5.3.1 From Coordination to Coordinating Connectives

Mithun (Reference Mithun, Haiman and Thompson1988: 331) sees coordination to be “the most basic and universal” syntactic construction. At the same time, she observes “surprising variety in the types of coordinating constructions exhibited by different languages, and even in the degree to which coordination is grammaticized at all” (p. 331). In other words, not all languages have developed grammaticalized forms of coordination, and when they have, they come with a high degree of both cross-linguistic and intra-linguistic variation, some paths being widely shared, others being specific to some languages or language families (see also Haspelmath, Reference Haspelmath and Shopen2007; Mauri & Van der Auwera, Reference Mauri and van der Auwera2012). Giacalone Ramat and Mauri (Reference Giacalone Ramat, Mauri, Heine and Heiko2011: 654–55) focus on the three main types of coordination relations recognized in the literature: conjunction (‘and’), disjunction (‘or’), and adversativity (‘but’). Diachronically, the three groups of coordinating connectives show partially distinct paths of evolution.

According to Giacalone Ramat and Mauri (Reference Giacalone Ramat, Mauri, Heine and Heiko2011: 661), conjunctive ‘and’-connectives often find their source in spatio-temporal adverbs and prepositions “typically indicating a linear succession in time ‘before, after’ or a linear organization in space ‘in front, beside’.” Thus, through a metaphorical process concrete temporal and spatial notions evolve into more abstract, logical notions. While this metaphorical process is a fairly standard process of meaning extension (Xu, Malt & Srinivasan, Reference Xu, Malt and Srinivasan2017), Traugott (Reference Traugott1986: 137) already noted that the precise details of exactly which spatial and temporal meanings are selected to express which logical connectives deserve more attention. This grammaticalization path from (spatial or temporal) linearization expression to coordinator can proceed at lower or higher levels of syntax.

Conjunction between nominal phrases typically starts in focal additive particles (‘also, too’) or comitative markers (‘with’) (see Bril (Reference Bril2011) for an illustration of this phenomenon in contemporary Austronesian languages, and Mithun (Reference Mithun, Haiman and Thompson1988: 339–43) for a detailed illustration of the diachronic processes involved in this grammaticalization path in Sarcee (Athapascan), Kwa languages (West Africa) and Cayuga (Northern Iroquoian)). Beyond the nominal phrase, paragraph-linking devices, discourse markers, verbs with dislocative meaning (‘go, bring’) or anaphoric pronominal roots may “grammaticalize at the higher levels as connectives between clauses” (Giacalone Ramat & Mauri, Reference Giacalone Ramat, Mauri, Heine and Heiko2011: 661). We again refer to Mithun (Reference Mithun, Haiman and Thompson1988) for detailed descriptions in typologically diverse languages. Of importance here is her observation that for many of these emerging forms, “[t]heir primary function is to provide a semantic or pragmatic link to previous discourse, not to specify a syntactic one” (Mithun, Reference Mithun, Haiman and Thompson1988: 346). This explains that in many languages discourse adverbials and syntactic conjunctions are often difficult to tell apart because of this common functionality. Finally, while there is not a single path leading to coordinating conjunctions, Mithun (Reference Mithun, Haiman and Thompson1988) points out two additional noteworthy observations: (i) nowadays differences in placement of conjunctions (pre-posed or post-posed) can stem from the different diachronic origins of the markers themselves (p. 351), (ii) coordinating conjunctions grammaticalized fairly recently across languages over the world, with some still in progress, and other clearly relating to borrowing in language contact situations (pp. 351–52).

As for the diachronic sources of disjunctive connectives, Giacalone Ramat and Mauri (Reference Giacalone Ramat, Mauri, Heine and Heiko2011) present a list comprising distal elements, interrogative particles, free choice verbs, dubitative particles, negative particles and denied conditionals. According to the authors, these elements may acquire a disjunctive meaning

In other words, disjunctive connectives emerge in a context where an alternative is presented between two possible choices for the hearer.

For adversative connectives, Giacalone Ramat and Mauri first observe that they come with a higher degree of intra-linguistic variation. In other words, in most languages the stock of adversative connectives is more important than that of conjunctive or disjunctive markers. They illustrate this with French “which only shows et for conjunction and ou (ou bien) for disjunction, but a number of different connectives for contrast relations, e.g. toutefois, mais, par contre, alors que, pourtant” (p. 658). Furthermore, they observe that adversative connectives are more prone to renewal over time and that they are more easily borrowed than conjunctive and disjunctive ones, which may be related to the fact that adversative meaning is strongly intersubjective, with speakers “constantly in search of new and expressive ways of conveying contrast, determining a high synchronic intra-linguistic variation and a quicker renewal” (p. 659). The diachronic sources for adversative connectives are expressions of spatial distance or opposition, expressions of temporal overlap or continuity, causals, comparatives and emphatic expressions. In the case of spatial and temporal source expressions, adversative meaning emerges from the contrast or opposition of two simultaneous states of affairs that may appear as surprising or unexpected. In her study of Italian adversative connectives anzi (‘rather’, ‘on the contrary’) and invece (‘instead’), which both find their source in the spatio-temporal domain of posteriority and anteriority, Musi (Reference Musi2016) shows that the emergent notion of oppositive and counterexpectative contrast encodes the speaker’s inferred link between (staged) simultaneity and contrast. Thus, “the contrastive meaning emerges as an invited inference in so-called bridging contexts, where occurrences present a set of recurrent semantic and syntactic features that favor the reanalysis” (p. 27) (see also Section 5.3.1). Causal meanings may also give rise to adversative meanings through denial of expectation, while comparative and emphatic expressions put a contrast between two elements of which one is more in focus (see also Section 5.3.2).

Tracing the evolution of any (coordinating) connective involves analysing its context of occurrence through time, where changing contexts reveal changes of function and meaning. Typically, the grammaticalization of connectives starts in contexts that are semantically and syntactically ambiguous between the original (concrete, propositional) meaning and the emerging (abstract) connective function. In such ‘critical’ contexts (Diewald, Reference Diewald, Wischer and Diewald2002) or ‘bridging’ contexts (Heine, Reference Heine, Wischer and Diewald2002), speakers will activate pragmatic inferences concerning the presence and/or type of interclausal relation (see also Musi, Reference Musi2016).

Concerning Lehmann’s (Reference Lehmann1995) grammaticalization criteria (see Section 5.2), Giacalone Ramat and Mauri observe that phonological reduction and coalescence are very often attested in the early stages of the diachronic evolution of coordinating connectives, for example, Old English be utan > Engl. but. Yet, as already mentioned, other grammaticalization features do not readily apply. The authors mention obligatorification, paradigmaticization, and scope reduction, that seem incompatible with the function and morphosyntactic properties of interclausal connectives as such (p. 664). Scope reduction does not apply because coordinating connectives typically link two clauses, taking wide scope over the two segments. Obligatorification, in the sense of obligatory presence of a linguistic element (e.g. inflectional morphemes), seems inconsistent with coordinating connectives because in many contexts they may be left out (cf. Chapter 1), that is, speakers have the choice between the syndetic and asyndetic options, even in languages with a well-developed connective system. Finally, paradigmaticization is a problematic feature because, while they form a closed set, coordinating connectives do not really form a paradigm in the sense of number or gender inflection paradigms. It follows that these three features cannot be taken as indicators of the degree of grammaticalization of interclausal connectives.

5.3.2 From Subordination to Subordinating Connectives

It is cross-linguistically widely acknowledged that subordinate clauses arise from simple, main clause structures, whereby propositional lexical material grammaticalizes into markers of clause subordination (Hopper & Traugott, Reference Hopper and Traugott2003: 184–90; Givón, Reference Givón2009). Grammatically, subordinate clauses come in three main types: (i) a complement clause functions as either subject or object of the main clause; (ii) a relative clause modifies a noun, similarly to an adjective does; and (iii) an adverbial clause occupies an adjunct position modifying the main clause. These three types of subordinate clauses seem to be cross-linguistically and typologically confirmed (Gast & Diessel, Reference Gast and Diessel2012a). Ohori (Reference Ohori, Heine and Narrog2011: 636) further observes that adverbial subordination is less homogeneous than the other two, both structurally and semantically, and that the border between adverbial subordination and coordination may be viewed as a continuum. Structurally, Ohori summarizes the general tendency that dominates the grammaticalization of subordination as one of “less to more clause integration”, that is, an evolution from syntactically independent, juxtaposed, simple clauses to a syntactically integrated complex clause (see, e.g. Lehmann, Reference Lehmann, Haiman and Thompson1988; Muller, Reference Muller1996, but see Combettes, Reference Combettes and Debaisieux2013 for a more nuanced view). Here, we will focus on adverbial subordination mainly when reviewing the emergence of subordinating conjunctions used as discourse connectives.

Regarding the historical development of adverbial subordinators, Kortmann (Reference Kortmann1996: chapter 10) presents an overview study based on exhaustive inventories in Old English (OE: 10th–11th c.), Middle English (ME: 14th–15th c.), Early Modern English (EModE: 16th–17th c.) and Present-Day English (PDE: 20th c.). The study is presented as illustrative of the diachronic path of subordinating connectives in European languages. His focus is on the evolution of the inventories, sketching the major morphological and semantic changes through time. Quantitatively, the “stock” of available subordinators rises between the OE and the EModE periods and then drops in PDE, with markers emerging, changing and disappearing over time. Regarding the markers that “survived”, Kortmann finds that most subordinators go back far in time (OE and mainly ME) and that more recent innovations (last 500 years) are less numerous. In particular, many highly frequent PDE adverbial subordinators find their source in the OE ancestors, among them markers such as after, as, as long as, as soon as, if, since, so that, though and while. This being said, the ME period appears to be even more important for today’s subordinator inventory, with a massive increase in the number of markers, including borrowings (from French) and functional innovations of existing markers, of which more than 50 percent made it to the present time. Among these, we find again highly frequent markers, such as although, because, before, until, when, where. Kortmann links the emergence of markers in this period to the development of a written standard language for which “the development of an inventory of clause-connecting devices specifying interpropositional relations, such as adverbial subordinators, was indispensable” (p. 303).

From a morphological point of view, the evolution of subordinators is characterized by the following features. First, the preferred categorial sources of adverbial subordinators in OE are adverbs and demonstratives. This is a feature OE shares with other old Germanic dialects for the formation of adverbial subordinators. From ME on, the inventory sees a considerable increase in prepositions, nouns, verbs, and complementizers as source categories. The increasing importance of verbs as a source of adverbial subordinators follows from the reanalysis of participles with or without complementizer that. Examples are: considering (that), notwithstanding, in case (that), provided (that), seeing (that), etc. “In addition, Middle English and Early Modern English employ several subordinators incorporating the base form of the verb be (e.g., EModE be it so (that, as), (if) so be (that, as), if case be (that), albeit (that), howbeit (that))” (p. 307). Another “new” source category for subordinators in ME are nouns (often as part of a prepositional phrase) following from direct and indirect borrowings and calques from French. Examples are by (the) cause (that), to that forward (that), to the end (that), in respect (that), not all of which survived. Kortmann further observes that “the inflationary use of phrasal subordinators involving the complementizer that in Middle English times, should not be viewed entirely independently of the formation pattern in Old French producing phrasal subordinators with que as final element” (p. 309). Furthermore, the often obligatory use of complementizer that in OE and ME became optional over time and has now mostly disappeared giving rise to “one-word subordinators”, such as after, because, before, if, in case, since, though, till, unless, until, when, where, while, whilst, as though (p. 309).

The second morphological feature that we will present here concerns the syntactic polyfunctionality of adverbial subordinators. It is worthwhile noting that “Old English did not have a single primary adverbial subordinator, i.e. no one-word item serving exclusively as adverbial subordinator” (p. 310). Such “special-purpose” subordinators did not develop before the ME period and have remained stable with around 14 percent of the inventory of adverbial subordinators until today. Compared to the other West-Germanic languages, this proportion is remarkably low and in contrast with Present-Day Dutch (approx. 38 percent) and German (approx. 40 percent) (p. 311).

Concerning the semantic evolution of adverbial subordinators, there is a clear general tendency towards decreasing polyfunctionality (or increase in semantic precision) throughout the history of English, in the sense that the proportion of polyfunctional subordinators compared to the monofunctional ones decreases, especially (again) between the OE and ME periods. More precisely, comparing the proportion of mono- and polyfunctional subordinators for the whole subordinator inventory, Kortmann finds that “the proportions of polyfunctional items decrease and those of the monofunctional items increase from Old to Early Modern English” (p. 314), especially for the one-word and monomorphemic adverbial subordinators. Yet, this is not to say that polyfunctionality disappears in the individual markers. To the contrary, “there are many instances of polyfunctional subordinators whose meaning range has not become smaller or has, indeed, grown larger over the centuries (Kortmann, Reference Kortmann1996: 315). To illustrate this, Kortmann compares the semantic evolution of so and as, which he describes as the markers “which have undergone the most dramatic changes in their meaning range in the history of English” (p. 315). Whereas both markers have modality as their core meaning in OE, from ME on their semantic paths diverge: the (subordinator) meaning range of so decreases, while that of as increases. Concretely, so lost all of the meanings it had as an adverbial subordinator in OE, among them ‘simultaneity’, ‘condition’, ‘similarity’, ‘place’, and others. In ME, it gained ‘result’ and ‘purpose’ meanings, which are still the only subordinating meanings available today, as illustrated in (5–6):

1. (5)

   It was dark, so I couldn’t see what was happening.

1. (6)

   I packed him a little food so he wouldn’t get hungry.

   [from Kortmann, Reference Kortmann1996: 317]

Other PDE interclausal uses are not subordinating. In this area, we do indeed witness a decrease in semantic polyfunctionality. For a counterexample, namely one of polyfunctionality increase in this area, we refer the reader to Kortmann (Reference Kortmann1996: 317–18) for “the success story” of as in English, “i.e., one that is characterized by an enormous expansion of the range of interclausal relations it came to express in later periods than Old English.”

A second general finding concerning adverbial subordinators is that semantic changes involving ‘cause’, ‘condition’, ‘concession’ (or “CCC relations”, also including ‘purpose’ and ‘result’) emerge later, and generally derive from locative, temporal or modal senses. In Kortmann’s dataset “there is not a single subordinator which expresses some CCC relation before it (additionally or alternatively) comes to be used as a locative, temporal, or modal marker” (p. 319). These include the development of causal markers from temporal markers of anteriority or simultaneity, as observable for English since or now (that), but also German weil. In the following Section, we will see that this development seems to be attested cross-linguistically. Noteworthy is also the development of contrastive-concessive whereas from a locative marker, and the observation that ‘concession’ is generally the last meaning to emerge (see Section 5.3.1, for a case study in French). About such “CCC markers”, Kortmann finally observes that their proportion steadily expands in the inventory of adverbial subordinators, at the expense of the markers expressing temporal relations.

5.3.3 Reconstructing Semantic Paths towards Relational Meanings

Before turning to a number of illustrative case studies, we would like to emphasize the fact that the emergence of subordinating and coordinating conjunctions shares many features. In particular, there is striking evidence across typologically different language families that discourse connectives follow a semantic path that starts off with concrete propositional meanings that give rise to abstract relational meanings through metaphorical extension and pragmatic strengthening. To illustrate this point, we will take as our starting point the “Target-Source List” in Heine and Kuteva’s (Reference Heine and Kuteva2002) famous World Lexicon of Grammaticalization. This list gives an overview of present-day (semantic and syntactic) concepts and relates them to possible source forms in over 500 (documented) languages across the world. With an eye on uncovering the semantic paths leading towards nowadays connective use, we retrieved all target concepts related to the diachrony of connectives. We restricted our retrieval to relational meanings also found in the main taxonomies of discourse relations (Chapter 3). These are (in alphabetical order): adversative, cause, concessive, conditional, purpose, and temporal.

When it comes to the (lexical) source of these relational meanings, Heine and Kuteva distinguish the following:

• adversative < temporal

• cause < back, here, locative, matter, purpose, say, since (temporal), temporal

• concessive < conditional, temporal

• conditional < copula, s-question, say, temporal

• purpose < allative, benefactive, come to, complementizer, give, go to, matter, say

• temporal < allative, comitative, hour, in (spatial), interior, locative, time

For the sake of illustration, we will review here the evolution towards causal relational meanings only (marked in bold in the list above), that is, those source items that have given rise to causal connectives, very often expressed by means of coordinating and subordinating conjunctions.

The lexical source back (body part) is a “successful” source for several grammatical forms. According to Heine and Kuteva (Reference Heine and Kuteva2002: 46–50), this process is an instance of a more general process whereby body parts are grammaticalized to spatial concepts and markers, which again are used to express temporal concepts, but also markers for more abstract grammatical relations, such as cause. A striking illustration comes from Wolof ginnaaw, which synchronically expresses three different grammatical categories: as a noun with the meaning ‘back’ (body part), as a preposition with the meaning ‘behind’ and as a subordinating conjunction meaning (causal) ‘since’. Robert (Reference Robert1997) proposes a unitary semantic analysis according to which ginnaaw refers to a location behind a given structured space marked by a landmark. “When ginnaaw is in nominal function, no other term in the utterance plays the role of the landmark; the morpheme has an extra-linguistic reference (…) i.e. the human body. In prepositional use, the landmark is the noun governed by ginnaaw. In subordinating use, the landmark is the clause introduced by ginnaaw” (Robert, Reference Robert1997: 125). Thus, the causal subordinating use proceeds from the conceptualization of causality in spatial terms, that is, one event behind the other, but also from the metaphorical structuring of “discourse as landscape”, in which argumentative sequences are located behind and following one another, thus creating argumentative inferences (Robert, Reference Robert1997: 124).

Other “spatial concepts” that give rise to causal conjunctions are here, and more generally locatives. For here, Heine and Kuteva (Reference Heine and Kuteva2002: 171) cite examples from Lingala and Albanian, where the locative adverb ‘here’ (respectively áwa and ke) comes to be used as a causal conjunction corresponding to ‘since’, ‘because’. As for purpose, the authors observe that it is often part of “the same polysemy set” as cause (Heine & Kuteva, Reference Heine and Kuteva2002: 247). While there is no conclusive historical evidence to support their hypothesis, the authors nevertheless argue on the basis of data available to them that diachronically purpose precedes cause in time. Causal subordinating conjunctions may also result from a path that takes say as its source item. According to the authors, say-verbs may develop into markers of purpose, cause, and temporal adverbial clauses at later stages of grammaticalization (Heine & Kuteva, Reference Heine and Kuteva2002: 261–70). They see this as “an instance of a process whereby process verbs, on account of some salient semantic property, give rise to grammatical markers used for clause combining” (p. 269). Examples are given for languages as diverse as Baka or Lezgian. Finally, temporal expressions are an important source for causal conjunctions. “This appears to be an instance of a widespread process whereby spatial and temporal markers are grammaticalized in specific contexts to markers of ‘logical’ grammatical relations, such as adversative, causal, concern, concessive, and conditional relations” (Heine & Kuteva, Reference Heine and Kuteva2002: 275–76). A well-known example is Latin posteaquam ‘after’, ‘ever since’ > French causal subordinator puisque ‘since’. Other instances described by Traugott and König (Reference Traugott, König, Traugott and Heine1991: 194–97) include the semantic paths: Old High German dia wila so ‘so long as’ > German weil ‘because’, English temporal since > English causal subordinator since. They also mention the regular polysemy of conjunctions with temporal and causal meaning: French temporal subordinator quand ‘when’ also used with causal meaning ‘because’, Finnish temporal subordinator kun ‘when’, ‘as’ also used with causal meaning ‘since’, ‘because’, Estonian temporal parast ‘after’ also used as causal ‘because of’ or temporal kuna ‘while’ with causal meaning ‘as’, ‘since’, ‘because’, or Romanian din moment ce ‘from the moment’ or ‘because’, and so on.

5.4.1 From Temporal to Concessive in French

Marchello-Nizia (Reference Marchello-Nizia2009) studies the development of the French concessive connectives cependant (‘yet’), toutefois (‘however’) and pourtant (‘though’). She shows that these linguistic forms first change their syntactic category through a process of grammaticalization (from verbal and nominal expressions to adverbials), in absence of any concessive meaning. Then, through what she calls a process of pragmaticalization, these adverbials acquire a concessive meaning. Originally, none of these linguistic expressions were neither adverbs, nor connectives, and they did not express a concessive meaning. Rather, stepwise, they changed their grammatical category and their semantic meaning. To illustrate this, we will focus on her analysis of cependant (‘yet’) (see also Marchello-Nizia, Reference Marchello-Nizia2007).

The origin of cependant lies in the anaphoric demonstrative ce (‘this’) followed by the present participle of the verb pendre (‘to hang’). This corresponds to Diessel’s (Reference Diessel1999) observation mentioned above that many connectives find their source in the combination of a pronominal demonstrative and some other element expressing a semantic relation. Here, pendant (‘hanging’) is at the origin of the temporal (durative) meaning, that developed with its adjectival use in technical judicial vocabulary at the end of the thirteenth century, for example, le jugement pendant, meaning ‘while waiting for the judgment’. In the course of the 14th–15th c. durative (adjectival) pendant extends to new, non-judicial, contexts. During the fifteenth century, pendant starts occurring in prenominal position leading to a reanalysis and recategorization as a preposition (adjective > preposition). In parallel, from the end of the thirteenth century, pendant is used in the combination: en [preposition] ce [anaphoric demonstrative] pendant [verb], with durative meaning in judicial contexts. In the course of the fourteenth and fifteenth centuries, the expression ce pendant or cependant gains a clear temporal (concomitant) meaning (‘during that same time’) in non-judicial contexts as exemplified in example (7) from the fifteenth century.

1. (7)

   … il ne feroit que demy guet, c’est assavoir depuis la mynuyt jusques au matin seulement, et [que] si ce pendant il vouloit venir parler a elle, elle orroit voluntiers ses devises. (Cent nouvelles nouvelles, 1460, p. 389)

   ‘He would keep only half a period of watch, namely from midnight to morning only, and if ce pendant (‘during that time’) he wanted to come and talk to her, she would gladly listen to his words.’

   [from Marchello-Nizia, Reference Marchello-Nizia2009: 12]

When this temporal meaning occurs in concessive contexts, either explicitly marked by concessive connectives such as mais (‘but’) or in contexts where the temporal simultaneity receives meaning nuances of opposition or contrast, cependant is in a favourable position to add the concessive value to its temporal meaning. These are contexts referred to by Heine (Reference Heine, Wischer and Diewald2002) as “bridging contexts”, that is, contexts in which the source meaning is still available, but in which the target meaning is more likely to be inferred. Thus, at this stage, cependant has not yet acquired its new concessive meaning. According to Marchello-Nizia, this does not happen before the mid-sixteenth century, when the temporal meaning leaves room to the encoded concessive meaning and cependant is used independently as a concessive connective.

Summarizing the grammatical and semantic evolution of cependant, we witness a case of polygrammaticalization where pendant, on the one hand, evolves from verbal phrase to adjective to preposition, and on the other hand, from verbal phrase to adverb (combined with ce). Semantically, the evolution is from spatial (the verb pendre ‘to hang’ is originally spatial) to temporal to concessive logical meaning, in line with the semantic path identified by Heine and Kuteva (Reference Heine and Kuteva2002) (cf. Section 5.2.3). A final note is in place here concerning the present-day (restricted) polysemy of cependant, which until today may occur with its temporal meaning of simultaneity, but only in formal writing.

Strikingly similar semantic paths have been described for Italian anzi and invece (Musi, Reference Musi2016), the Romance markers Fr. toutefois, It. tuttavia, Sp. todavía and Port. todavia (Giacolone Ramat & Mauri, Reference Giacalone Ramat and Mauri2012), English nevertheless, nonetheless, notwithstanding, still, yet and counterparts in many languages (König, Reference König1985), English while (among others) (Traugott & König, Reference Traugott, König, Traugott and Heine1991), etc.

5.4.2 From Cause to Contrast: Italian Però

In their study, Giacalone Ramat and Mauri (Reference Giacalone Ramat, Mauri and Verhoeven2008) reconstruct the semantic evolution of Italian però from causal to adversative marker. They find this evolution “intriguing” because it involves a so-called functional reversal, from cause to counterexpectative contrast, or denial of an expected causal sequence (Giacalone Ramat & Mauri, Reference Giacalone Ramat, Mauri and Verhoeven2008: 304), rather than the more classical evolution from temporal to causal and/or concessive/adversative (see Sections 5.3.2 and 5.4.1; Heine & Kuteva, Reference Heine and Kuteva2002: 291). A present-day example of the adversative però is (8). It is a use that is fairly recent in Italian with first unambiguous attestations in the seventeenth century.

1. (8)

   Mario gioca bene però perde in continuazione.

   ‘Mario plays well però (but) always loses.’

   [from Giacalone Ramat & Mauri, Reference Giacalone Ramat, Mauri and Verhoeven2008: 303]

Reconstructing the history of però, Giacalone Ramat and Mauri indicate that the connective is attested from its earliest occurrences in the twelfth century with a resultative function, thus directly continuing the meaning of its late Latin antecedent per hoc. In Old Italian, però expresses both causal (‘since’) and resultative (‘therefore’) meanings. It is the occurrence of this latter resultative meaning in specific contexts with negative scope that will give rise to the adversative meaning. The details of this evolution are as follows (Giacalone Ramat & Mauri, Reference Giacalone Ramat, Mauri and Verhoeven2008: 307–12):

In Old Italian (13th c.–15th c.) peró occurs with causal meaning in the widespread sequence però que (‘since, because’). This use becomes rarer in the sixteenth century, almost disappearing from the seventeenth century on, when it gets replaced by the forms perché and poiché. During this same period of time, però is also very frequent in its resultative function, being replaced by perciò from the early eighteenth century on. Thus, after the seventeenth century, però loses its causal and resultative meanings to give way to its adversative use. The emergence of this new meaning occurs in contexts that are characterized by the presence of some wide scope negation. “In such cases, però introduces some consequence that does not take place despite expectations, determining a contrast between the cause and the denial of the expected effect” (Giacalone Ramat & Mauri, Reference Giacalone Ramat, Mauri and Verhoeven2008: 307). An example is (9).

1. (9)

   Si fu la loro immensa gloria spesso dalla invidiosa fortuna interrupta, non però fu denegata alla virtù (16th c.)

   ‘If their immense glory was often obstructed by adverse fortune, non però (not for that) was it denied to virtue.’

   [from Giacalone Ramat & Mauri, Reference Giacalone Ramat, Mauri and Verhoeven2008: 303)

While però is not contrastive per se in such negative contexts, speakers may have reanalyzed it as a marker of a specific type of contrast generated by the denial of an expected causal sequence. In other words, negated causality is reinterpreted as concession (cf. König & Siemund, Reference König, Siemund, Couper-Kuhlen and Kortmann2000). Giacalone Ramat and Mauri see an increasing frequency of the form non però from the fifteenth and especially sixteenth century onwards. Before the fifteenth century this negated form is rather rare. These negated resultative constructions frequently occur in the environment of other contrastive markers (concessive clauses or the connective ma ‘but’), thus strengthening the adversative interpretation through an invited inference (see also König, Reference König and Hawkins1988; König & Traugott, Reference König, Traugott, Hüllen and Schulze1988). During this long period of time, the adversative interpretation of però was strictly connected to the presence of negation. It is not before the early seventeenth century that occurrences are found of adversative però without negation. At that point, “the process of reanalysis is complete and però has been re-semanticized as a marker of counterexpectative contrast” (Giacalone Ramat & Mauri, Reference Giacalone Ramat, Mauri and Verhoeven2008: 311). For two additional centuries, però is semantically ambiguous between its resultative meaning and its adversative one, with the two coexisting. Yet, the syntactic context in which the two meanings are expressed is different, with resultative però being restricted to clause-initial position, so that the ambiguity is resolved through the syntactic context. From the second half of the nineteenth century, the resultative meaning of però starts to fade, and the syntactic contexts of the item with adversative meaning starts to include the clause-initial position too, which was hitherto reserved to its resultative use.

From the above-sketched evolution, it appears that the grammaticalization of però as an adversative coordinating marker (occurring mainly in clause-initial position) has fairly recently come to completion (between the end of the nineteenth and twentieth century). Several factors played a role in this evolution including competition between two forms, specific semantic and pragmatic contexts of use, including syntactic restrictions. In the following section, we will see that similar positional constraints have played a role in the semantic evolution from French temporal alors to its causal and discursive uses. Yet, in contrast with però where the semantic ambiguity has disappeared with the completion of the grammaticalization process, this is not the case with alors which remains polysemous in present-day French.

5.4.3 Today’s Polysemy Explained: The Case of French Alors

Degand and Fagard (Reference Degand and Fagard2011) take as their starting point the polysemous and polypositional use of the French connective alors (‘then’, ‘so’, ‘now/well’, ‘right’). They are interested in finding out how this linguistic expression acquired both new grammatical affordances and new meanings. Grammatically, alors evolved from a sentence adverbial with temporal meaning, to a (syntactically detached) connective marking temporal, causal or conditional relations, and eventually to a discourse-structuring marker with conversational management uses. The three main uses are illustrated in examples (10), temporal sentence adverbial, (11) causal (consequential) connective, and (12), discourse structuring marker:

1. (10)

   Mais le soir tomba sans que la pluie eût cessé. Alors, la Comtesse commit une imprudence … (20th c.)

   ‘But the night fell and the rain still hadn’t stopped. Alors, the countess got careless …’

1. (11)

   ah il adore ça alors ben tu penses bien avec moi euh il était aux aux anges hein

   ‘oh he loves it alors well you’ll guess that with me he was in seventh heaven’

1. (12)

   mais alors ce qui était marrant c’est que euh / tout à coup il s’arrêtait / et alors euh / assez vite alors xx se disait maintenant vous vous dirigez vers telle porte // mais alors

   ‘but alors the funny thing was that er / suddenly he stopped / and alors er / quite quickly alors xx was saying now you go towards the door // but alors’

   [from Degand & Fagard, Reference Degand and Fagard2011: 31–35]

The original meaning of alors is one of temporal simultaneity. It appeared in twelfth century Old French as a compound of the preposition and prefix à ‘at’ and lors ‘then’, which in turn originates in the Latin ablative form ‘illa hora’ with the meaning ‘at that hour’ (see example (13)).

1. (13)

   … selonc la costyme romaine. Et sachiés que c’estoit une feste qu’il coltivoient alors mout hautement. (Tristan en prose, 13th c.)

   ‘… according to Roman customs. And you should know that this celebration was alors held in the highest regards.’

   [from Degand and Fagard, Reference Degand and Fagard2011: 31]

Grammatically, in its use as a temporal adverbial, alors is restricted to a clause-internal position (integrated, syntactically dependent, adjunct). At the end of the thirteenth century, it appeared in contexts where it retained its temporal meaning but could take causal meanings, and occasionally also conditional meanings (from the fourteenth century on). In their corpus-based study, Degand and Fagard show that the frequency of these non-temporal uses rose steadily to 35 percent at the beginning of the fourteenth century, while 65 percent were still clearly temporal. In terms of grammaticalization, there was no recategorization (no shift of word class), yet the marker gained syntactic detachment moving from sentence-internal to sentence-peripheral (initial) position.

From the Middle French period on (15th–16th c.), clause-initial position became the preferred position for alors. Strikingly, it is in this new syntactic position that new meanings emerged: first temporal, causal and conditional meanings with a connecting function (from the seventeenth century on), then new discourse structuring meanings (mainly topic shifters and transition markers) from the twentieth century on. Thus, this new syntactic position with extended grammatical scope gave way to new meanings with extended scope over the whole clause. This is in line with Heine et al.’s (Reference Heine, Kaltenböck, Kuteva and Long2021: 64) observation that new contexts of use raise new functional meanings (see Section 5.1). It is also confirmed by the many studies observing the movement from sentence-internal to sentence-peripheral position, with a tendency for the expression of subjective meanings in the left periphery (LP) and for intersubjective meanings in the right periphery (RP). This phenomenon was described first in the contributions in Beeching and Detges (Reference Beeching and Detges2014), and has, since, been labelled the Subjectivity, Intersubjectivity and Peripheries Hypothesis (SIPH; Salameh, Estélles & Pons Bordería, Reference Salameh, Estellés, Pons Bordería and Beeching2018), giving rise to quite some discussion regarding its generalizability (Pons Bordería, Reference Pons Bordería, Pons Bordería and Loureda2018). Notwithstanding this discussion, there is agreement that peripheral position plays an important role in the development of discourse markers and discourse connectives as a locus for the development of new meanings (see, e.g. Onodera & Traugott, Reference Onodera and Closs Traugott2016; Heine et al., Reference Heine, Kaltenböck, Kuteva and Long2021; Van Olmen & Šinkūnienė, Reference Van Olmen and Šinkūnienė2021) The connective alors is but one of the many examples of this phenomenon.