1 Introduction

1 From reported speech to reported thought

Speakers routinely discuss the words of others, whether they have actually heard these words being spoken or merely attribute them, with constructions as in (1)-(4):

1. Sasha said: "This will be a beautiful day."

2. Sasha said that it would be a beautiful day.

3. Sasha thought: "This will be a beautiful day."

4. Sasha thought that it would be a beautiful day.

The first two of these sentences are examples of reported speech, traditionally divided into direct speech (1), which purports to represent an utterance without significant modification by the current speaker, and indirect speech (2), which has no such implications (Coulmas 1986). While many questions remain in the study of reported speech, it has received relatively wide attention, including in several influential volume-length treatments (e.g. Lucy 1993) and in many individual studies.

This cannot be said for reported thought, as exemplified by (3) and (4). As these example sentences show, expressions of reported thought parallel those of reported speech in many ways, at least in English and most other commonly described languages. We could therefore label (3) 'direct thought' and (4) 'indirect thought' and this classification is straightforward on structural grounds: in both (1) and (3) deictic pronouns and tense are indexing the origo of the subject of the matrix clause and in the written form they include quotation marks, both (2) and (4) appear to involve complementation with an embedded proposition, and the only distinction between the two pairs of examples of reported speech and reported thought is the choice of the lexical verbs say and think.

But what does 'direct thought' mean and what would motivate its choice over ­'indirect thought'? While authors have frequently remarked that direct speech is rarely actually a 'verbatim' representation of some one's words (see Vandelanotte 2009: 118?130 for discussion), the question of what it means to reflect someone's thoughts as they were thought, instead of by approximation, seems absurd. There is an obvious, fundamental difference between the usage context of reported speech and reported thought: a thought, by its very definition, remains unspoken. No one has direct access to a thought apart from the person who thinks it, the so-called 'cognisant' (apart from the exceptional case of omniscient narrators in a literary text). Therefore, the main questions of the reported speech literature, that primarily deal with the representation and manipulation of utterances and the coordination of the perspectives of multiple speakers and hearers, are irrelevant for understanding reported thought. As a consequence, the phenomenon remains largely unstudied.

This is not to suggest that reported thought has gone completely unmentioned in studies of reported speech. Authors have, roughly, dealt with it the phenomenon in one of three ways:

1. assuming that reported thought is mostly equivalent to reported speech (the equivalence assumption);

2. assuming that reported thought is derived from reported speech (the derivation assumption);

3. assuming that reported thought is irrelevant for understanding reported speech and can therefore be ignored in studies on the latter phenomenon (the dissimilarity assumption).

As the initial examples in (1)-(4) illustrated, the equivalence assumption is not unreasonable, especially for languages like English. From a syntactic viewpoint, structures involving reported speech and reported thought are often similar (Palmer 1986: 135; Spronck & Nikitina 2019). Moreover, the predicates used in reported speech (e.g. say, tell) and reported thought (e.g. think) behave as bridge verbs allowing for a number of syntactic phenomena, including extraction across wh-questions, embedded V2 in Germanic languages, and complementizer deletion (Erteschik-Shir 1973; Vikner 1995; Cocchi & Poletto 2002; Dor 2005; see also Salvesen & Walkden 2017). Apart from these syntactic similarities, across nearly all continents languages have been found that do not even make a lexical distinction between 'say' and 'think' (Larson 1978; Saxena 1988; Rumsey 1990; Reesink 1993; Güldemann 2008; Spronck 2015), resulting in an even greater structural resemblance between reported speech and reported thought.

The derivation assumption is also quite plausible. On the one hand, it serves as a diachronic explanation for the equivalence assumption (a reported speech expression often constitutes the source construction for reported thought in languages in which the two are indistinguishable, cf. Spronck & Casartelli 2021). This explanation is made even stronger by the fact that no language has so far been demonstrated to show the opposite development path, i.e. from reported thought to reported speech. On the other hand, even languages in which the verbs of speech and thought are consistently distinguished, speech verbs are routinely used to express mental states (Pascual 2014), as in expressions like 'I said to myself. . .' or 'Her smile spoke volumes'.

A particularly interesting psychological motivation for the derivation as­­­­­­­­­­­­­­­sumption lies in Vygotsky's (1987) notion of 'inner speech', which interprets thought as a form of speaking to oneself (cf. Volosinov 1973). In a more recent study, Mercier & Sperber (2017) build on this idea by proposing that our human ability to reason is a consequence of internalizing dialogue, with conscious thought being no more than practicing how we are going to explain our judgements and intentions before vocalizing them. Where expressions of reported thought are diachronically or structurally based on reported speech constructions, they mirror this process of internalization in a remarkably iconic way.

The dissimilarity assumption can be adopted for many reasons, for example: a traditional formal semantic analysis of quotation may choose to exclude reported thought because it does not involve statements with a verifiable truth-value and for a similar reason a functionalist account of quotative meaning (cf. Boye 2012) may disregard reported thought since it does not involve a fully embedded illocution or speech act.

Such analyses build on an important observation: the fact that there are empirical and theoretical differences between reported speech and reported thought. These emerge, for example, in the differences in the syntactic structure of complement clauses involved in reported speech and reported thought: authors have pointed out that the complements of thought predicates seem to lack an independent illocutionary force, while speech predicates select for clauses with a full structure licensing root phenomena (Hooper & Thompson 1973; Heycock 2006). Such distinctions also lead to different diachronic processes leading to (semi-)grammaticalized constructions, parenthetical expressions, evidential or epistemic adverb(ial)s (cf. English methinks, Spanish dizque, Greek lei), discourse markers and modal particles, and grammatical elements such as complementizers (Cruschina & Remberger 2008; Thompson & Mulac 1991; Posio 2014; Cruschina 2015; Wiemer 2018, and references therein).

Furthermore, reported speech and thought may involve different complementizers (see, e.g. Ledgeway 2005), or have different moods in the embedded clause (see, e.g., Laca 2013), and behave differently with respect to other phenomena such as negation raising (Horn 1989) and the omission of complementizers in spontaneous speech (e.g., Posio & Pesková 2020).

Therefore, while none of the assumptions that previous authors have taken to the discussion of reported thought are necessarily false for the purposes of the respective studies, or even mutually exclusive, they have often tipped the balance to studying reported speech instead of reported thought. This is where the current volume aims to rebalance the scales, not by assuming that reported thought is the same, derived or even necessarily different in all languages from reported speech, but by asking the question: what happens if we study reported thought as an autonomous research topic? In the chapters of this volume, authors take up this challenge from a range of backgrounds, theoretical persuasions and linguistic sub-disciplines.1