Left and right peripheries in discourse: an introduction

1 Left and right peripheries in discourse: research aims and questions

Over the past one or two decades, there has been an increasing amount of research on the internal structure of discourse units (henceforth DUs) in general, and the left and right peripheries of DUs in particular, which is reflected in a large number of studies investigating the peripheries and their role in the structuring of discourse across a wide range of languages and from both synchronic and diachronic perspectives (see, e.g., the contributions in Beeching and Detges 2014a; Hancil et al. 2015; Van Olmen and Šinkūnienė 2021a). While the internal structure of – and the linguistic elements occurring at – the left periphery (henceforth LP) and the right periphery (henceforth RP) have been largely ignored in traditional, sentence-based accounts of grammar, recent discourse-oriented approaches to grammar – including Systemic Functional Grammar (Halliday et al. 2014), Functional Discourse Grammar (Hengeveld and Mackenzie 2008) or Discourse Grammar (Heine et al. 2013) – have come to acknowledge the important role that the peripheries and their constituent linguistic elements play for the structuring of discourse, thus contributing to providing “a complete and adequate account of the grammatical organization of connected discourse” (Dik 1997a: 12; see also Fetzer 2014a: 50; Mackenzie 2016: 476).

Research on LP and RP has so far addressed a number of important issues from various perspectives (see Sections 2 and 3 for a more detailed discussion), including the question of granularity (i.e. the nature of the unit in relation to which LP and RP can be defined), the question of how the peripheries of DUs are internally structured and linguistically realized (i.e. which linguistic elements LP and RP consist of), and the question of where to draw the boundary between the core and the peripheries of DUs (i.e. whether to consider the boundary between core and peripheries as discrete or gradient). The overall aim of this special issue is to further elaborate on these key topics by providing fresh perspectives on – and further insight into – both the internal structure of LP and RP across languages (in terms of which linguistic categories occur at the peripheries and how they are sequentially ordered) and the conceptualization of the boundary between the core and the peripheries of DUs. In order to achieve this aim, this special issue addresses the following research questions:

1. Which structural slots do DUs in general – and LP and RP in particular – consist of?

2. How (i.e. by means of which linguistic material) are the structural slots at LP and RP realized across languages?

3. How can the co-occurrence and sequential ordering of linguistic elements at the peripheries be modelled?

4. How can the boundary between the core and the peripheries within DUs be defined and conceptualized?

The six contributions to this special issue focus on the topics and questions outlined above from theoretical and empirical as well as synchronic and diachronic perspectives. They investigate linguistic phenomena at LP and RP across various languages, including English (Lewis; Mustafa and Kaltenböck; Pinson), Italian, Spanish and Basque (Coniglio and Monforte), Danish (Heegård Petersen) as well as German and Persian (Lehmann et al.).

2 Discourse and its constitutive parts

Discourse may be examined from various perspectives, each of which addresses issues that are fundamental to any analysis of – and approach to – discourse. From the perspective of discourse syntax, on the one hand, discourse may be conceptualized as a structural configuration consisting of a varying number of concatenated and linearized DUs that constitute stretches of spoken and/or written discourse. The meaning of this linearized whole is understood as being more than the sum of the meanings of its constitutive parts (i.e. the individual units of discourse), with the particular sequential organization of DUs triggering specific discursive meanings which cannot be deduced from the propositional meanings of the individual DUs themselves (see, e.g., Fetzer 2010: 17–18; Hodge 2017: 521–522). Discourse semantics and discourse pragmatics, on the other hand, consider any piece of discourse to underlie “a general default principle of coherence” (Bublitz and Lenk 1999: 157). Discourse is thus conceived of as a coherent, meaningful whole, and the main questions to be addressed from a discourse-semantic and discourse-pragmatic perspective are how the constitutive parts of discourse are semantically and pragmatically connected and how discourse participants “construe a coherent representation of discourse both at a global stage for discourse-as-a-whole and locally for embedded discourse units within the discourse” (Klumm et al. 2023: 5).

While there has over the past few decades been a vast amount of research adopting discourse-semantic and discourse-pragmatic perspectives on discourse and investigating the nature of the connectedness of DUs by way of discourse relations (for an overview, see Klumm et al. 2023), this special issue adopts a largely discourse-syntactic approach to discourse. From this perspective, particular emphasis is being placed on the questions of (i) how DUs can be defined and conceptualized, and (ii) how these constitutive parts are internally structured and linguistically realized. The questions of granularity (i.e. the nature of the basic unit of investigation) and the internal structure of DUs are addressed in Sections 2.1 and 2.2, respectively, while Section 3 zooms in on the left and right peripheries of DUs, focusing on their internal structure and linguistic realization.

3 The left and right peripheries of discourse units

While Section 2 has provided an overview of the internal structure of DUs in general, this section zooms in on the left and right peripheries of DUs, focusing in more detail on three widely discussed and interrelated issues, i.e. (i) the linguistic elements which may occur at LP and RP, (ii) the internal structure of LP and RP in terms of structural slots, and (iii) the conceptualization of the boundary between the core of a DU and its peripheries.

As far as their linguistic realization is concerned, the peripheries of DUs are generally agreed upon by scholars to be occupied by a wide range of linguistic categories, including – among many others – discourse connectives (see, e.g., Celle and Huart 2007), comment clauses (see, e.g., Brinton 2008), question tags (see, e.g., Martínez Caro 2020), general extenders (see, e.g., Traugott 2015) or left/right dislocations (see, e.g., Averintseva-Klisch 2009). The contributions to this special issue focus on several further categories occurring at LP and/or RP: Diana Lewis investigates the sentence adverbs apparently and sadly as well as the discourse connector instead in English; Ozan Mustafa and Gunther Kaltenböck focus on evidential last I (ELI) fragments in English; Mathilde Pinson examines one particular type of the N-is construction in English (i.e. (the) bottom line (is)) and its surrounding elements at LP; Marco Coniglio and Sergio Monforte focus on left-peripheral optative particles in Italian, Spanish and Basque; Jan Heegård Petersen investigates the use of different types of adverbials in left-dislocated position in spoken Danish; and Nico Lehmann, Vahid Mortezapour, Motahare Sameri, Elisabeth Verhoeven and Aria Adli compare the use of different types of right-peripheral subjects (i.e. extraposition, right dislocation and afterthought) in German and Persian.

The various categories outlined above have been subsumed under different umbrella terms such as pragmatic markers (following the broad definition by Fraser 2009), parentheticals (Dehé and Kavalova 2007a), theticals (Heine et al. 2013; Kaltenböck et al. 2011) and extra-clausal constituents (Dik 1997b; Kaltenböck et al. 2016a). For the sake of consistency, it is the notion of extra-clausal constituent (henceforth ECC) that will be used in the remainder of this introductory chapter, given that the term nicely captures the idea that those linguistic elements classified as ECCs are – in some way or another – ‘outside’ the clause to which they are attached. Dik (1997a: 310) defines ECCs as linguistic elements “which are not part of the clause proper, but [which are] more loosely associated with it in ways which can most adequately be described in terms of pragmatic functionality”. More specifically, ECCs are characterized by a number of formal and functional features, the most important of which are summarized in Table 1 below.

Looking at the functions of ECCs from a diachronic perspective, elements at LP have been hypothesized to have undergone a process of subjectification and thus to be subjective, whereas those at RP have been hypothesized to have undergone a process of intersubjectification and thus to be intersubjective (see Beeching and Detges 2014b; Salameh Jiménez et al. 2018). The universality of this hypothesized functional asymmetry between LP and RP has been tested – and challenged – in numerous studies across languages (see, e.g., the contributions in Beeching and Detges 2014a; Van Olmen and Šinkūnienė 2021a).

While the features provided in Table 1 are generally considered sufficient for assigning linguistic elements to LP/RP and for distinguishing them from elements belonging to the core of a DU, Kaltenböck (2007: 31) claims that the formal and functional properties of ECCs are often presented “without clear indication as to which criteria are taken as primary” for the assignment of extra-clausal status. Indeed, previous research has demonstrated that not all characteristic features apply to all left- and right-peripheral elements to the same extent. Traugott (2015), for instance, convincingly shows that some of the features are applicable to a limited number of ECCs only, which leads her to conclude that the characteristic features of ECCs “cited repeatedly are weak tendencies […] and cannot be used as diagnostics” (Traugott 2015: 127). Given both the wide range of types of ECCs and the considerable amount of variation with regard to their formal and functional characteristics, it can thus be argued from a cognitive-prototype perspective that some linguistic elements generally considered to belong to the peripheries of DUs are more prototypical members of LP/RP than others. In other words, the more of the formal and functional features described in Table 1 apply to linguistic elements, the more they are prototypical ECCs and thus prototypical members of LP/RP.

While the internal structure of DUs in general has been modelled in different ways in various languages (see Section 2.2), scholarly attention has only relatively recently come to focus on a more specific modelling of the internal structure of LP and RP in terms of patterned co-occurrences of peripheral elements and their sequential ordering within LP and/or RP. The structural slots within LP and RP and the ordering preferences of peripheral elements has been investigated in more detail in several languages, including French (e.g. Crible and Degand 2021; Detges and Waltereit 2014), Spanish/Catalan (e.g. Cuenca and Marín 2009; Pons Bordería 2018), German (e.g. Averintseva-Klisch 2009; Frey 2018, 2023), Dutch (e.g. van der Wouden and Foolen 2015, 2021) and English (e.g. Cuenca and Ludivine 2019; Fetzer 2014b; Fraser 2015; Klumm 2024; Koops and Lohmann 2015, 2022; Lohmann and Koops 2016). While most studies investigating the co-occurrence of linguistic elements at the peripheries of DUs have focused on sequences of ECCs at LP (see also Section 2.2 on multiple Themes in Systemic Functional Grammar), there are only few studies which have examined in more detail the sequential ordering of linguistic elements at RP (see, e.g., Haselow 2019; Izutsu and Izutsu 2021; Klumm 2024; van der Wouden and Foolen 2015). What most scholars working on the internal structure of LP and RP agree on, however, is the finding that the particular ordering of linguistic elements at LP and/or RP is highly constrained and systematic in that it is determined by the function that the respective ECCs fulfil within the discourse (e.g. discourse-structuring vs. (inter)subjective). This argument is also pursued by Pinson (this issue), who investigates the various linguistic elements surrounding the English construction (the) bottom line (is) at LP with regard to their respective discourse-pragmatic functions. Finally, as far as the number of potential structural slots at LP and RP is concerned, the studies mentioned above vary widely in their scope, focusing on a minimum number of two slots at LP and/or RP and up to a maximum number of four potential slots at LP in Dutch (see van der Wouden and Foolen 2015) and nine potential slots at LP in English (see Pinson this issue).

A final issue that has received increasing scholarly attention over the past few years relates to the question of how the boundary between the core of a DU and its peripheries can best be conceptualized. Generally speaking, a broad distinction between elements belonging to the core of a DU and elements belonging to LP/RP can be made on the basis of the formal and functional features listed in Table 1: While elements at LP/RP are typically characterized as being prosodically independent, syntactically optional, semantically non-restrictive and non-truth-conditional, elements belonging to the core of a DU are typically defined by way of exclusion, i.e. as being prosodically integrated, syntactically obligatory, semantically restrictive and truth-conditional. That such a broad distinction is too simplistic has already been shown in the discussion of ECCs above, which has revealed that linguistic elements at the peripheries of DUs are best approached from a cognitive-prototype perspective, given that not all characteristic features provided in Table 1 apply to all ECCs to the same extent. Such a prototype-anchored conceptualization cannot only be applied to the internal structure of LP and RP (as described above), but also to the boundary between the core of a DU and its peripheries. As has been argued for in recent research (see, e.g., Klumm 2024; Rizzi and Cinque 2016; Traugott 2015), the boundary between elements at the peripheries and elements belonging to the core of DUs is best described as fuzzy and gradient (rather than as clear-cut and discrete) because many linguistic elements have an intermediate status with respect to the features listed in Table 1 in that they may fulfil some of the features (and may thus be considered to belong to the peripheries) while they may not fulfil others (which would suggest classifying them as belonging to the core). Klumm (2024), for instance, discusses the borderline status of sentence adjuncts in English, which are characterized by features applying both to core constituents (i.e. they contribute to the propositional content, and affect the truth conditions, of a DU) and to elements at the peripheries (i.e. they are syntactically optional and positionally mobile). The gradient nature between core and peripheries is also argued for in the contribution to this special issue by Diana Lewis, who investigates the development of sentence adverbs and discourse connectors in English on the basis of apparently, sadly and instead. By adopting a diachronic perspective, Lewis shows that the distinction between core and peripheries is blurred during language change, with the boundary being permeated progressively, imperceptibly and via hybrid uses. This is in line with Traugott (2016: 27), who argues that LP and RP (as well as the linguistic elements considered to belong to the peripheries) “are on a continuum from ‘inner core clause’ position to ‘outer’ position, since many originate in adverbs and other expressions used within the argument structure of the clause”.^[2]

4 Contributions to this special issue

In her contribution, Diana Lewis adopts a diachronic viewpoint on the English left periphery and the boundary between core and periphery of a DU. She compares the development of three speaker-oriented sentence adverbs that started life in the core of the clause, but now typically occur in the periphery, namely the evidential adverb apparently, the evaluative adverb sadly and the discourse-connecting adverb instead. All three evolved into peripheral stance expressions by similar mechanisms. English recruited a large number of evidential and, later, evaluative sentence adverbs from the Early Modern English period onwards. Lewis traces their development using several contemporary and historical corpora of British English, e.g. the Old Bailey Corpus, the EEBO collection, and the spoken section of the British National Corpus. Apparently developed from a modal adverb, denoting visibility to observer and speaker (placed in the core) towards an evidential adverb of hedged validity (placed at the periphery). The bridging context were probably stative predicates, especially cases in which an adjectival (or participial) predicate was modified. With sadly, the bridging context also seems to have been the modification of an adjectival or participial predicate. Here, the evolvement from a modal adverb to a sentence adverb needs an intermediate step of usage as sadly as negative intensifier. Instead did not undergo any semantic change, but a change in the possible conjuncts, from entities (typically expressed by noun phrases) via predicates (typically represented by gerunds) to propositions. In all three cases, the shift in meaning was accompanied by a syntactic shift of these elements towards the left periphery, eventually crossing the fuzzy boundary between core and periphery. Typically, while being in the borderland, so to say, there are many instances in which the usage is ambiguous between the modal core reading and the speaker-oriented peripheral reading, which facilitates the semantic and syntactic change of these items.

The paper by Ozan Mustafa and Gunther Kaltenböck focuses on an extra-clausal fragment in English, namely last I/we + V, (e.g. last I checked), and traces its formal and functional development. The authors draw their conclusions from a corpus study of the Corpora of Historical and Contemporary American English (COHA and COCA). The origin of these fragments are full-fledged sentences in specificational function (i.e. a copula construction in which one element, the variable, is defined by another element, the value, or equated with it) that start to become truncated around 1860 and are used purely as downtoners in the beginning, indicating that the speaker does not want to wholly commit themselves to the truth of the proposition because of limited available evidence. The area of usage extends in the course of the 20th century to booster functions (where, just opposite to the original use, the reliability of the source is emphasized) and ironic usage (exploiting a pragmatic incompatibility of fragment and host clause). The verbs used in this construction were originally verbs of physical or mental sensation such as hear, which remains the most common one up to now. In the 1940s, the verb check entered the picture and soon became the second-most frequent one. This predicate (and others first used in the last I/we + V-construction) indicates a semantic change in this construction, as with this predicate presentational sentences are not possible anymore, but only readings in which last functions as a temporal adjunct. This indicates a specialization of this fragmentary form towards evidential use, while the full versions remain non-evidential. At the same time, a shift in allegiance occurs, in which this construction in general is increasingly associated with temporal adjuncts. An increasing independence from their host clause is indicated by several formal changes, such as the loss of compulsory temporal agreement between the last I/we + V fragment and the host clause. Eventually, the fragment has acquired discourse-structuring functions, indicating topic shifts.

In investigating N-is sequences, especially the sequence (the) bottom line (is), Mathilde Pinson shows that LP in English can be divided into nine optional slots, the first four being reserved for expressions marking disjunction with the previous discourse, the latter five being more argumentative in character, enhancing the validity of the upcoming assertion and trying to convince the listener of it. Introductory fragments in general started life diachronically as main clauses but were reduced and are now parenthetical fragments introducing an independent clause. They do not contribute to the truth-conditional meaning of the clause introduced by them, but they have several other functions, for example giving the speaker extra time to plan their utterance, and signalling the pragmatic function of the upcoming utterance. The introductory fragment under consideration, (the) bottom line (is), is topic-oriented and has the function of signalling that the previous discourse is being summarized, a change in sub-topic is launched, or a digression is ended. Using the spoken section of the Corpus of Contemporary American English, Pinson focuses on left-peripheral elements surrounding bottom line in such sequences and explores whether there are ordering preferences. Many elements show very clear tendencies: Connectors like but, so etc., attention getters and well nearly categorically precede bottom line, whereas addresses, expressions of reportative evidentials and though are almost always found after bottom line. Taking the relative ordering of the elements co-occurring with bottom line into account, Pinson manages to establish nine slots at LP preceding the core clause: 1: well, 2: connectors encoding discursive links to the preceding discourse, 3: I mean, 4: attention getters, 5: subjectifiers, 6: bottom line, 7: though, 8: desubjectifiers, 9: addresses. The order of these slots displays movement from establishing a (disjunctive) link to the previous discourse to the (positive) validation of the present utterance.

While the three papers forming the first part of the special issue (Lewis, Mustafa and Kaltenböck, and Pinson) discuss the development of English left-peripheral elements, the three papers forming the second part (Coniglio and Monforte, Petersen, Lehmann et al.) focus on other languages, including as German, Danish, Italian, Spanish, Persian, and Basque. Marco Coniglio and Sergio Monforte investigate optative particles in various languages. Optative particles fall into two classes, i.e. vacuous optative particles (or optative I-particles) on the one hand, in which the optative meaning is only emphasized, but other linguistic means are necessary to convey the notion of optativity (such as German ‘nur’, English ‘only’, Italian ‘solo’), and optative markers (or optative C-particles) on the other hand, in which the particle is a sufficient denotator of optativity (such as Italian ‘magari’, Spanish ‘ojalá’, Basque ‘agian’). Vacuous optative particles tend to occur within the clause proper, while optative markers tend to occur at the left periphery of clauses. The grammaticalization path of both classes of particles is different: While vacuous optative particles tend to emerge from lower adverbial elements, optative markers usually stem from former matrix clauses with meanings like ‘I wish’, ‘it may be’ etc. In this respect, they are comparable to the English left-peripheral elements discussed by Pinson as well as Mustafa and Kaltenböck. All optative particles have in common that LP must be activated, either by the particle itself (in the case of optative markers) or by other means such as verbal movement (in the case of vacuous optative particles). The term left periphery is understood here in the generative sense as the leftmost part of the clause proper, quite unlike the usage in the more functionalist literature represented by the papers in the first part of the volume, in which the periphery is an area outside the clause proper. The different positions and properties of these particles are a consequence of their being licensed in different projections in the syntax tree.

The contribution by Jan Heegård Petersen describes the distribution and functions of fronted adverbials in spoken Danish. There is variation in positioning them in the prefield (i.e. the position immediately before the finite verb in a verb-second language such as Danish) or in a position even further to the left, the so-called Extra-left position, which is presumably outside the core clause, that is, a left periphery conceptualized in the same way as in the papers dealing with English by Lewis, Mustafa and Kaltenböck, and Pinson. If adverbials occur in Extra-left position, a resumptive pronoun occurs in the prefield. The construction is thus quite comparable to German left dislocation and occurs in other Scandinavian languages as well. Some classes of adverbials, e.g. conditional clauses or temporal clauses (especially the ones introduced by når ‘when’), are very prone to appear in Extra-left position, while others, like attitude and modal adverbials (e.g. heldigvis ‘luckily’ or alligevel ‘in spite of’) are found there only in roughly half of the cases, otherwise standing in the prefield. The variation is induced by grammatical and discourse-structuring factors, and to a lesser extent by lexical weight, which has been suggested as the main factor in earlier literature. To identify these factors, Petersen conducted a corpus study of transcripts of sociolinguistic interviews. One important factor is the type of adverbial: While clausal adverbials are in general more prone to the Extra-left position, there are differences in with regard to their semantic function, with conditional clauses being more frequently found in Extra-left position than temporal clauses. As for non-clausal adverbials, summarizing adverbials are found in Extra-left position in 70 % of cases, whereas other adverbials occur there less frequently. Only 17 % of attitude adverbials, being the least frequent group, are found in Extra-left position. As for discourse structure, the construction has been claimed in previous research to indicate the closing of a discourse unit and the introduction of a new point of departure, which is partly corroborated by Petersen’s study.

In contrast to the other five studies, the paper by Nico Lehmann, Vahid Mortezapour, Motahare Sameri, Elisabeth Verhoeven and Aria Adli is concerned with the right periphery rather than the left periphery. The main question leading their investigation is whether subjects in the right periphery behave similarly with respect to registers in German and Persian, both being structurally verb-final languages (the verb-second effect in German comes about by movement of the verb to the left in main clauses). In both languages, as in all languages that have been researched in this respect, placement of the subject to the right periphery is a rare phenomenon. The area after the clause-final verbal complex is argued to count as right periphery. Distinguishing several structural slots for such right-peripheral material with a different degree of bondage to the core clause (as indicated e.g. by prosody or the compatibility with correlates), the authors identify four formally distinct types of right-peripheral subjects, i.e. extraposed subjects, right-dislocated subjects, afterthought subjects with a pronoun correlate in the core clause, and afterthought subjects with a full NP correlate. It is known that the use of peripheries is highly dependent on register and production mode, in the sense that peripheries are used at all, and in the sense that the discourse-structuring and information-structuring functions may differ across registers. Based on a corpus study on the multilingual Leg*Reg corpus, the authors found that the two types of afterthought serve the same function, namely the attribution of further qualities to the subject referent in the case of afterthought with full NP correlate, and the clarification of the referent (if it is ambiguous or too little salient) in the case of afterthought with pronoun correlate. Likewise, the functions of Persian extraposition and German right dislocation are quite similar, establishing the topic of the following discourse. German extraposition has a different function, being restricted to focused information, while in the Persian data, right dislocation is not attested. Overall, right-peripheral subjects are more often found in interactive communicative events than in monological ones. The effect has been found to be significant in Persian, but not in German, which might have to do with the fact that Persian is more ready to position subjects in the right periphery than German.