Solega defenestration: Underspecified perspective shift in an unwritten Dravidian language
Abstract
Based on original fieldwork, this paper discusses reported speech and thought constructions in Solega (Dravidian). Following McGregor (1994)McGregor, William B. 1994 “The Grammar of Reported Speech and Thought in Gooniyandi.” Australian Journal of Linguistics 14 (1): 63–92. 
we claim that reported speech can only be comprehensively characterised if it is identified as a syntactic construction in its own right, a construction we label a framing construction. In natural discourse, elements of the framing construction, particularly clauses referring to the reporting event, may be left unexpressed. We term framing constructions without a matrix clause ‘defenestrated clauses’. While defenestrated clauses in Solega leave perspective shifts underspecified, they include several distinctive strategies that allow us to reconsider the role of morpho-syntactic marking in the expression of perspective shifts.
Keywords:
Publication history
1.Introduction
Analyses of reported speech11.The linguistic phenomenon whereby speech or thought is typically attributed to an entity other than the current speaker at the current speech moment, or presented as a speech/thought event in a reality/speech event that is distinct from the current speech reality. often explore the boundaries between grammatical marking and extra-syntactic expression (Banfield 1973Banfield, Ann 1973 “Narrative Style and the Grammar Direct and Indirect Speech.” Foundations Language 10 (1): 1–39. ; D’Arcy 2015D’Arcy, Alexandra 2015 “Quotation and Advances in Understanding Syntactic Systems.” Annual Review of Linguistics 1 (1): 43–61. ; Dancygier and Vandelanotte 2016Dancygier, Barbara, and Lieven Vandelanotte 2016 “Discourse Viewpoint Network.” In Viewpoint and the Fabric of Meaning: Form and Use of Viewpoint Tools across Languages and Modalities, ed. by Barbara Dancygier, Wei-lun Lu, and Arie Verhagen, 13–40. Berlin/Boston: De Gruyter Mouton. ). Not only are the grammatical structures involved in reported speech unlike any other complex clause types in many languages (McGregor 1994McGregor, William B. 1994 “The Grammar of Reported Speech and Thought in Gooniyandi.” Australian Journal of Linguistics 14 (1): 63–92. ), reported speech frequently involves no apparent marking at all. Consider the frequently discussed category of ‘free indirect speech’, exemplified in (1) (also cf. Vandelanotte, this volume and Van Duijn & Verhagen, this volume).
It was too blooming dull (Joyce, cited in Banfield 1982 1982 Unspeakable Sentences: Narration and Representation in the Language of Fiction. Boston: Routledge & Kegan Paul., 186)
Structurally, none of the elements in (1) signal a shift in perspective from the current speaker to a reported one. Yet, semantically, with (1) the narrator attributes the proposition ‘it is too blooming dull’ to another speaker. Although this shift is not signalled with a speech clause as in the direct speech construction in (2) or the indirect speech construction in (3), semantically (1)–(3) all belong to the class of reported speech expressions.
He said: “It is too blooming dull”
He said that it was too dull
The semantic complexity of the perspective shift involved in reported speech apparently makes the lack of structural marking all the more remarkable (cf. Sharvit 2008Sharvit, Yael 2008 “The Puzzle of Free Indirect Discourse.” Linguistics & Philosophy 31: 353–395. ). However, it has become increasingly clear that reported speech in fact frequently involves un(der)marking or irregular marking in the languages of the world. In recent years, the traditional opposition between direct-indirect speech, from which notions such as ‘free indirect speech’ derive, has been increasingly challenged by studies of reported speech in (unwritten) under-described languages (Aikhenvald 2008 2008 “Semidirect Speech in Manambu.” Language Sciences 30: 383–422. ; Munro et al. 2012Munro, Robert, Rainer Ludwig, Uli Sauerland, and David W. Fleck 2012 “Reported Speech in Matses: Perspective Persistence in Evidential Narratives.” International Journal of American Linguistics 78 (1): 41–75. ; also see De Roeck 1994Roeck, Marijke De 1994 “A Functional Typology of Speech Reports.” In Function and Expression in Universal Grammar, ed. by E. Engberg-Pedersen, L. Jakobson, and S. Rasmussen, 331–351. Berlin/New York: De Gruyter Mouton.). Evans (2013) 2013 “Some Problems in the Typology of Quotation: A Canonical Approach.” In Canonical Morphology and Syntax, ed. by Dunstan Brown, Marina Chumakina, and Greville G. Corbett, 66–98. Oxford etc.: Oxford University Press. even characterises direct speech, such as (2), in which the perspective in the reported clause shifts entirely to that of the reported speaker (i.e. the referent of the grammatical subject of the matrix clause) and indirect speech, as in (3), where the perspective remains with the current speaker, as relatively exceptional cases. Languages often employ intermediate structures that do not fully represent either type and signal aspects of both the current and reported speaker’s perspective (cf. also Vandelanotte, Van Duijn and Verhagen, Zeman, this volume), a category Evans calls bi-perspectival speech. The free indirect speech sentence in (1) could be seen as an example of this type, with the evaluation ‘it is dull’ being formulated from the perspective of the current speaker, as signalled by the past tense verb and the expletive ‘blooming’ directly echoing an utterance by the reported speaker (also see Banfield 1982 1982 Unspeakable Sentences: Narration and Representation in the Language of Fiction. Boston: Routledge & Kegan Paul.).
The phenomenon of unmarked or undermarked reported speech has been mostly studied for written, European languages and is commonly seen as a literary device, a claim going back to Vološinov (1973)Vološinov, Valentin N. 1973 Marxism and the Philosophy of Language. Ed. by Ladislav Matejka and I. R. Titunik. New York/London: Seminar Press.. While analyses of underdescribed languages have demonstrated that the wealth and variety of structures involved in reported speech is considerably greater than the traditional direct-indirect speech opposition, the general impression that free indirect speech belongs to the realm of written language has remained. This impression is unwarranted, as we aim to demonstrate in the present study. In this paper we introduce newly recorded examples of reported speech from the Indian Indigenous language Solega (Dravidian), such as in (4).
‘You arrive and start clanging the pots around, and your children are sitting here looking at you’
There’s one sitting there, “When will they give it (food) to us?
O Shiva! When will our father give us (something to eat)?”
Here their intestines are twisting around (from hunger), aren’t they?’ (KG, BasaveGConv_2010-10-19, 00:21:54)
Like other languages, Solega displays a range of strategies for expressing reported speech, including the type shown in (4), which involves no apparent marking.22.Note that the quotation marks have been added in the transcription: Solega does not have a written tradition. For details about the prosodic features of Solega reported speech, please see below. The perspective shift between the narrator and the hypothetical speech of the children introduced in the first sentence in ‘“eshṭottiga ada koṭṭa:valla? namma appanavaru eshṭottiga koṭṭranallavalla shivane?” ‘“When will he give us some? O Shiva! When will our father give us something to eat?”’, is not accompanied by a matrix speech clause. Such examples of ‘free direct speech’ or ‘zero quotatives’ (Mathis and Yule 1994Mathis, Terrie, and George Yule 1994 “Zero quotatives.” Discourse Processes 18 (1): 63–76. ) are frequent in the Solega corpus.
Our aims in this paper are three-fold. First, we attempt to provide a descriptive account of narrative perspective shift in Solega, a previously undescribed Indigenous language of India (but see Si 2016Si, Aung 2016 The Traditional Ecological Knowledge of the Solega: A Linguistic Perspective. Dordrecht: Springer. ). Second, we aim to present a new account of un(der)marking in reported speech, which does not focus on the marking of perspective shifts (i.e. the indexing devices that are there), but approaches it as a problem of grammatical optionality. For reasons to be explained below, we label this phenomenon ‘defenestration’ (Spronck 2017 2017 “Defenestration: Deconstructing the Frame-In Relation in Ungarinyin.” Journal of Pragmatics 114: 104–133. ). Finally, our third goal in this article is to develop an approach to defenestration that allows the notion to be applied cross-linguistically as a method for analysing structural optionality in the expression of reported speech.
The paper is structured as follows: Section 2.1 briefly situates the Solega language typologically, geographically and culturally, and 2.2 introduces the data examined in the present paper. In Section 3 we introduce the various ways in which Solega expresses reported speech, identifying its general features (3.1) and the attested construction types with explicit marking (3.2), and instances in which marking is apparently absent (3.3). Section 4 proposes a basic semantic representation for reported speech and Section 5 discusses the defenestration approach and applies it to the observations of the Solega data. Section 6 offers a brief conclusion.
2.Field site and methods
2.1Solega speakers
The Five-Clan Solega (also Soliga, Sholaga) are a ‘Scheduled Tribe’ of India, whose traditional lands are to be found in the southern Indian state of Karnataka, in what is now the Biligiri Rangaswamy Hills Tiger Reserve. With an estimated population of 20,000–30,000, the Solega speak a Dravidian language that is closely related to the official State language, Kannada. Until recently, they were hunter-gatherers, but they now live in permanent settlements due to government pressure. The following data were collected during the course of a documentation of Solega ethnobiological knowledge between 2008–2012 (see Si 2016Si, Aung 2016 The Traditional Ecological Knowledge of the Solega: A Linguistic Perspective. Dordrecht: Springer. ).
2.2Data
The data in this paper are based on four recordings (three audio and one video) involving three male speakers, and with a combined duration of around 80 minutes. As can be seen from Table 1, the texts cover a range of speech genres, from personal reminiscences through to mythological narratives. Recordings 1 and 2 involved speech directed at a Solega field assistant who was asked to engage with the speaker. These recordings were made at a time when the first author did not have the proficiency in Solega to interview the speaker personally, and was content to operate the audio recorder, while his assistant responded to the speaker’s narrative with the appropriate cues. Recording 3 was made by a non-Solega colleague of the first author, who was accompanied by Solega field assistants, but spoke directly to speaker DG in Kannada. Finally, recording 4 was made by a Solega field assistant working independently, and operating a video camera while speaker BG and his brother engaged in conversation.
| Recording title | Description | Speakers | Duration |
|---|---|---|---|
| 1. Genasu | Mythological narrative | BG | 4min 27sec |
| 2. Forest | Personal narrative, descriptions of cultural practices, folk tales | BG | ~35min |
| 3. Savana | Mythological narrative | DG | ~ 13min (extract from a 45min-long recording) |
| 4. Conversation | Conversation between two brothers, current events, reminiscences about traditional life | BG and KG | ~28min |
3.Solega reported speech
3.1Solega reported speech (RS) practices
Based on the corpus described in Table 1 we extracted all instances of speech and thought representation, regardless of how they were structurally marked. Table 2 shows that of the total number of instances of reported speech (93) 68 (i.e. 73%) did not co-occur with a clausal expression marking it as reported speech.
| Matrix clauses | # |
|---|---|
| In pre-position | 8 |
| In post-position only | 17 |
| Reported clauses without matrix | 68 |
| Total no. of examples of RS in the corpus | 93 |
Solega does offer a clear and stable strategy for marking reported speech, however. Before addressing the unmarked cases, we would first like to review those instances in which reported speech is marked most explicitly in Solega.
3.2The canonical RS construction and its variations
A full canonical reported speech construction makes use of one or more matrix clauses incorporating the verbs he:ḷu or ennu ‘say’. The matrix clause(s) may occur before, after, or on either side of a stretch of reported speech, as illustrated by Examples (5)–(7).
(So) what did Madappa go and say (to Savana)? “Oh, greetings, my Lord!”
“Oh, it’s a good omen! Look, my lord!” he said.
That bird says its own name, “kuṭru-kuṭru”.
The ‘say’ verb can carry a non-finite inflection, as illustrated by Example (8), where the speaker reports a claim that government officials frequently make about the Solega. Here, the verb ennu ‘say’ is marked with a GERund suffix, and behaves as a kind of complementiser. Typically, ennadu ‘saying’ appears in constructions which take the form TOPIC-ennadu-COMMENT. In the present corpus, ennadu behaves quite differently from the canonical, more standard complementiser endu (see Examples (9), (11), (12)), as the latter is never followed by a comment on the RS that precedes it. Example (9) shows a common strategy where both ‘say’ verbs appear in the matrix clause that follows a report. In this construction, the first verb ennu invariably appears in a different non-finite construction, in combination with the CONVerb suffix to give rise to the complementiser endu. The second verb he:ḷu ‘say’ forms the head of the matrix clause, and appears fully inflected for tense, gender and number.
So what is the governmentx doing to our peopley? (Theirx) saying “They’rey eating the animals! They’rey eating the animals!” is a lie.
“You have to bring me a woman every day!” he said to Rangappa.
Speakers also have the choice of bracketing a section of RS between two matrix clauses containing ‘say’ verbs, as in Example (10). The ‘say’ verbs are shown in bold, as (10) is a rather long and complex example. The verb he:ḷu in the first matrix clause occurs in a non-finite construction (similar to ennadu ‘saying’ in (8)). Here, he:ḷidadu ‘saying’ occurs in gerund form, and is another instance of insubordination. Ennu ‘say’ and he:ḷu ‘say’ are both present in the clause following the report, but the former appears in a Past Participle construction (enna), and qualifies the noun lekka ‘account’. The adverb ha:ge ‘thus’ frequently occurs phonologically bound to ennu ‘say’, as can be seen in both (9) and (10). Finally, in the second matrix clause of (10), the verb he:ḷinã ‘he said’ takes the full set of inflections as the head of the clause.
Now look here, this blacksmith, Acugã, he said, “I completely avoided it [western medicine], I haven’t taken any pills from the time I was born, I did everything with just traditional medicines. There’s (something) for headache, for stomach ache, we have something that works for absolutely every (condition).” That was the kind of account he gave.
Instances of matrix clauses only occurring before a report (5) are rare in our corpus, as the majority tend to be placed following the RS. Note that in (5) the object slot of the ‘say’ verb he:ḷu is occupied by the interrogative e:na ‘what’, so that the matrix clause resembles a rhetorical question that is answered by the speaker himself. In this instance it could be argued that there is no strict syntactic relation between the speech clause and the following reported clause. (6) illustrates the simplest type of construction with a following matrix clause, while (7) is a rare example of a report embedded within a matrix.
Finally, we have also classified, as matrix clauses, those instances which make use of certain non-speech verbs that nevertheless strongly imply a reported speech or thought event. These include the verbs kare ‘call’, ke:ḷu ‘ask’, oppu ‘agree’, ma:tukate ma:ḍu ‘have a chat’, ma:ta:ku ‘make a plan’ or hesara kaṭṭu ‘name’. Most instances of the use of such verbs also involve the use of the ennu ‘speak’-derived complementiser endu, as in (11) and (12).
After that happened, he agreed, saying, “Fine!”
He hailed and called everyone, saying, “Everybody come here!”
In summary, the four sentential reported speech constructions in Solega could be schematically represented as in (13).
| a. | [ (SAY) | [ | ] | SAY ] |
| b. | [ (SAY) | [ | ] | ha:ge ‘thus’ SAY ] |
| c. | [ (SAY) | [ | ] | enn-adu ‘saying’ ] |
| d. | [ (SAY) | [ | ] | endu ‘comp’ (V) ] |
| (Matrix) | Reported | Matrix |
Given the weak empirical evidence for pre-posed matrix clauses in the data, we consider matrix clauses following a reported message primary, although the round brackets in first position in (13) indicate that Matrix clauses are not restricted to final sentence position.33.This analytical position does mean, however, that we suggest that a pre-posed ‘matrix’ clause without a post-posed matrix clause is not an example of a fully specified syntactic reported speech construction in the Solega. The bracketed V position in the Matrix clause allows for a range of speech and thought verbs, but need not necessarily be filled, given that the complementiser/quotative forms enn-adu and endu derive from the lexical verb ‘say’ (as its GERund or CONVerb form, respectively), and neither form is fully grammaticalised, particularly in the case of enn-adu (‘say-ger’).
It is likely that the constructions in (13) display subtle semantic differences, for example in the way in which they provide epistemic access to the reported content or allow for modal evaluation of this content. Such subtle meaning differences between, e.g., direct/indirect speech constructions and their variations are common cross-linguistically (Spronck and Nikitina 2019Spronck, Stef, and Tatiana Nikitina 2019 “Reported Speech Constructions Are a Dedicated Syntactic Domain: Typological Arguments and Observations.” LinguisticTypology 23: 1. ; Spronck 2012Spronck, Stef 2012 “Minds Divided: Speaker Attitudes in Quotatives.” In Quotatives: Cross-Linguistic and Cross-Disciplinary Perspectives, ed. by Isabelle Buchstaller and Ingrid Van Alphen, 71–116. Amsterdam/Philadelphia: John Benjamins. ). However, in this study we will focus on a reported speech type that does not involve any of the constructions in (13), at least not in instantiations in which all positions are filled. These cases of un(der)specified reported speech are frequent in the data, as Table 2 demonstrates, and we will analyse these in the remainder of this paper.
3.3Un(der)specified reported speech
It was mentioned in passing in Si (2016Si, Aung 2016 The Traditional Ecological Knowledge of the Solega: A Linguistic Perspective. Dordrecht: Springer. , 155) that the speaker BG often used the following discursive strategy in his descriptions of events, processes or entities: he would incorporate into his description a mini-conversation between himself and an unspecified addressee or between two (or more) unspecified interlocutors. BG would usually launch into these mini-conversations without any overt indication, in the immediately preceding narrative, that a section of reported speech was to follow. Below is an example of one such instance of reported speech, produced during an elicited description of a landscape term, oḍḍuga:ḍu ‘boulder field’.
Consider Aḍkugallu [place name], and E:ru kallu [place name], the rocks in these places are piled up; that’s what an oḍḍuga:ḍu or kalloḍḍu is like. Lots of rocks piled on top of one another. There are bears there, tigers also live in those places. The place where they live is the oḍḍu. “You shouldn’t go to the oḍḍu ! There are bears there!” (people would say.)
Lacking any explicit morphological marking, the reported speech sections of (14) can only be distinguished from the preceding narrative by means of the prosodic marking imparted by the current speaker, as he imitates the admonishments of the reported speaker (presumably an elderly person talking to a child). The final two clauses shown in Figure 1, which represent the reported speech segment, can be distinguished by their generally raised pitch as well as unusual boundary tones towards the end of each clause. Clause 3, for instance, is characterized by a sharply rising pitch contour, while clause 4 ends with a sharp fall. In contrast, the ‘narrative’ clauses 1 and 2 exhibit gradual downdrift as expected.
Prosodic contrast is undoubtedly an important correlate of underspecified reported speech constructions in Solega, but is not sufficient to explain why the phenomenon of underspecified reported speech occurs, and with the frequency with which it does. None of the patterns of rising intonation are unique to reported speech environments, and imitating prosody does not preclude the expression of a Matrix clause as well. Within the current paper we intend to develop a different approach. We propose that rather than pursuing a marking-based analysis, focusing on potential alternative ways of structurally and contextually ‘signalling’ a perspective shift, the absence of marking in Solega reported speech in narratives would be more fruitfully analysed through the prism of grammatical optionality (McGregor 2013 2013 “Optionality in Grammar and Language Use.” Linguistics 51 (6): 1147–1204. ; Spronck 2017 2017 “Defenestration: Deconstructing the Frame-In Relation in Ungarinyin.” Journal of Pragmatics 114: 104–133. ). In order to demonstrate this, we will first need to provide a fuller semantic account of reported speech and the structural elements involved in marking this meaning. We will subsequently examine what conditions allow aspects of this semantic representation to remain unexpressed.
4.Semantics of reported speech syntax
What type of a syntactic structure is a reported speech construction? As McGregor (1994)McGregor, William B. 1994 “The Grammar of Reported Speech and Thought in Gooniyandi.” Australian Journal of Linguistics 14 (1): 63–92. points out, while many languages ostensibly use regular coordinated or subordinated structures for marking reported speech, on closer inspection few reported speech constructions display features that are commonly associated with coordination or subordination. For example, matrix clauses in reported speech can often occur on either side of reported clauses, which is uncharacteristic for main clauses in subordinating structures, and the reported clause typically introduces prominent information, which is unexpected for both coordinated sentences (in which the information status of both clauses can be assumed to be equal) and subordinate clauses (which typically provide backgrounded information; for a fuller list of arguments for why reported speech constructions are exceptional, see Spronck 2017 2017 “Defenestration: Deconstructing the Frame-In Relation in Ungarinyin.” Journal of Pragmatics 114: 104–133. , 106–107). McGregor (1994)McGregor, William B. 1994 “The Grammar of Reported Speech and Thought in Gooniyandi.” Australian Journal of Linguistics 14 (1): 63–92. concludes on the basis of these observations that reported speech constructions involve a syntactic relation in its own right, which he labels a ‘framing’ construction.
We define the semantics of a framing construction as in (15) (also see Spronck and Nikitina 2019Spronck, Stef, and Tatiana Nikitina 2019 “Reported Speech Constructions Are a Dedicated Syntactic Domain: Typological Arguments and Observations.” LinguisticTypology 23: 1. ).
A framing construction minimally includes three meaning components:
-
Evidentiality (reflects a deictic relation between two events: the alleged original situation of discourse production and the current speech moment);
-
Semiotic exceptionality (qualifies the reported message as a representation of an utterance, as demonstrated discourse);
-
Modality (reflects an epistemic evaluation of the represented utterance, or, more accurately, cancels implicatures that the current speaker commits to the truth of the discourse, allowing for epistemic qualification of the reported message, both by strengthening or weakening truth commitment).
We may illustrate the semantic properties of reported speech syntax described in (15) on the basis of the Solega reported speech construction in (6), repeated below as (16), overlaid with the schematic representation of the construction through square brackets, as in (13). The labels for the Reported clause (R) and the Matrix clause (M) appear in subscript.
“Oh, it’s a good omen! Look, my lord!” he said.
The evidential meaning as intended in (15) can be gleaned from (16) from the deictic relation the sentence sets up between the current speech event, with respect to which the Matrix avã he:ḷ-in-ã ‘he said’ is deictically anchored, and the speech event to which the Reported utterance o: oḷḷe sakuna no:ḍ-i buddi ‘Oh, it’s a good omen! Look, my lord!’ is attributed. The interpretation of evidentiality we follow here goes back to that in Jakobson (1957)Jakobson, Roman 1957 Shifters, Verbal Categories and the Russian verb. Department of Slavic Languages and Literatures, Cambridge, MA: Harvard University., and slightly differs in focus from the more generally cited definition of evidentiality as signalling ‘source of information’ (Aikhenvald 2004Aikhenvald, Alexandra Y. 2004 Evidentiality. Oxford etc.: Oxford University Press.). There is no fundamental distinction between these two interpretations, however: defining evidentiality as signalling a source of information suggests a perception event in which the current speaker witnessed the information described in the proposition. The information from the perception event (in the case of reported speech, a conversation), is brought into the current speech event, which sets up a contrast between the current speech event and the perception event. Jakobson’s (1957)Jakobson, Roman 1957 Shifters, Verbal Categories and the Russian verb. Department of Slavic Languages and Literatures, Cambridge, MA: Harvard University. definition addresses the deictic relation between these two events, rather than solely characterising evidentiality around the notion of a ‘perception’ event, i.e. an information ‘source’, but both definitions naturally imply each other.55.For further discussion of this interpretation relating to the meaning of evidentiality, see Spronck (2015) 2015 “Stance as Participant Structure: A Jakobsonian Approach to the Pragmatics and Semantics of Evidentiality.” Belgian Journal of Linguistics 29: 193–216. .
What Jakobson’s (1957)Jakobson, Roman 1957 Shifters, Verbal Categories and the Russian verb. Department of Slavic Languages and Literatures, Cambridge, MA: Harvard University. definition allows us to do, however, is to explore the relation between evidential meanings and other ‘shifter’ categories, which can be done through the notion of deixis. The deictic relation in the evidential meaning of reported speech is supported by tense (past tense in m, tenseless imperative in R) and pronouns (third person, as the reported speaker is in a third-person relation to the current speaker in m, and an address term from the perspective of the reported speaker buddi ‘my lord’ in R). However, these deictic expressions do not themselves constitute the evidential meaning: in our interpretation, the evidential meaning as specified in (15a) is an inherent semantic component of the framing construction. Given that this meaning is based on a deictic relation (also see Vandelanotte 2004Vandelanotte, Lieven 2004 “Deixis and Grounding in Speech and Thought Representation.” Journal of Pragmatics 36 (3): 489–520. ), it is expected that deictic categories in reported speech adjust to this deictic environment in accordance with the grammar of the respective language (also see Nikitina 2012Nikitina, Tatiana 2012 “Personal Deixis in Reported Discourse: Towards a Typology of Person Alignment.” Linguistic Typology 16 (2): 233–263.).
The second semantic component of a framing construction, semiotic exceptionality, is shown in (16) by the fact that R is interpreted, at least on some level, as a single unit. The entire sentence o: oḷḷe sakuna no:ḍ-i buddi ‘Oh, it’s a good omen! Look, my lord!’ is interpreted as a representation of the reported message as a whole. In other words, it is being ‘demonstrated’ in the current speech situation (Clark and Gerrig 1990Clark, Herbert H., and Richard J. Gerrig 1990 “Quotations as Demonstrations.” Language 66: 764–805. ). Linguistic reflexes of this semiotic status can be prosodic signals of prominence and the interpretation of R as a unit that allows limited modification by the current speaker, which we suggest both apply to the Solega example in (16). While our interpretation of ‘demonstration’ derives from Clark and Gerrig (1990)Clark, Herbert H., and Richard J. Gerrig 1990 “Quotations as Demonstrations.” Language 66: 764–805. and Clark (2016)Clark, Herbert H. 2016 “Depicting as a Method of Communication.” Psychological Review 123 (3): 324–347. , our interpretation is slightly more restricted in the sense that we apply the notion of ‘demonstrated’ entirely to the semiotic status of the reported utterance (closer to ‘iconic’ in the sense of Recanati 2001Recanati, François 2001 “Open Quotation.” Mind 110 (439): 637–687. ), and not to the act of demonstrating itself. On the other hand, we apply the term to a broader range of phenomena, in the sense that we potentially allow for degrees in which an utterance can be demonstrated, not just as ostensibly ‘as-is’/‘verbatim’ in direct speech constructions, but also filtered through the perspective of the current speaker in indirect speech (McGregor 1994McGregor, William B. 1994 “The Grammar of Reported Speech and Thought in Gooniyandi.” Australian Journal of Linguistics 14 (1): 63–92. , 1997 1997 Semiotic Grammar. Oxford: Clarendon Press.) and bi-perspectival speech (Evans 2013 2013 “Some Problems in the Typology of Quotation: A Canonical Approach.” In Canonical Morphology and Syntax, ed. by Dunstan Brown, Marina Chumakina, and Greville G. Corbett, 66–98. Oxford etc.: Oxford University Press.).66.Cf. also Vandelanotte (this volume) and Van Duijn and Verhagen (this volume). Again, none of the grammatical phenomena associated with the semiotic status of R explicitly mark semiotic exceptionality. Within our approach, the demonstrated nature of R is a conventional aspect of the framing construction, that more naturally affects linguistic signals that deal with syntactic ‘unithood’ or prominence, such as prosody or clause boundaries.
Finally, the modal meaning involved in framing constructions comes out in (16) through the interpretation of the degree of commitment by the narrator with respect to the reported message in R. While R contains evaluative elements, such as exclamative intonation and the adjective ‘good’, these are not interpreted with respect to the current speaker. In this sense there is a modal distance between R and the narrator. Reported speech is an example of ‘non-serious’ language use, as pointed out by Goffman (1981)Goffman, Erving 1981 Forms of Talk. Philadelphia: University of Pennsylvania Press.: the speaker does not assert the content of R. On the other hand, while the current speaker need not agree with or commit to the meaning of R, s/he can usually modulate the credibility of R by choosing particular speech verbs, referring to a highly authoritative reported speaker, or even adding further wide-scope modal markers (Evans 2006Evans, Nicholas 2006 “A View with a View: Towards a Typology of Multiple Perspective Constructions.” In Proceedings of the thirty-first annual meeting of the Berkeley Linguistics Society, ed. by Rebecca T. Cover and Yuni Kim, 93–120. Berkeley: Berkeley Linguistics Society.; Spronck 2012Spronck, Stef 2012 “Minds Divided: Speaker Attitudes in Quotatives.” In Quotatives: Cross-Linguistic and Cross-Disciplinary Perspectives, ed. by Isabelle Buchstaller and Ingrid Van Alphen, 71–116. Amsterdam/Philadelphia: John Benjamins. ).
Having laid out the semantic building blocks of a ‘full’ reported speech construction, how can we qualify ‘incomplete’ reported speech constructions, i.e. Solega examples such as in (14), in which the perspective shift to the reported speaker remains unspecified? We propose that as grammatical constructions, the only component missing in such examples of un(der)specified perspective shift is an explicit M-clause. With the definition of a framing construction presented in this section, we can now qualify constructions as in (14) as a framing construction without full M-marking.
The resulting interpretation of R-clauses without M-clauses is not dissimilar in form to subordinate clauses without a main clause in ‘insubordination’ (Evans 2007 2007 “Insubordination and its uses.” In Finiteness: Theoretical and Empirical Foundations, edited by Irina Nikolaeva, 366–431. Oxford etc.: Oxford University Press.). Since in our analysis framing constructions do not (necessarily) involve subordination to begin with, the term ‘insubordinate clauses’ would be inappropriate for Matrix-less reported speech constructions. Instead, we label these framing clauses without a Matrix-clause ‘defenestrated clauses’ (Spronck 2017 2017 “Defenestration: Deconstructing the Frame-In Relation in Ungarinyin.” Journal of Pragmatics 114: 104–133. ). We use the term ‘defenestration’ both for the process of defenestration, i.e. ‘leaving out’ M-clauses in online conversation, and for the resulting class of expressions, i.e. defenestrated clauses. Here we will exclusively focus on the latter interpretation, however, more particularly, on the class of defenestrated expressions in Solega.
5.Defenestration
What do Solega M-clauses mark? As the examples of Solega framing constructions in (5)–(12) demonstrate, M-clauses in Solega (and in many other languages) have two main functions: first, they provide the lexical and pronominal referential expressions that allow the current addressee to identify the reported speaker, who is commonly marked as the subject of the M-clause, and/or the reported addressee. Second, they specify lexical aspects of the reported speech situation, such as marking it as a speech or thought event, or characterising a manner of speaking. Consider the list of frequent item types in Matrix, Non-matrix preceding and Reported clauses, as summarised in Table 3.
| CLAUSE TYPE | ITEMS |
|---|---|
| Matrix | speech event verbs (he:ḷu ‘say’, ennu ‘say’, kare ‘call’, ke:ḷu ‘ask’, oppu ‘agree’, ma:tukate ma:ḍu ‘have a chat’, ma:ta:ku ‘make a plan’ or hesara kaṭṭu ‘name’); endu ‘comp’; reported-addressee identity; reported-speaker identity |
| Preceding, non-matrix | motion verbs (ho:gu ‘go’, baru ‘come’ se:ru ‘approach’); non-finite constructions (converb suffixes -(t/d)i/u ‘having done X’, -ru ‘while doing X’, continuous suffix -ta ‘X-ing’); temporal adverbs or adverbial suffixes (a:ga ‘then’, me:le ‘after’, ṭaiminalli ‘at the time of’, -da:ge ‘as soon as’) |
| Reported | kinterm (usually directed at addressee: appa ‘father’, aṇṇa ‘brother’, etc.); pronouns (na: ‘1sg’, ni: ‘2sg’, ella(:ruve) ‘all/everybody’); other address terms (buddi ‘Lord’, proper name, titles); interjections (o:(ho)! ‘Oh!’, aiyo! ‘Oh no!’, e:! ‘Hey!’ ḍou! ‘Hey!’); illocutionary change to interrogative, imperative, hortative; discourse markers (a:tu ‘fine/OK’, sari ‘fine/OK’, hã ‘yes’, ille ‘no’, koṇo/kopo ‘EMPH’, matte ‘so/then’, =ḍou ‘EMPH’) |
The lexical items appearing in the Matrix-clause according to Table 3 are exactly the type of elements that can serve the two main functions of M-clauses: specifying types of speech event and specifying participants in the reported speech event. The column also includes the SAY-derived complementiser endu, which does not clearly belong to either the M- or the R-clause, but serves as an explicit marker of a subtype of the framing construction.
With these observations in mind, we are now in a position to characterise the meaning of Matrix-less reported speech clauses in more precise terms: none of the functions performed by M-clauses correspond to the semantic core components of framing constructions as defined in (15). Therefore, the framing meaning of Matrix-less reported speech clauses is fully intact. In order to be interpretable, defenestrated clauses need to be interpreted as carrying an evidential meaning, semiotic exceptionality meaning and modal meaning as discussed, and these constitute the perspective shifting meaning of reported speech constructions. While M-clauses, as fuller instantiations of framing constructions, help identify a structure as a framing construction, they are not a necessary component in order to constitute any of these meanings and they may be omitted if the speaker judges that omission of the M-clause does not jeopardise recognition of the framing construction. In our analysis, Solega defenestrated clauses are framing constructions in which the M-clause is treated as an optional element.
While we do not claim that any of the aspects of our analysis is entirely novel, our analysis departs in important ways from more traditional analyses and several recent accounts of un(der)marking perspective shift. Our semantic definition of framing relies heavily on earlier semantic descriptions of reported speech (see Spronck 2017 2017 “Defenestration: Deconstructing the Frame-In Relation in Ungarinyin.” Journal of Pragmatics 114: 104–133. for a fuller literature review), and the approach to optionality is primarily indebted to McGregor (2013) 2013 “Optionality in Grammar and Language Use.” Linguistics 51 (6): 1147–1204. , but it provides an alternative to a recently popular school of thought that situates the meaning of reported speech outside ‘core’ grammar (Blackwell, Perlman, and Tree 2015Blackwell, Natalia L., Marcus Perlman, and Jean E. Fox Tree 2015 “Quotation as a Multimodal Construction.” Journal of Pragmatics 81: 1–7. ; Dancygier 2016Dancygier, Barbara 2016 “Concluding Remarks: Why Viewpoint Matters.” In Viewpoint and the Fabric of Meaning: Form and Use of Viewpoint Tools across Languages and Modalities, ed. by Barbara Dancygier, Wei-lun Lu, and Arie Verhagen, 281–288. Berlin/Boston: De Gruyter Mouton. ; D’Arcy 2015D’Arcy, Alexandra 2015 “Quotation and Advances in Understanding Syntactic Systems.” Annual Review of Linguistics 1 (1): 43–61. ; Lampert 2013Lampert, Martina 2013 “Cognitive Semantics Goes Multimodal: Looking at Quot(ativ)es in Face-to-face-settings.” International Journal of Cognitive Linguistics 4 (2): 103–132. http://search.proquest.com/docview/1625137999?accountid=8330; Stec, Huiskes, and Redeker 2015Stec, Kashmiri, Mike Huiskes, and Gisela Redeker 2015 “A Multimodal Analysis of Quotation in Oral Narratives.” OpenLinguistics 1: 531–554.). Faced with common examples, such as the free indirect speech construction in (1), or the frequency and regularity of suprasegmental and extra-linguistic signals in the expression of reported speech, such as eye-gaze or voice quality, these accounts arrive at the conclusion that reported speech cannot be characterised at a morpho-syntactic level alone. Authors within this approach analyse reported speech as a ‘multi-modal’ construction, in the sense of being constructed through, e.g., oral, visual and textual modalities (cf. also Vandelanotte, this volume for discussion). Although this proposal seemingly breaks new ground in opening up the definition of what belongs to grammar and linguistic meaning making (D’Arcy 2015D’Arcy, Alexandra 2015 “Quotation and Advances in Understanding Syntactic Systems.” Annual Review of Linguistics 1 (1): 43–61. ), the underlying motivation behind it is in fact conservative: the approach emerges from the assumption that if morphosyntactic marking underspecifies a semantic meaning (and perhaps especially a meaning as seemingly complex as perspective shift) this must be compensated for elsewhere in the communicative signal. We acknowledge that the multi-modal approach to underspecified perspective shift has unearthed interesting and valuable communicative practises and presents exiting challenges to syntacticians who insufficiently engage with interactional data. We also believe, however, that it assumes a wrong causality between multi-modality and morpho-syntactic marking. Faced with the evidence that reported speech is accompanied by multimodal signals, and that these signals are present both when reported speech receives a typical morphosyntactic marking, the multimodal approach suggests that all signals must be part of a multifaceted multimodal construction. Although gestures, eye-gaze, voice quality etc. occur in patterns, most of these are actually not that clearly conventionalised (Stec, Huiskes, and Redeker 2015Stec, Kashmiri, Mike Huiskes, and Gisela Redeker 2015 “A Multimodal Analysis of Quotation in Oral Narratives.” OpenLinguistics 1: 531–554.), and such observations are puzzling if the assumption is that linguistic semantic interpretation relies on them. The underlying assumption, that a meaning must be marked somehow, combined with the finding that marking is imperfectly conventionalised, then leads to the dramatic conclusion that grammar itself is an unreliable indicator of linguistic meaning, in reported speech and, possibly, beyond.
The optionality-oriented approach we advocate here provides a much simpler explanation for the observed patterns: if the ‘loosely’ conventionalised patterns do not compensate for morphosyntactic marking, but provide possible contexts in which the expression of an M-clause can become optional, it is expected that there is variation in both the degree to which a speaker displays regular multimodal patterns, and the actual omission of M-clauses in relation to these patterns.
If our analysis is on the right track, we need to provide evidence that our semantic definition of framing provides an encompassing meaning description for both fully specified framing clauses and defenestrated clauses. We supplied such an analysis for the function of M-clauses in relation to the core semantics of reported speech, as specified in (15) above, but will now turn to the elements in Table 2 that potentially remain expressed in defenestrated clauses: those in Reported-clause preceding position, and those within the Reported-clauses themselves.
What is striking about the collection of ‘Preceding’ elements in Table 2 is that they almost exclusively relate to movement actions or temporal positioning. This is consistent with the class of elements Verstraete (2011)Verstraete, Jean-Christophe 2011 “The Functions of Represented Speech and Thought in Umpithamu Narratives.” Australian Journal of Linguistics 31 (4): 491–517. calls ‘perspectivising clauses’ in the Australian Aboriginal language Umpithamu. Spronck (2017) 2017 “Defenestration: Deconstructing the Frame-In Relation in Ungarinyin.” Journal of Pragmatics 114: 104–133. finds a similar class of predicates in this position in the unrelated Australian language Ungarinyin, and Sotirova (2004)Sotirova, Violeta 2004 “Connectives in Free Indirect Style: Continuity or Shift?” Language Literature 13 (3): 216–234. discusses the role of discourse connectives in free indirect speech. Example (17) below shows an instance where both motion and temporal positioning are combined in the word banda:ga ‘when SUB came/comes’.
Yes, so when everyone came, “What are your offerings?”
A marking-centric approach would conclude on the basis of such observations, that there is a marking pattern associated with, e.g., movement verbs and that they therefore signal a perspective shift when they occur right before instances in which a shift in perspective is required. The consequences of this analysis are highly problematic: movement verbs and temporal expressions occur throughout a narrative, not just in reported speech preceding position, and the mechanism through which they gain perspective-shifting meanings is mysterious. Again, the defenestration approach provides a simple alternative: the relevant movement clauses are typically co-indexical with the reported speaker, which helps identify the speaker, reducing the need for an explicit M-clause, but not substituting it. With their event meanings, movement and temporal expressions help to identify the evidential meaning of the framing construction/defenestrated clause, and they may highlight the initial boundary of the Reported clause, thereby facilitating its interpretation as a semiotic unit. While the class of preceding elements can therefore be expected to form a relatively regular set, and display loose patterning, we can acknowledge this under a defenestration approach, without lending it a similar marking status as a full framing clause.
The elements found within Reported clauses, both within a full framing construction and under defenestration, we believe, further supports our approach. The elements we find here include address terms and local pronouns, expressive elements such as interjections and a phenomenon we have labelled ‘illocutionary change’, displaying an illocutionary contrast with the immediately preceding clause, mostly of clearly interactional type,77.Cf. also Sandler & Pascual (this volume) for examples that would fit this analysis. such as an imperative or an interrogative immediately following, e.g., a statement. (18) illustrates the use of the discourse marker ‘so’ and the clear signalling of a defenestrated Reported clause by means of an illocutionary change to imperative and hortative mood. Again, the (apparent) clustering of such elements in Reported clauses is not restricted to Solega. For example, Eckhardt (2012)Eckhardt, Regine 2012 “Particles as Speaker Indexicals in Free Indirect Discourse.” Linguistische Datenverarbeitung, ed. by Lotte Hogeweg and Eric McCready and Grégoire Winterstein, (35–36): 99–109. discusses the dominance of indexical elements in reported speech, and individual grammarians have frequently observed that interjections tend to show up at the boundary of R-clauses.
You hammer it [a yam stick], make a proper point, hold it in your hand, and, “So, get moving! Let’s go into this boulder field!”
These elements are not restricted to defenestrated R-clauses, but are found in framing constructions in general, as Figure 2 shows.
The pairs of bars in Figure 2 show the proportion of full framing constructions and defenestrated constructions which have a particular item. The percentages add up to more than 100%, because multiple items can be present in a single construction. What is most relevant for our analysis is that the fully framed R-clauses and defenestrated clauses overall show very similar distributions of elements. Low-frequency items such as motion/position and perception verbs, non-finite verbs, temporal expressions, the complementiser endu, address and kin terms all appear in roughly equal proportions in both framed R- and defenestrated clauses. Interjections and discourse markers are both more frequent in defenestrated clauses, and speaker references and pronouns in framing clauses, which given the frequency of these items in M-clauses can be simply accounted for. Given the relatively low number of items in the corpus we have refrained from providing details about the statistical significance of the differences between distributions, but based on these observations we may suggest that indexical items, such as pronouns and illocutions, are frequent in framing constructions in general, and slightly more frequent in defenestrated clauses. Here, Solega mirrors a pattern we also observe in Ungarinyin (Spronck 2017 2017 “Defenestration: Deconstructing the Frame-In Relation in Ungarinyin.” Journal of Pragmatics 114: 104–133. ). Indexical elements may be slightly more represented in defenestrated clauses, but again, we propose that this does not mean that they signal perspectival shift per se. Rather, they create a context in which the function of M-clauses is sufficiently established, so that they can be left out, or, defenestrated.
The deictic evidential meaning, the attitudinal modal meaning and the unit-building ‘demonstrated’ semiotic meaning are compatible with the indexical, attitudinal and discourse marking elements that are prevalent in the Solega (defenestrated) R-clauses. We claim that the indexical elements, as well as conversation imitating strategies such as illocutionary change (also cf. Sandler and Pascual, this volume) are frequent in framing and defenestrated constructions based on their semantic compatibility with the meaning components of reported speech constructions described in (15). A comparison of the distribution of the relevant terms over sentences of reported speech as opposed to non-reported speech sentences in Figure 3 indeed suggests that they are.
In this case, the dominance of the relevant elements in reported speech seems overwhelming, and additional observations can be made on the instances where local pronouns and illocutionary change occur in non-reported speech: Non-RS clauses which have 1st person pronouns are from narratives or conversations where people are talking of personal experiences. The non-RS illocutionary change tokens are mostly rhetorical questions which are a common discursive strategy in Solega. These include instances along the lines of, “He saw an elephant, so what did he do? He ran away.” In rare cases, an (current) addressee-directed imperative is employed, as in, “Now look here…”
Nevertheless, almost none of the items that appear to cluster in reported speech, are absent in non-RS contexts. If we assume that such items are part of the marking of reported speech, this presents a problem. Distributions of items, significant or not, of the type ‘i occurs more in X, but not only in X’ are problematic if we assume that ‘i in X’ signals meaning Y, as would be the expectation if, e.g., interjections marked RS. Within the optionality-oriented defenestration approach we have proposed here the distributions are exactly what would be expected: in a narrative monologue, interactional lexemes are expected in a context of reported speech, and they support environments in which perspective shift can remain un(der)specified. If ambiguity looms, however, a full framing construction is available in Solega as a morphosyntactic tool to provide disambiguation.88.In response to a reviewer’s comment we would like to point out that we certainly do not aim to imply that ambiguity resolution is the only factor explaining the expression of M-clauses in Solega (or, indeed, cross-linguistically). There may be stylistic and a range of other pragmatic or semantic factors that affect the optionality of M-clauses in certain contexts.
6.Conclusion
The approach to the marking of reported speech in Solega developed in this article has aimed to contribute to the more general discussion about the syntax and pragmatics of irregular perspective shift in three ways: First, our observations about the optionality of marking allow for a fundamentally different approach to the problem of free (in)direct speech. Traditionally, free indirect speech has been described as a discourse pattern, or even as a construction type, and for such analytic units linguists tend to seek for conventional meaning contrasts. Within this approach it makes sense to ask questions such as ‘what is the function of free (in)direct speech?’ ‘What are the pragmatic and semantic differences with, e.g., direct speech?’ ‘Why use free (in)direct speech?’ A cursory look at the prolific literature on free (in)direct speech shows that very few of these questions have been convincingly answered.
Our analysis suggests a possible reason for this: we do not believe that defenestrated clauses in Solega form a coherent class of constructions or discourse patterns. If our optionality-oriented approach is on the right track, un(der)marked instances of reported speech are instantiations of framing constructions (which lend them their specific meaning and functions), but instantiations in which not all structural marking needs to be explicit. For such structures, the regular linguistic questions, such as sketched above are less clearly motivated: the function of a defenestrated clause, as well as its (other) pragmatic and semantic features do not need to diverge from the full framing construction it represents. Obviously we should be careful not to suggest that all examples of free (in)direct speech are defenestrated clauses in this sense. For example, the careful criteria Vandelanotte (2009 2009 Speech and Thought Representation in English: A Cognitive-Functional Approach. Berlin/New York: De Gruyter Mouton. , this volume) describes for defining free indirect speech and distancing speech are likely to produce a set of structures that can be meaningfully described as distinct functional units. However, we predict that more traditional (and less careful) definitions of free (in)direct speech in the literature as reported speech without clausal marking will include as least several instances of defenestrated clauses of the type we have described for Solega. Although we have demonstrated in Section 5 that defenestrated clauses have certain predictable properties that may provide sufficient structural conditions for Matrix clauses to remain unexpressed, these do not mark defenestrated clauses. For these structures there can be no expectation that they form a structurally, semantically or pragmatically coherent class.
A second implication of our account for the study of irregular perspective shift, and un(der)marked reported speech in particular, is that we have shown that these phenomena are not confined to written, literary languages, as many have claimed since the study of the interrelation between forms of reported speech and the evolution of the European novel in Vološinov (1973)Vološinov, Valentin N. 1973 Marxism and the Philosophy of Language. Ed. by Ladislav Matejka and I. R. Titunik. New York/London: Seminar Press.. Although this point was previously demonstrated by analyses such as by Mathis and Yule (1994)Mathis, Terrie, and George Yule 1994 “Zero quotatives.” Discourse Processes 18 (1): 63–76. and others, the view that un(der)marked reported speech is mostly confined to Western literary languages is still common.
Finally, even though we believe that the question why defenestrated clauses exist is less relevant for these structures, they do beg a more fundamental question: How can defenestrated clauses possibly be interpreted, if their full meaning is not explicitly or unambiguously marked? The answer we would like to suggest may at least hint at why there can be defenestrated clauses: un(der)marked reported speech reflects how fundamental dialogic interaction is to telling stories, to transferring information and to using language more generally (also see Vandelanotte, Van Duijn and Verhagen, Sandler and Pascual, this volume). Yes, languages can have a range of multimodal or even more subtle structural signals to indicate that we are talking about other speakers’ perspectives and utterances, like those listed in Table 3. It is essential to acknowledge that such meanings can be structurally expressed if maximum clarity is required in conversation. But the optional marking analysis developed here also proposes that it is often not necessary to be maximally explicit about perspective meanings. Apart from true pragmatic motivations for not specifying perspective shifts (such as when a speaker aims to be conspicuous about a source), perspective shifts are actually rarely ambiguous in dialogue. This is not because perspective shifts in language are irrelevant, or cannot be structurally marked, it is rather the opposite: a fundamental property of narratives and dialogue is taking and shifting perspectives, and speakers tend to such shifts with an acuteness that allows these to be left unmarked. Defenestrated clauses are an example of irregular perspective shift, requiring the addressee to understand a shift in perspective when the grammar provides no clear indication to this end. Yet the lack of structural expression when talking about such perspective meanings does not indicate that perspective meanings are linguistically irrelevant. The fact that defenestrated clauses exist across languages shows that speakers expect, anticipate and recognise perspective meanings as a fundamental property of language.
Notes
.Glosses:
ablative
accusative
adverb
complementizer
converb
continuous
dative
desiderative
focus
genitive
gerund
honorific
hortative
human
imperative
infinitive
instrumental
interjection
locative
masculine
neuter
negative
nominative
non-past
onomatopoeia
plural
past participle
present
proximal
past
reduplication
rhetorical
singular
simultaneous
temporal
vocative.
References
Address for correspondence
University of Helsinki
P.O Box 24 (Unioninkatu 40, Room B 622)
00014 Helsingin Yliopisto, Helsinki
Finland
[email protected]Biographical notes
Aung Si is an Alexander von Humboldt Postdoctoral Fellow at the University of Cologne. He has completed PhDs in Biology and Linguistics, and has carried out fieldwork in southern India, central Myanmar and Arnhem Land, Australia.
Stef Spronck is a postdoctoral researcher at the University of Helsinki, Finland. His research focuses on typology, dialogic linguistics and reported speech/TAME.
