On the exceptionality of reported speech

1 Introduction

Spronck and Nikitina (S&N) have taken on the task of defining a linguistic phenomenon that has managed to elude definition, despite playing a key role in many subfields of linguistics. In formal semantics in particular, speech reports have been at the center of attention from the very beginning (Frege, 1892). S&N’s endeavor presupposes that there is something worth defining, i.e. that reported speech is indeed a linguistic category of its own, not an arbitrary intersection of various other, larger linguistic categories such as clausal embedding and evidentiality. In this response I want to provide additional, semantic evidence for S&N’s claim that reported speech should be treated as a linguistic category.

2 The traditional picture: Attitude ascription, quotation, and evidentiality

As S&N point out, the cluster of phenomena that they call reported speech is not always considered a linguistic category in its own right. In semantics, in particular, it is rather considered a mixture of (at least) three fundamentally distinct subtypes, viz. direct and indirect reported speech and reportative evidentials, each reducible to a larger category. Let me start by reviewing this traditional picture.

First, indirect reported speech is considered just a special case of the larger natural class of attitude ascriptions:

Syntactically these are all clausal complement constructions, and semantically they are analyzed as intensional operators, i.e. quantifiers over possible worlds: (1) is true iff Sue stole Maria’s idea in all the possible worlds compatible with what she said/believed/etc.^[1]

Second, direct reported speech is considered a special case of the larger natural class of quotations, which includes not only speech report uses of quotation marks, but also some thought reports (2b), and so-called pure quotation or mention (2c), perhaps even scare quotes, role shift, demonstration, etc.:

Semantically, what unifies the broad class of quotations is metalinguistic reference, i.e. reference to linguistic expressions.

Reportative evidentials, finally, are considered a species of evidential marking:^[2]

Semantically, the unifying characteristic of evidentiality is that it is a way of specifying what evidence one has for a specific claim, in a such a way that this evidence information is ‘backgrounded’. In current terminology, the at-issue contribution of (3) is that it is raining, and the not-at-issue contribution is that the speaker has either direct perceptual or indirect hearsay evidence for this (Murray, 2017).

Summing up, on the traditional picture there is no category of reported speech, but a cluster of distinct phenomena that can be used for various things, including but not limited to reporting what someone said. Yet, over the years various isolated semantic phenomena have been brought to light that seem to support S&N’s thesis that there really is something special about speech reports after all.

3 The special status of speech reports

I aim to establish that, within each of the traditional broad semantic categories reviewed above, the speech reporting subclasses tend to behave rather differently from the remainders.

4 A note about narrative perspective shifts

I end with a brief, more speculative note about Free Indirect Discourse (FID) and perspective shift in narrative. FID is a way of describing a character’s thoughts and utterances in a way that combines features of both direct and indirect discourse (Banfield, 1982).

Within semantics there has been considerable interest in this phenomenon (Schlenker, 2004; Maier, 2015; Eckardt, 2014). One popular analysis emerging from this debate starts from the idea that these sentences are to be evaluated with respect to two ‘perspectives’, i.e. formally, two distinct context parameters: the context of narration and the protagonist’s context of thought. More recently, this ‘bicontextual’ approach has been extended to deal with other kinds of perspective shifts, such as ‘Protagonist Projection’, (10a), and ‘Viewpoint Shift’, (10b).

The idea is that in (10a) the ring is studded with diamonds from the perspective of the woman receiving the ring, but it’s glass from the narrator’s perspective. In (10b) the T-Rex is familiar from the perspective of the narrator and the reader, hence the definite article in the first sentence, but unfamiliar from the perspective of the little dinosaur character Billy, hence the indefinite article in the third sentence.

While Stokke (2013) and Abrusáan (2018) stress the similarities between the phenomena in (9) and (10) – with Abrusáan even hinting at a continuum of perspective shifting –, Hinterwimmer stresses the differences. He describes three linguistic tests to tease apart FID and Viewpoint Shift. For example, while FID is a root phenomenon, Viewpoint Shift may occur in embedded positions. In fact, decades earlier, Banfield (1982) had likewise proposed linguistic tests for distinguishing FID from a type of narrative perspectival shift that she called ‘the representation of non-reflective consciousness’. For example, Banfield shows that adding expressive elements like exclamations or vocatives turn a perspective shift into a full-fledged FID report (again in line with S&N’s ‘demonstratedness’ criterion for reported speech).^[10]

The picture that emerges from the Banfield/Hinterwimmer observations is one where FID is a quotation-like type of speech reporting (where speech includes inner speech, i.e. thought), while Viewpoint Shift (and presumably at least some examples of Protagonist Projection) is a way of describing a character’s perceptual sensations or other non-linguistic, non-reflective mental states. In other words, within the broad class of perspective shift phenomena, the reportative variety known as FID stands out as a separate phenomenon with its own quotation-like syntax and semantics.

5 Conclusion

Looking over the formal semantics literature on reported speech constructions, I find myself in agreement with much of what S&N conclude from their typological studies. On future occasions we could quibble over the details of the proposed definition – Do cases of standard indirect discourse or hearsay evidentiality really always involve ‘demonstratedness’? And in what sense does a free indirect speech or thought representation in a fictional narrative mark ‘evidentiality’ or ‘source of information’? But in this commentary I’ve restricted myself to providing support for S&N’s general project from a semanticist’s perspective. More specifically I’ve listed a number of ways in which speech reports differ semantically from various related phenomena to which they have been reduced by other semanticists. My hope is that these observations may prove helpful in the emancipation of speech reports as a dedicated linguistic category.