Quotatives, or reports of past speech acts, are traditionally viewed as split between two types: oratio recta and oratio obliqua, direct and indirect speech. These terms originate with the grammatical description of Ancient Greek and Latin (Liberati 2017 and others), but a similar distinction is found in many present-day European languages (Coulmas 1986). The examples in (1) illustrate the contrast between direct (1a) and indirect (1b) speech.

Because in a famous speech at the height of the cold war John F. Kennedy produced the German words “Ich bin ein Berliner”, which can be translated into English as “I am a Berliner”, both (1a) and (1b) are acceptable as true. But (1a) would not be acceptable if we changed the subject from first person I to third person he or the verbal tense from the present tense am to the past tense was. This is because in English direct speech the coordinates of evaluation are obligatorily shifted from me, the current author, and the time of my speaking back to John F. Kennedy speaking on June 26, 1963 in Berlin. So, he can’t be used to refer to Kennedy and was could not be used to refer to 1963.

A second important grammatical property of quotatives in English and other European languages is syntactic islandhood. It is ungrammatical to question a constituent of a quotative as in (2a), but English allows a parallel structure in indirect speech.

Much of the recent interest in speech reports such as these originates with two independent events: a renewed interest in recursion following the publication of Hauser, Chomsky & Fitch (2002), and the discovery of indexical shift. Of course, recursion is found in languages at many levels, and many theoreticians would agree that any three-word sentence must involve a recursive structure. Still, recursion of one sentence contained within another as is found in speech reports may constitute a special case. This is justified from a cognitive science perspective, as for example Bühler (1934) has proposed that speech reports specifically involve meta-representation in their interpretation, i.e., rather than directly evaluating expressions from the current perspective, it is the general relationship established between a perspective and an evaluation that determines the interpretation. Furthermore, there is some evidence that theory of mind ability, i.e., the ability to ascribe mental states different from one’s own to other individuals, a cognitive ability at least closely related to Bühler’s meta-representation, is in part linked to the ability to form embedded sentences.

But the properties of quotatives are not everywhere as uniform as they are in English and the few other European languages. An important discovery has been the phenomenon of indexical shift that has been found in many languages across the globe, especially since Schlenker’s (2003) work on Amharic (Hamito-Semitic) pointed out the theoretical relevance of these cases. A central example of his discussion is presented in example (3), where the second person in the speech report is not understood as referring to the addressee of the reporting speech act, but as referring to the addressee of the reported speech act. At the same time, the question word what cannot be part of the speech that is reported. So, (3) would seem to correspond to the English structure “I didn’t hear what he said to me ‘you bring’”, which like (2a) is ungrammatical in English.1

Schlenker’s analysis is that (3) is not a full quotative; Amharic allows an operation of indexical shift that shifts some aspects of an indirect speech report back to the original speech event. The result is a type of indirect speech that shares some characteristics with direct speech.

How surprising the discovery of indexical shift was for theorists can be seen from the fact that the philosopher David Kaplan (1978) called the type of context-shifting operator that makes indexical shift possible a “Monster” and claimed that such operators did not exist in human language. Making reference to this, the current literature on the topic frequently uses a typographic monster from the computer game Pacman to indicate the operator.

Interestingly, indexical shift is not uniform across languages. Consider examples from the Amazonian language Matses (Panoan). The Matses example in (4) was elicited by means of a translation task from Spanish to Matses for the Spanish indirect speech report “Danny said yesterday that he will come tomorrow”. However, the Matses translation in (4) obligatorily contains the first person instead of third person, tomorrow instead of today, and go instead of come. In all three cases, the indexical elements in the speech report are interpreted not relative to the current time of the speaker giving the speech report as in indirect speech, but are shifted back to the event of the original speech event that is being reported as in quotations.

At the same time, Matses quotations differ from English direct speech in two ways. As illustrated by (5), Matses allows extracted structures similar to (2a), which are not possible in English.

Furthermore, (6) shows another case where two perspectives are available in a speech report in Matses in what is called a de re interpretation. Here, the belief of Dashe’s reported speech was not that it is impossible that there are actual sloths that are also termite nests, but (6) is acceptable in a scenario where Dashe saw an actual termite nest in the dark and thought it was a termite nest. Note that English direct speech (Dashe thought saying “the termite nest is a sloth”) does not allow such de re interpretations.

Deal (2020) provides an overview of recent theoretical work on indexical shift. In addition to Amharic, languages important to the discussion have included Nez Perce, Slave (Anand & Nevins 2004), Uyghur (Shklovsky & Sudo 2014), Tamil (Sundaresan 2013), Tatar (Podobryaev 2014), Telugu (Messick 2017), Zazaki (Anand & Nevins 2004), and more recently Tabasaran (Bogomolova 2023) and Aqusha Dargwa (Ganenkov 2023).

The Matses example shows that data on quotatives from Amazonian languages are of great theoretical interest. Up to now quotatives in Amazonian languages have been primarily discussed within grammars or under the general topic of subordination (van Gijn et al. 2011). Some Amazonian languages have been described as only allowing direct speech reports: for example, in addition to Matses (Fleck 2006), Michael (2001) reports this for Nanti (Arawakan). Brazilian and/or Amazonian languages are of great interest for understanding quotatives since they represent different families and, in some areas, indexical shift may be a regional phenomenon across different language families. The topic of quotation in Amazonian languages is for this reason (among others) of broad theoretical interest.

In Section 2, in addition to exploring the contrast between direct and indirect speech, we discuss other aspects of quotatives based on the descriptive and typological literature (cf. Spronck & Nikitina 2019) in light of the patterns described by the authors of this volume. These include, for example: the argument structure of speech verbs, characteristic features of quotatives as well as of the main clauses that may co-occur with them (such as I said to her in “I said to her: ‘I will leave soon’”), and the impact of text genre in the choice of quotative structure.

In this section we present an overview of some features observed across the Indigenous languages spoken in Brazil that are represented in this volume. Those languages are:

The data reported on combine both analyses of texts and data from elicitation sessions. The authors were provided with a previously unpublished questionnaire prepared by Tonjes Veenstra, Denny Moore, and Hein van der Voort which includes a list of relevant phenomena for the investigation of quotatives. (Veenstra, Moore, and van der Voort’s questionnaire is published as part of the present collection.) When describing quotatives in a language, some of the questions researchers were encouraged to explore are:

In the remainder of this section, we present an overview of some features observed across the languages studied in light of these questions.

In this paper, we discussed some of the features that characterize reported speech in some Indigenous languages spoken in Brazil. In light of typological studies on the topic, we have drawn attention to some common features observed across such languages where direct speech seems to prevail. Multiple strategies may be employed in quotative constructions, including speech verbs, absence of quote framing, and partial presence of functional morphology without a speech verb (as observed in Kuikuro). The complete or limited marking of the matrix clause (the framer clause that may identify the reported speaker), also known as defenestration, is a common feature across languages (Spronck & Nikitina 2019: 126). Reported speech can also be analyzed as a multimodal phenomenon given that not only morphosyntactic cues but also prosody and gestures may be relevant components of reported speech cross-linguistically (Spronck & Nikitina 2019: 142). Finally, we find that the type of text has an effect in determining which quotative structure is going to be used in a language.

One last significant dimension of reported speech that is explored in this volume and that we wish to highlight is the pragmatic role of these structures. The contribution by Stenzel & Williams shows that exploring the use of quotatives in the discourse is an important yet understudied field of research, particularly in lesser studied languages.