Introduction: the development of manner expressions into complementizers or quotatives

Saxena (1995) drew attention to the fact that manner expressions equivalent to English thus or so may develop into complementizers or quotative markers. She suggested a pathway of development along which manner expressions first develop a function as quotative marker, which then gives them the potential to acquire the same range of functions as verbs of speaking that developed into complementizers. Saxena further proposed that this development proceeds unidirectionally along the following implicational hierarchy (including other functional extensions of manner expressions):

In later work, the specific aptitude for manner expressions to develop into quotatives as in (2) or complementizers as in (3) has been confirmed to be widespread (Boye and Kehayov 2016; Buchstaller and Alphen 2012; Güldemann 2008: 317–328; Heine and Kuteva 2002: 274, 258; Jędrzejowski and Umbach 2023; König 2015; Treis and Vanhove 2017).

At the same time, Saxena’s proposal has been refined in various ways. Attention has, for instance, been drawn to the different types of manner expressions that can be the source of quotatives and/or complementizers: these can be manner demonstratives, e.g., thus or so in (1) and (2), but also manner question words (e.g., how), manner nouns (e.g., way), manner affixes, and closely related similative verbs (e.g., resemble) or similative adpositions (e.g., like). The different morphosyntactic status and constructional environments of the source element may, however, have a tendency towards different developmental pathways (e.g., Diessel and Breunesse 2020: 334 on different types of demonstratives as sources), but for a better understanding of this, we need more detailed historical descriptions of change. Moreover, markers of comparison or similarity may be the input to the change as well as the output as in (1), which opens up different directionalities of change than Saxena’s (e.g., Treis and Vanhove 2017; Wiemer, this issue). Finally, it has been shown that steps in Saxena’s pathway can be skipped. Specifically, complementizers may develop directly from manner expressions without quotatives as an intermediate step: the quotative stage is often not attested, or it is infrequent.

Still, the diachronic development of manner expressions remains understudied. This has been attributed to the semantic and syntactic versatility of manner expressions (e.g., König 2015: 39–40), and to the concomitant issue of where to draw the line between manner expressions “proper” and expressions of similarity, comparison, instrument, or means (Kortmann 1997: 81, 84, 146). Recent years have seen a surge in interest in the functions that manner expressions acquire across languages, however (Boye and Kehayov 2016; Jędrzejowski and Umbach 2023; König 2015; Næss et al. 2020; Treis and Vanhove 2017).

The aim of this special issue is to add to the knowledge about manner expressions and their development by contributing a number of in-depth studies of the development of specific manner expressions in specific languages. In continuation of Saxena’s (1995) study and König’s (2015) wider overview of the development of manner deictics, we are especially interested in the development of manner expressions into complementizers and into quotative markers. These two developments therefore constitute the two main themes in the contributions to the special issue.

While the contributions provide further evidence for these developments, they also emphasize the considerable differences covered by the generalization into development types or paths. For instance, it is well known that elements that derive from manner expressions may differ in terms of epistemic meaning. Some are functionally specialized for non-presupposed clausal contents in the domain of direct and/or indirect speech and thought reporting, for instance into quotative, i.e., direct speech-related, uses (e.g., so in [2] and be like) (see especially Buchstaller and Alphen 2012; Güldemann 2008: 317–328). Others specialize for so-called presupposed, “factive”, propositions (e.g., how in [3]) rather than illocutions (Boye and Kehayov 2016; Legate 2010; Nye 2013). Yet another type of marker deriving from manner expressions does not necessarily involve complement clauses but involves a marker perspectivizing a clause as counterfactual or proven wrong. So-called “mistaken belief” clauses as in (4), for instance, have been in part argued to relate to similative uses (cf. the preposition ‘like’) (e.g., Breen 1984; Evans 1995, 2003; McGregor 2023).

In line with this, the contributions to the present issue find that the development of manner expressions into complementizers may be accompanied by distinct semantic developments. In Basque, Russian, and Semitic, the resulting complementizer has been argued to be factive, at least at certain stages of development, while in Estonian, Finnish, and Polish, the resulting complementizers represent a lower degree of certainty.

In terms of pathways of change, Saxena’s pathway might capture the development of various quotatives and complementizers derived from cataphoric manner deictics (Guz, this issue; Teptiuk and Tuuling, this issue), but most of the papers in this issue propose alternative developmental paths starting from manner expressions, namely:

1. from a similative adposition ‘(be) like, as’ to complementizer, possibly through a reanalysis of the similative part as a relative/content clause or as constituent of a complex transitive clause (Hernaíz);

2. from a similative marker ‘as’ with variable scope univerbated with the irrealis enclitic by to a complementizer or particle in contexts of hearsay (Wiemer);

3. from an interrogative pronoun ‘how’ with a broad meaning centered on manner to a complementizer, possibly through the integration of an independent how-clause as a second object or relative clause in constructions of the type ‘look at/think about X, how X …’ (Serdobolskaya and Kobozeva);

4. from a reinforced manner expression ‘(it is) so’ which developed uses as a focus particle to a complementizer verbal prefix (Krajewska);

5. from a derivational suffix forming deverbal nouns which can focus on the manner in which an event takes place to a bound complementizer scoping over full clauses (Rentzsch).

The described developments highlight various features central to the trajectories undergone by manner expressions. For manner deictics, it has been proposed that the central semantic feature is that of expressing similarity. This allows manner deictics to compare entities based on the content dimensions of manner, quality/type, and degree (König and Umbach 2018). As König and Umbach point out, similarity need not involve identity; as a result, the expression of similarity naturally yields approximative meanings besides equative/alignment meanings. This contrast may lead to highly context-dependent contrasting developments in terms of speaker-related commitment or epistemic distancing, which can be more transparent in the presence of other markers of uncertainty, hedging, hypotheticality, or counterfactuality (see e.g., Wiemer, this issue; Teptiuk and Tuuling, this issue). The notion of similarity is moreover central to the cognitive mechanism of classification, allowing for the creation of ad-hoc “kinds”, i.e., “similarity classes” (König and Umbach 2018: 304). This typifying function of similarity expressions may play a role in their development into nominalizing suffixes, with the resulting deverbal nouns referring to types of entities (Rentzsch, this issue). The relevance of similarity-based categorization also shows in the possibility for the resulting quotative markers to come to be used together with the quotes they introduce to predicate qualities (properties or states) on the part of the reported speaker (Guz, this issue; Teptiuk and Tuuling, this issue). Moreover, the fact that manner expressions may invoke similarity with different types of antecedents, depending on whether they foreground manner, quality, or degree, makes them highly flexible in terms of scope (see also Wiemer, this issue). This feature may be an important factor underlying continued patterns of semantic extension from the earliest attestations onwards. The scope of the manner expression may as such easily be extended, for instance from nominal to clausal attachment sites (Rentzsch, this issue) or narrowed, e.g., either losing the potential to introduce propositional contents (Serdobolskaya and Kobozeva, this issue) or developing conventionalized subordinating functions from an original discursive use as a focus particle (Krajewska, this issue). With manner question words as source, this flexibility may result in different types of complementizers (manner complementizers, eventive complementizers, propositional complementizers) either through bleaching of the manner semantics and neutralization of the declarative-interrogative contrast in complement functions, or through integration of independent clauses with the same broad semantic potential already in (rhetorical) interrogatives directly into complement functions (Serdobolskaya and Kobozeva, this issue). In what follows, we conclude this introduction with a summary of each of the contributions.

Guz focuses on quotative functions of Polish tak and tak-i in contemporary spoken data. Both forms combine similative and demonstrative semantics (cf. ‘like this’) in one word, and can ultimately be traced back to the same root in Proto-Slavic, *takъ ‘such’. In quotative constructions tak and tak-i are often interchangeable but show restrictions and meaning differences reflecting a different source form, with tak relating to an invariable adverbial form with manner semantics ‘in this manner, so’, and tak-i being an inflected adjectival form with quality semantics ‘such (a), like this’. These differences also show in two verbless quotative strategies of the type ‘(Conj) NP tak/tak-i QUOTE’, cf. ‘And he thus: “Aah!”’, for which two different developmental trajectories are suggested. The verbless quotative construction with the adverbial form tak is argued to derive from a more schematic, possibly reportative, full clause (NP Verb tak), through constructionalization and conventionalization for quotative uses. The verbless quotative construction with the adjectival form tak-i maintains its quality semantics in such a way that the quote it accompanies is either modified or hedged itself, or is presented as describing a state or property of the reported speaker. This second verbless quotative construction is therefore argued not to stem from reduced clauses but to involve a natural extension of tak-i’s quality-based similarity semantics to approximative, hedging, and quotative uses in combination with other quotative strategies.

Teptiuk and Tuuling describe the range of functions that manner deictics (cf. ‘thus’) and similative adpositions (‘like, in the manner of’) have in contemporary social network data from different Finno-Ugric languages (Finnish and Estonian, with a comparison to initial findings for Komi, Udmurt, and Erzya). The two types of manner expressions can function in typical quotative constructions to introduce reported discourse with predicates like ‘say’, ‘think’, or ‘be’, but also in the context of predicates of perception or knowledge. With a focus on such perception and knowledge predicates, the authors find that the manner demonstratives and similative adpositions can be used in a range of different functions with respect to the clause they introduce, which can involve endophoric reference to verbatim or approximate representations of reports or interpretations, a stress on the subjective interpretation by the reported speaker, a generalized use of the report as describing a characteristic property or stance on the part of the reported speaker, and even counterfactual or desiderative functions. These different functions can be grouped together as non-factive functions, as they stress either the (approximate representation of the) reported discourse itself, or highlight the subjective, reported speaker-describing, less than certain, hypothetical, or counterfactual nature of the content of the introduced clause. The choice of main clause predicate and of manner expression, as well as lexical and grammatical markers of uncertainty, hedging, hypotheticality, or counterfactuality (e.g., expressions like ‘somehow, kind of’; conditional mood; adversative clauses) are important cues in distinguishing the different semantic overtones in different contexts.

Hernaíz reexamines existing hypotheses for the source of k-complementizers in Semitic, based on documented occurrences of (multifunctional) k-markers in Old Akkadian, Old Babylonian, and Old Assyrian data. Earlier work and existing grammaticalization pathways suggest that these may have originated in reanalyses of causal clauses, embedded how-questions, quotative uses, or similative uses. The author argues that there is not sufficient evidence in favor of any of the first three proposals. Instead, it is suggested that complement clauses introduced by k-markers may stem from two different types of constructions in which the main predicate takes an NP followed by a clause introduced by a k-marker, with the k-element going back to a (non-deictic) similative marker (‘[be] like, as’). These are (i) constructions with mainly abstract locution nouns such as message followed by a clause representing its content, and (ii) constructions with a first-order entity NP (e.g., ‘him’) and a clause predicating information about the referent in the first NP. Omission or integration of the initial NP would then result in the k-marker introducing a complement clause. In contrast to earlier proposals, this scenario is argued to be a better fit for attested distributional contexts (explaining contexts of speech-related predicates as well as those of more frequent cognition and perception predicates) and is overall more in line with attested polysemies and contexts of use across the Semitic language family. There are also manner deictics derived from the same root as the k-complementizers that function as endophoric markers pointing to the beginning or end of directly reported discourse.

Krajewska similarly looks into the diachronic source of a subordinator, i.e., that of the verbal prefix bait- in Basque. Bait- was already multifunctional in the earliest texts in the corpus; in subordinate contexts it could already function as a subordinator in relative, adverbial, and complement clauses. Still, Krajewska’s analyses of attested ambiguities and changes in distribution, in combination with considerations of etymology and documented pathways of grammaticalization, allow her to propose a stepwise development of different functions from an original manner expression bai ‘as, so’ with anaphoric uses. Bai is argued to have independently given rise to uses as an affirmative particle (‘yes’), focus particle (cf. ‘also, even’) and to a reinforced form bait. A regular occurrence in subordinate contexts of focus particle uses, or of broader discursive uses marking e.g., emphasis or contrast, could then become conventionalized in different types of subordinate constructions and be reanalyzed as a marker of subordination. In terms of semantic specialization, bait- is mainly associated with clausal contents that the speaker presents as informationally backgrounded or true: besides uses in relative clauses and complements to nouns (e.g., ‘the reason that …’), subordinator bait- occurs in adverbial clauses of reason, result, or manner, and in complements to implicative predicates like ‘happen’ and emotive-evaluative predicates like ‘be happy; it’s a pity’. The data suggests that the first subordinator uses may have been those in relative clauses, with adverbial reason clauses becoming predominant later, and the use in complements to emotive factive predicates being even more recent, with potential bridging contexts connecting these.

Serdobolskaya and Kobozeva investigate the history of the Russian subordinator kak ‘how’. They show that in Old Russian, kako/kakъ had a central manner meaning, but was further used in a wide range of functions in both independent and subordinate (complement and adverbial) clauses. Over time the data attest to an increasing specialization: most notably, kako/kakъ could introduce both eventive and propositional clauses in Old Russian but has lost the propositional function over time. In terms of development, the authors examine the pathway from question to subordination (cf. Heine and Kuteva 2006), through which a question word targeting manner in independent sentences (e.g., ‘How did you do this?’) could be embedded as a question word targeting manner in indirect questions (‘I asked how you did this’) before resulting in a manner complementizer (‘I know how you did this’). In the latter function, prior work has argued that the manner meaning could be bleached, thus giving rise to eventive and/or propositional complementizer functions. Serdobolskaya and Kobozeva argue that the intermediary step of manner complements is not required for the development of eventive and propositional functions: independent questions with kako/kakъ ‘how’ can contextualize to a broad set of meanings already in Old Russian, as they are used not only for questions about manner but also for questions about a situation as a whole, in (rhetorical) polar questions, in questions about time, reason, etc. For the integration of these functions in complement clauses, it is suggested that a possible bridging context can be found in a double object construction (of the type ‘think about/look at X, how X …!’), in which a shared referent can lead to the integration of a formerly independent kako-clause as a complement. This construction has also been highlighted for the development of English how from manner question word to propositional complementizer (López-Couso and Méndez-Naya 1996).

Wiemer presents a diachronic study of Polish jakoby, which has a prominent function in present-day usage as a particle (cf. ‘allegedly’) or complementizer in contexts of hearsay. The historical development of the reportive complementizer runs inverse to Saxena’s proposed pathway, as it starts rather than ends with a similative marker jako ‘as’, in this case univerbated with the irrealis enclitic by. The source domain of similarity marking lends itself well to the development of the reportive complementizer: firstly, the similarity marking itself can already scope over propositions and involve complements (e.g., of predicates of perception like ‘seem as if …’); secondly, the marking of similarity in combination with irrealis also implies a lack of full epistemic support. The shift for jakoby is reminiscent of Gipper’s (2018) proposed pathway from similative to reportive function, but lacks an intermediate inferential use, which marks that the conveyed judgment is based on inference as a source of evidence.

Rentzsch examines the Turkish suffix -(y)Iş, which had the function of forming deverbal nouns already in Old Turkic. A productive subset of the resulting nouns has the added semantic value of referring to the manner in which an event takes place (e.g., duruş ‘the manner of standing’, from dur- ‘to stand’), which suggests that a manner component is an important semantic feature of the nominalizing function of -(y)Iş. Besides noun formation, it is argued that the suffix has developed into an inflectional marker on nominalized clauses, including those in complement functions. Within the paradigm of bound verbal noun markers that can assume the function of marking complement clauses, it is proposed that the main competitors -DIK and -mA respectively emphasize ‘knowledge about an event’ and ‘the event itself’, while the emergent marker -(y)Iş retains a more specialized interpretation focusing on the concrete circumstances or nature of the event in the nominalized clause. This may involve a degree of persistence of the original manner meaning, whereas in other Turkic languages, cognate nominalizers have developed further into a more general verbal noun marker for clauses referring to states-of-affairs.