language evolves over time. In diachronic change, exemplars track frequency and variation. High-frequency exemplars become entrenched, leading to phonetic reduction or the maintenance of irregular forms, while marginal exemplars may decay and be lost. Exemplar models are therefore well suited to explaining gradual change, variation and frequency effects. Networks allow for analogy and the emergence of new units. Links establish how parts of words (e.g., affixes) are recognized and how irregular patterns are maintained through similarity. Networks also allow for gradience: for instance, as a word moves from a lexical to a function word, its position and links within the network shift gradually. New constructions arise from specific exemplars of existing constructions. Over time, constructions can become autonomous, losing their semantic and syntactic transparency as they move away from their source forms.

As mentioned above, the diachronic perspectives of grammaticalization theory and usage-based approaches are highly compatible, as they share fundamental beliefs: both frameworks are usage-based and functionalist, rejecting formalist models that separate language from its use. They both prioritize understanding the relationship between structure and use, as well as the role of human cognition and social behavior in language change. Diachronic Construction Grammar and grammaticalization theory are particularly in tune with each other: it has been recently claimed that the concept of grammaticalization is a natural precursor to Diachronic Construction Grammar [13]. Modern grammaticalization studies increasingly emphasize that grammaticalization occurs within specific syntactic contexts, or constructions, the basic units of grammar in Construction Grammar. Both approaches invoke the same explanatory factors, such as frequency, iconicity, metaphor, and metonymy, and Diachronic Construction Grammar is uniquely equipped to map the gradual nature of language change, allowing for the representation of non-discrete categories and indeterminacy, which have long been central concerns in grammaticalization theory. The authors propose that grammaticalization should not be viewed as an independent mechanism of change but as a process, a sequence of steps or stages similar to biological evolution, that is driven by iterative interactions between lower-level mechanisms like reanalysis and actualization [13]. They argue that the entire enterprise of grammaticalization is modelled perfectly within Diachronic Construction Grammar through constructions.

One of the core tenets of usage-based theories is that grammar emerges from language use, an assumption that rejects a strict separation between grammar and usage, and, on the contrary, recognizes the indissoluble relation between form and meaning (witness the form–meaning pairings that are the base for constructions). The close interconnection between grammar, semantics and pragmatics is particularly salient in the usage-based approach to language change, where semantic-pragmatic change typically precedes and motivates structural change, that is, meanings change first, and grammatical structure adapts afterward to conventionalize those new meanings. This position directly contrasts with the generativists’ perspective that locates change in the acquisition of grammar via reanalysis of input. With their main goal being the demonstration that semantic change is regular (in that it is not arbitrary, it follows recurrent pathways, and it can be explained by general principles of language use and interpretation, constrained pragmatic processes shared by speakers and hearers), Traugott and Dasher [46] (p. 42) express very clearly the functionalists’ view on language change: ‘[t]he hypothesis is that innovation and change does not occur primarily in the process of perception and acquisition, but rather in the process of strategic choicemaking on the part of SP[eaker]/W[riter] and interactional negotiation with AD[dressee]/R[eader]’. Speaker–hearer interaction and the constant negotiation of the common ground are where change arises and where it is motivated. Therefore, semantic change originates in contextualized language use; speakers use expressions in particular contexts, and hearers draw so-called invited inferences, pragmatic enrichments encouraged by context (and possibly by the speakers themselves). These inferences recur in similar contexts, and recurrent inferences may become conventionalized, that is, encoded as part of a word’s (new) meaning. This is formalized in Traugott and Dasher’s [46] Invited Inferencing Theory of Semantic Change.

In this section, we have explored the main tenets of usage-based theories about language and language change. Language is understood as a social and embodied human behavior that evolves and changes through use. Diachrony is essential to explain synchronic structure: without an understanding of its diachronic development, it is impossible to explain the synchronic behavior of a given element. Diachronic language change is seen as the natural consequence of repeated language use, shaped by frequency, inference, and interaction, and resulting in systematic, cross-linguistically recurrent patterns. Recurrent pathways are explained through cognitive and communicative pressures, not innate parameters. Usage-based approaches to diachronic language change are perfectly aligned with theories of grammaticalization, particularly in Diachronic Construction Grammar. Semantics and pragmatics are intrinsic parts of grammar: semantic change is brought about by pragmatic processes, and only subsequently does grammar adapt to the new, conventionalized meaning. Semantic change mechanisms such as metaphorical extension and conventionalization of implicature are linked to recurrent grammatical paths. These mechanisms are cognitive and pragmatic in nature.

Table 1. Summary comparison of the generativist and functionalist views.

Usage-Based	Formal
Language as the object of investigation is a social and embodied human behavior	Language as the object of investigation is the native speaker’s internal, innate and genetically endowed system (I-language)
Grammar emerges from use	Grammar is innate
Broad approach to data suitable for investigation: corpora, diachronic data, psycholinguistic experiments, cross-linguistic comparison and child language development	Narrow approach to data suitable for investigation: speakers’ I-language
Diachrony is essential to understand synchrony	Diachrony is problematic, since native speakers of extinct languages are unavailable
Change is gradual	Change is abrupt
Change is triggered in language use	Change is triggered by re-analysis in the process of acquisition
Semantics (and pragmatics) are integral parts of grammar	Grammar is autonomous and dominates
Frequency is causal	Frequency is epiphenomenal
Grammaticalization clines are the outcome of the universality of cognitive processes	Grammaticalization is neither explanatory nor a primitive

offers an overview of the main differences between the generativists’ and the functionalists’ perspectives.

The next section explores some of the factors that are considered motivations and triggers for language change in the usage-based approach through a few case studies.

3. Motivations for Diachronic Language Change: The View from the Usage-Based Perspective

The factors that usage-based approaches identify as triggers for language change fall out of their understanding of the nature of language and diachronic change. Usage-based theories see diachronic change as gradual, incremental and driven by communicative interactions in the speaker–hearer dyad. Change arises in the negotiation of meaning in discourse: context changes with every new turn when communication proceeds in time and becomes conventionalized through repetition across generations. Text is also seen as reflecting the unfolding discourse, and the interpretation of the meaning of elements in a given stretch of text depends on the preceding stretches and reveals itself in the following text. In diachronic studies, the focus of attention has shifted from stable to dynamic features and from fixed categories and inherent word meanings to discursive forms and negotiated meanings, in turn shifting the focus of analysis from text and the text producer to the interaction between communication partners. Innovation occurs when speakers use forms in new contexts and when hearers draw pragmatic inferences [46].

Therefore, it is no surprise that under this view, change is motivated by how language is processed, interpreted, and used in discourse. In what follows, through the discussion of short case studies, we explore four triggers for language change.

3.1. Pragmatic Inferences: The Case of ‘As Long As’

It is a natural property of communication that speakers routinely convey more than they explicitly encode, knowingly or not, therefore offering hearers the space to pragmatically enrich the meaning of a given element in that particular context (these are the so-called ‘invited inferences’ [46]). If these inferences occur frequently enough and expand their usage to different contexts, the enriched meaning becomes dominant, then conventionalized, and eventually encoded as the semantic or grammatical meaning of that element. In this process, both speaker and hearer have equally active roles in this rhetorical strategizing act [46] (p. 5).

Diachronically, the meaning expansion (or shift) starts with an expression having a meaning A semantically encoded. In given contexts (known as ‘bridging contexts’ [47,48]), an opportunity arises for that expression to take on meaning B, resulting from an enrichment brought about by pragmatic inferences. Meaning B is repeatedly inferred, in broadening contexts, so that, in time, it becomes predominant in generalized contexts, and it is eventually conventionalized into the new (or additional, in the case of polysemies) encoded meaning of the expression.

The development of English as/so long as can be seen as a case in point. The starting point was presumably the spatial coded meaning that the expression already had in earlier Germanic, but in both Old and Middle English, the spatial and temporal meanings coexisted. The examples in (1) and (2) (both from ref. [46] (p. 36), dating between 850 and 950 AD) show the basic spatial meaning and its extension from the temporal into the conditional meaning: in (1), the spatial meaning is the only possible interpretation, while in (2) there is an opening for an invited inference: there is a concurrence of the temporal extension of administering the medicine with the need for the medicine, inviting the inference of the enriched conditional meaning ‘provided that’, ‘squeeze the medication on the eye for the length of time that he needs it/until he no longer needs it’.

(1)

a

het

Ælfred

cyng

timbran

lang

scipu

ongen

ða

then

ordered

Alfred

king

to-build

long

ships

against

those

æscas;

ða

wæron

fulneah

tu

swa

lange

swa

þa

oðru.

warships

they

were

nearly

twice

as

long

as

the

others

‘then King Alfred ordered long ships to be built to battle the warships; they were almost twice as long as the other ships.’

(2)	wring	þurh	linenne	clað	on	þæt	eage	swa	lange	swa	him
	wring	through	linen	cloth	on	that	eye	as	long	as	him
	ðearf	sy
	need	be
	‘squeeze (the medication) through a linen cloth onto the eye as long as he needs.’

In both Old and Middle English texts as/so long as primarily cooccurs with verbs that denote temporary events, that is, existence (such as be) and living; Traugott and Dasher [46] (p. 37) note that ‘the conditional reading is available if the temporal clause refers to the future, or is generic, but never seems particularly salient’. This changes in Early Modern English: the invited inference carrying the conditional interpretation had been generalized to contexts in which the conditional meaning, although not the only one, was more salient. The contexts had extended to verbs denoting patterns of reasoning and cognition that were unlikely to change over time, but a change was not only possible but would be highly valued. This is exemplified in (3), from ref. [46] (p. 37), dating to 1614: the temporal interpretation is still available (‘for the length of time that they understand and speak as men’, i.e., ‘as long as they live’), but the conditional reading is more salient (‘provided that’).

(3)

They whose words doe most shew forth their wise vnderstanding, and whose lips doe vtter the purest knowledge, so as long as they vnderstand and speake as men, are they not faine sundry waies to excuse themselues? ‘(Even) the men whose words show their wise understanding and whose lips utter the purest knowledge, as long as they understand and speak like humans, are they not in many ways, forced to excuse themselves?’

By the mid-19th century, there are instances in which the temporal meaning is no longer available: in these cases, the enriched meaning has become conventionalized as a conditional polysemy of temporal as/so long as in (4):

(4)

“Would you tell me, please, which way I ought to go from here?” “That depends a good deal on where you want to get to,” said the Cat. “I don’t much care where – ” said Alice. “Then it doesn’t matter which way you go,” said the Cat. “ – so long as I get somewhere,” Alice added as an explanation. (from ref. [29] (p. 37))

Similar developmental paths are seen in French and Japanese, suggesting a more generalizable, recurrent path motivated by parallel pragmatic processes valid cross-linguistically.

3.2. Frequency: Gonna Construction

High frequency of use of a linguistic item is associated with several developments: strengthened mental representations, possibly due to a deeper entrenchment in memory and a faster (i.e., more automated) processing time; a certain propensity to phonetic reduction and reanalysis [41,42].

Grammar emerges from language use; it is a dynamic network of usage events stored in memory. Therefore, frequency of use is one of the most important forces shaping linguistic structure and language change [42]. Highly frequent items are accessed faster, and they tend to be stored as unanalyzable chunks.

Two types of frequencies need to be recognized: token and type frequency, each with different effects on language change. Token frequency refers to the frequency of a specific form, and it leads to entrenchment and faster processing. Highly frequent tokens are more prone to undergo grammaticalization. Type frequency refers to the frequency of a pattern or construction, and it leads to productivity. Highly frequent types are more likely to attract new members and to enable the extension of their pattern [42]. Bybee argues that frequency effects are universal and that cross-linguistic regularities arise because of shared cognitive constraints among speakers: faced with similar communicative pressures and needs, speakers produce similar outcomes.

To demonstrate that frequency is a crucial factor in language change, Bybee [42] presents a classic study, the grammaticalization of the English construction be going to into the intention/future marker gonna. It is a prototypical example of how token frequency, chunking, and reduction interact over time. The phonetic reduction undergone by be going to into gonna, in which the allative or infinitive marker is fused with the preceding -ing form, points to a concomitant loss of analyzability; this is supported by the contrast between the minimal pairs She’s going to apply for it and She’s going to the shops: when to is a preposition, no reduction is possible (*She’s gonna the shops). This also supports the idea that grammaticalization of lexical items has scope over constructions rather than over individual lexical items, and that it creates new constructions [49]: going to only grammaticalizes when it is followed by a verb (e.g., apply for it), but not when it is followed by a location (e.g., the shops).

The original sequence was a more general-purpose clause in which physical movement was necessary to perform an action: Alfred is going to York to see the Minster. This created the invited inference that the intention would lead to action, therefore suggesting the shift from intention to futurity. At this second stage, no physical movement was necessary, and the construction was extended to include non-movement verbs: go lost its original lexical meaning, becoming a member of a distinct construction [42] (p. 106). The shift was triggered by the frequency with which the original purpose construction going to, expressing motion towards a spatial goal, was occurring with infinitival complements denoting actions rather than locations. From having a fully compositional meaning of motion in order to carry out an action, the construction began to be interpreted as ‘movement towards X implies X being intended and likely to happen’. This was an invited inference in a bridging context, hence still pragmatic and not yet encoded semantically. Bybee stresses that these contexts were frequent, predictable and highly similar across tokens: as grammaticalization was taking place, there was a sharp rise in token frequency of going to in intention/future contexts. This high token frequency led to faster processing and a reduced need for full phonetic articulation. Furthermore, going to came to be stored as a chunk, from [going + to] to [going-to], in which internal word boundaries weakened, and the sequence started to behave differently from other going to sequences that were followed by a goal location. This, Bybee [42] argues, was driven by frequency; it applies only to the high-frequency future construction and does not apply to low-frequency go + to combinations. The same process can also be invoked for contractions of want to, have to and supposed to.

3.3. (Inter)Subjectification: Only

Diachronic studies have identified a tendency for meanings to shift from expressing objective properties to subjective perspectives related to the speaker’s attitude, and from expressing subjective views to an intersubjective awareness of the hearer’s perspective (starting from ref. [50]). The processes of subjectification and intersubjectification, therefore, broadly refer to a move towards meaning/functions more deeply rooted in the speaker’s attitude towards the proposition and in their relation to the addressee’s beliefs, respectively. In intersubjectivity, the attention to the ‘self’ of the addressee is meant both in an epistemic sense (paying attention to their presumed attitudes to the content of what is said), and in a more social sense (paying attention to their ‘face’ or ‘image needs’ associated with social stance and identity).

An example of subjectification is the shift in the meaning of English only. In present-day English, only typically functions as an adjective (e.g., the only reason) or a focusing adverb (e.g., only once), but it can also be a clause-linking conjunction which can either have an adversative meaning (similar to but, e.g., Do what you like, only don’t miss the train, from ref. [51], in ref. [52] (p. 97)) or an exceptive focusing function in which only behaves as a pragmatic marker (similar to except that, e.g., The flowers are lovely; only, they have no scent, from ref. [53]). Only developed from the numeral one into an adverb meaning ‘uniquely’, and shifted further into an exclusive focus marker, an element that, taking scope over the following element, picks that element out of a set and establishes that the state-of-affairs denoted by the verb applies to that element exclusively. For an element to become a scalar focus particle, it is necessary for the speaker to rank the selected element relatively to others in a set: this is, of course, the outcome of an evaluation on the part of the speaker, an expression of subjectivity. The process through time is an instance of subjectification, since only acquires speaker-based meanings (e.g., belief, stance). From meaning ‘uniquely’, which implies oneness, only develops to express the speaker’s evaluation in terms of excluding all other possibilities from a given set. This initial development is summarized in (5):

(5)	numeral one > polysemous adj./adv. only > exclusive focusing only (from ref. [53] (p. 26)).

The further development of only into a pragmatic marker with an exceptive meaning is also analyzed as an increase in subjectivity: ‘[i]n its conjunctive sense, only likewise rejects a presupposition, but in this case not the presupposition set of the focus item, but the presupposition of the entire preceding clause’ ([53] (p. 28)). This use could already be found in the 17th century, see in (6):

(6)

I do fully see the evidence of all that which you have said, and therefore I must needs be perswaded of it. I do heartily thanke God for it, and will endevor myselfe to put it in practise continually. Only here is the difficulty, how a Schoolemaster may do this, to teach his Scholler so to proceede with understanding, and how to give a reason of every matter which they learne, to make use of all their learning. (from ref. [53] (p. 24))

Here, only is used to block any possible pragmatic inference that, because of Brinsley’s hearty thanks and promise to continually engage with it, his endeavor will be easy: only, therefore, expresses the speaker/writer’s attitude and evaluation of the connection between the preceding clauses and the following one. Brinton [53] (p. 28) observes that many of the early examples in which only has this meaning are comments on the communicative situation and about the progress of the argument.

3.4. Discourse Management and Turn Taking: The Obligatorification of French Subject Pronouns

One of the core tenets of the functionalist perspective is that grammar evolves from language use to serve interactional needs; one of the most fundamental of these needs is the organization of conversational turns. Turn-taking is the manner in which orderly conversation normally takes place; it is a system involving both linguistic (e.g., intonation, specific structures of lexical elements) and non-linguistic cues (e.g., gestures, gaze) that determine how speakers alternate, share the communicative ground, and transition smoothly from one turn to another, minimizing overlaps and gaps [54].

Diachronic studies have observed how turn-taking pressures motivate the emergence and grammaticalization of linguistic elements that can signal turn boundaries, manage floor-holding, coordinate speaker–hearer alignment, and facilitate rapid interpretation. Language change is hence partly driven by interactional efficiency.

An interesting case study of how turn-taking management led to a prototypical example of grammaticalization is the emergence of obligatory subject pronouns in French [55]. Detges explores the obligatorification of subject pronouns in the transition from Old French to Modern French, contrasting it with other Romance languages. The core argument proposes that this significant typological shift was driven primarily by pragmatic factors rather than grammatical necessity or Germanic influence. Specifically, Detges details how the original function of Old French subject pronouns was to mark a referential contrast. Their frequent ‘rhetorical over-use’ in contexts such as self-topicalization and turn-taking led to a dramatic frequency increase and the subsequent devaluation of their pragmatic effect, a process termed rhetorical devaluation. This devaluation, driven by the overuse of optional first- (followed by second-) person pronouns as a rhetorical strategy to gain a turn in the conversation, seen in ‘strong’ speech acts, led to their eventual reanalysis as unmarked, subject markers with a purely grammatical function.

In Old French, subject pronouns (SP) were not used in the same way as in Modern French. As (7) shows, SP did not have an anaphorical function: they were used to mark a referential contrast (i.e., they were much rarer), unlike their counterpart in Modern French (8):

(7)

Nadabi,

le

fiz

Jeroboam,

regnad

sur

Israel

el

secund

Nadabi

the

son

of-Jeroboam

reigned

over

Israel

the

second

an

Asá,

le

rei

de

Judá,

é

dous

ans

øi

regnád.

Málement

with

Asá

the

king

of

Judah

and

two

years

reigned

badly

øi

uverad

vers

nostre

Seignur

é

øi

sewíd

les

behaved

towards

our

Lord

and

followed

the

males

traces

sun

perej

é

le

pechied

par

unt

ilj

fist

bad

steps

of-his

father

and

the

sin

through

which

he

made

pecchier

cez

de

Israel.

sin

those

of

Israel

‘Nadab, the son of Jeroboam, reigned over Israel as second king with Asá, king of Judah, for two years. He behaved badly towards our Lord and followed the bad example of his father and the sin through which the latter (i.e., his father) had made those of Israel sin’. (from ref. [56] (p. 370) in ref. [55] (p. 1)).

(8)	Nadabi le fils de Jeroboam régna sur Israel en second roi avec Asa, le roi de Juda, et ili régna deux ans. Ili agit mal envers notre Seigneur et øi suivit les mauvaises traces de son pèrej et le péché par lequel celui-cij avait induit dans le péché ceux d’Israel.

The process that saw SPs become obligatory in French, which started around the end of the 12th century and affected first-person pronouns first, involved a considerable frequency increase of SPs and a loss of their original contrastive effect. The increased frequency of SPs was first witnessed in main clauses: this was a distinct phenomenon from the already existing high frequency of SPs in subordinate clauses, which was linked to a technique of subordination. It was unlikely that the high frequency of SPs in subordinate clauses could have in any way triggered a frequency increase in main clauses, since main clauses are much more frequent than subordinate clauses in oral communication, and the direction of influence is usually from main to subordinate clauses and not vice versa. Furthermore, the high frequency of SPs in subordinate clauses followed a regularity that had remained stable since the 11th century. Therefore, it is reasonable to assume that the sudden increase in frequency of SPs in main clauses from the end of the 12th century onwards had no direct connection with their increase in embedded clauses [55] (p. 3)

The obligatorification of SPs is a polygenetic phenomenon: a type of process that has affected many languages in different parts of the world at different times. Polygenetic change cannot be explained either by grammatical necessity or by specific historical circumstances. Therefore, Detges proposes that the phenomenon in French is the result of certain rhetorical strategies motivated by the speakers’ desire to assure themselves of immediate communicative advantages. Support for such an analysis comes from cross-linguistic evidence from other Romance languages that have, at some point, seen a change in the use of their subject pronouns. The process leading to obligatorification often starts with the first person (cf. also Brazilian Portuguese, in which a similar development has taken place). This shift can also be currently observed underway in null-pronoun languages like Spanish, where SPs show an increased frequency in specific contexts, particularly with verbs of opinion (e.g., pensar ‘to think’, creer ‘to believe’). In the context of these verbs, the SPs are no longer exclusively used with their original contrastive value: they also perform the additional function of signaling that the speaker is self-promoting because they have something of value to contribute to the conversation. Formulas of self-topicalization and stressed SPs are highly efficient routines used in conversations to secure the right to speak (turn-taking), often realized under time pressure. Marking a contrast (even rhetorically) announces unexpected information, which justifies the incipient speaker’s attempt to take the floor (‘the more the incipient speaker’s contribution appears to contain unexpected information, the more his attempt to take over the right to speak seems justified’, ref. [55] (p. 8).

In Old French, a similar large-scale process of rhetorical devaluation occurred, specifically linked to ‘strong’ speech acts, such as directives, commissives and in collocation with verbs of saying. The stressed first-person pronoun was employed to underpin the speaker’s authorship or communicative responsibility for the proposition, as in (9):

(9)	Pur	ço	le	juz	jo	a	pendre	e	a	murir.
	for	this	him	judge	I	to	hang	and	to	die
	‘For this, I condemn him to hang and die’ (from ref. [55] (p. 12)).

The presence of the SP jo guarantees the illocutionary force of this ‘strong’ declarative speech act that would not have been possible without its overt realization. This strategy makes SP-marked speech acts appear more serious and efficient than null-subject speech acts. The frequent use of SP to mark speech acts as ‘strong’ results in rhetorical ‘overuse’, causing their frequency to increase dramatically while their contrastive value progressively weakens, the core mechanism of rhetorical devaluation. Evidence for this shift is observed in texts reflecting spoken language around the early 13th century, such as the Conquest of Constantinople (10), from ref. [55] (p. 14):

(10)

Sire,

nos

somes

venu

a

toi…

Et

saches

tu

que

il

Sire

we

we-are

come

to

you

and

know

you

that

they

te

reprovent…

Vos

lor

avez

juré,

vos

et

vostre

to-you

represent…

you

to-them

you-have

sworn

you

and

your

peres

la

convenance

a tenir…

Vos

ne

lor

avez

mie

father

the

promise

to keep

you

not

to-them

you-have

not

si

bien

tenu

com

vos

deussiez.

so

well

kept

as

you

should

‘Sire, we have come to you… and you know that they represent to you… You have sworn to them, you and your father, to keep the promise… You have not kept to it as well as you should have’.

This passage is an extremely impolite, bold challenge (‘strong’ speech act) characterized by a high number of SPs (nos, tu, il, vos). The frequent use of SP is a major stylistic device giving the text a direct and forceful character, yet the stylistic value of each occurrence is lost in the sheer mass, demonstrating rhetorical overuse.

The rhetorical devaluation of SPs led to a subsequent grammatical development, known as ‘pronominal reprise constructions’: since the original SPs lost their contrastive function, a new, ‘heavy’ procedure for self-topicalization was needed. This function was adopted by the pronominal reprise construction (e.g., modern French moi je ‘As for me, I’). The earliest examples of SP reprises (dating to 1617) appear in contexts already associated with the change: verbs of thinking and ‘strong’ speech acts in (11).

(11)	Et	jou	je	cuit	si	bien	feriés…
	and	I	I	think	if	well	you-did
	‘And as for me, I think that if you did well…’ (from ref. [55] (p. 17)).

In modern French, tonic moi is typically used, just like the stressed SP in Spanish or Italian, in contexts of self-topicalization and of turn-taking. Interestingly, in these constructions, moi no longer obligatorily marks a referential contrast.

The obligatorification of SPs in French was triggered by pragmatic factors, specifically, rhetorical strategies motivated by speakers seeking immediate communicative advantages in turn-taking. This is a powerful stimulus to employ the SP more often than necessary, including in situations where this is no longer justified by the real relevance of the speech act in question. The increased frequency led to an inflationary effect, so that the original contrast expressed by SPs was lost, and they became unstressed and obligatory pronouns.

Discourse management provides systematic motivation for recurrent types of change, especially the rise of discourse markers, particles, and interaction-oriented constructions. It is a key motivation for language change because conversational interaction is rapid, with minimal gaps within and between turns, competitive (speakers negotiate the floor), and incremental (meaning is built in real time). These pressures leave diachronic traces in grammar. Linguistic elements are repeatedly recruited to manage turn initiation, continuation, and transition, and through frequent use at turn boundaries, they undergo semantic bleaching, positional fixation, and grammaticalization.

4. Conclusions

In this overview of two of the main theoretical approaches to diachronic language change and the exploration of usage-based motivations for such change, we have focused on triggers that originate in language use and are, crucially, not dictated by autonomous, internal grammatical rules. At the core of language change are discourse participants and their individual subjectivities, the dynamics of managing and negotiating meaning and the common ground, discourse management, cognitive processes and functions, as well as frequency of use. Because these are features that pertain to communication in general, cross-linguistic recurring patterns are common, giving rise to paths of change (or grammaticalization clines). The presentation of four short case studies has offered a taster of what qualifies as triggers from the usage-based perspective, how these factors can be employed to explain specific changes and, in turn, how they represent the reasons why language changes.