The templatic structure of monoclausal sentences in Kotiria and Wa’ikhana is shown in (1), with the finite verb constituting the only obligatory element. The pre-verbal position is important for our discussion, as it is typically occupied by a syntactic object (patient, recipient, benefactive) or a nominalized complement of an auxiliary verb in constructions such as progressives and purposives (see (8b) below; cf. Stenzel 2013: chapter 11).1

Although sentences that include quoted speech are biclausal, the template and examples in (2) show that their construction mirrors (1) in that quoted speech, embedded and with its own finite marking, most often occurs in the pre-verbal, object/complement position. The Quoted speaker (Q-speaker) is usually identified at the end of the sentence (e.g., “‘ABC,’ said X.” as in (2a) and (2b)). However, positioning of the referential Q-speaker and Q-addressee (participant grammatically marked by the OBJ suffix -re) varies enormously—if those participants are overtly identified at all.

The template and examples in (3) show the second most common ordering of constituents, in which the quoted speaker and speech verb precede the quoted speech.

Additional orderings and constituencies are summarized in (4), the final column indicating which examples in this article exhibit each ordering/constituency. These alternatives tend to occur in contexts of extended dialog, sequences in which both speech verbs and identification of participants (with proper names, lexical nouns, or pronominals) are optional.

The sentences in (2) and (3) were produced as biclausal quotative constructions with first-person inflection regardless of whether the elicitation prompt (in Portuguese) had a direct or indirect (third person) structure. Kotiria and Wa’ikhana have no complementizers such as the Portuguese que ‘that’, nor are there shifts of pronominal, temporal, and spatial reference indicating a change in perspective from the original speaker’s utterance to that of the current speaker that might signal indirect speech. Throughout extensive elicitation, speakers consistently neutralized the indirect/direct distinctions in the Portuguese prompts, and, except for order variations (preposing of the Q-speaker often mirrored the elicitation prompts), produced quoted speech recognizable through references presented as “faithful” to the original speaker’s perspective (Coulmas 1986: 2; Aikhenvald 2008: 383). In Kotiria and Wa’ikhana, these include both the person-referencing inflection on the quoted speech verbs and other indexing constituents, such as the possessed noun ‘your chicken’ in (5).

Such shifts in person reference, whether evidenced in verbal inflection or use of overt pronominal forms (e.g., the third-person pronouns tiro/tikido in (7) below), are the most consistent clues signaling quoted speech. Examples (6) and (7), for instance, were elicited as indirect speech with different referential values for the proper name and third-person pronoun; Nick=he in (6) and Nick≠he in (7). Nevertheless, speakers produced quoted speech with appropriate grammatical shifts for both: the shift to first person verbal inflection in (6) indicates coreferentiality within the quoted clause, while in (7), the third person pronominals tiro/tikido and non-first person inflection code non-coreferentiality.

Thus, speakers of Kotiria and Wa’ikhana only reference talk using direct quotation, its use being highly diversified. Indeed, direct quotation occurs throughout all kinds of discourse, including everyday conversation, and all sentence types can be quoted. Framers likewise take many shapes, and participants can be identified in a variety of ways.

Table 1 gives the speech verbs employed to introduce or frame quoted speech, which are cognate in both languages. The most frequent is ~di ‘say’, which has semantics generic enough to allow its use to introduce quoted talk or quoted thought in virtually any type of speech situation. The other verbs listed emphasize specific types of speech acts (e.g., the request with ~sidí in (8c) below).

Though common, speech verbs are nevertheless contextually dispensable. In narratives, for example, speaker-narrators often “animate” (cf. Goffman 1981: 145) the talk of different characters for extensive periods with limited use of framing verbs, the short sequence in (8) being a typical example. Here, we see dialog framed only by use of the participants’ names: speaker Ñahori and addressee Yuhpi Diani. Although (8a) has no explicit speech verb, the pronominal and inflectional shifts to first-person in (8b) make it recognizable as quoted speech by Ñahori. The speech verb ~sidi ‘ask for/request’ in (8c) frames Yuhpi Diani’s (equally first-person) response in (8d). Repetition of the generic speech verb ~dí in the final line of talk is a common discourse-organizing tool indicating the end of the dialog sequence (see also (9) below).

Any kind of speech act can constitute a quote. Looking briefly at sentence types, besides realis declaratives, for which evidential marking constitutes finite inflection (e.g., the “inference” and “assertion” markers in (9b) and (9c)), quoted speech can contain interrogatives, as in (9a), framed by the explicit verb ~sidítu ‘ask for/demand’. In this sequence, speaker identities are inferred from context, with only the contrastive marker -se’e on the third-person pronominal in (9c) signaling the change of speaker.

Quoted questions also appear in the following excerpt from a (multilingual) conversation between a husband (EG) and wife (ES) in (10). Both quotes contain EG’s suggestions for what his wife should ask a guest arriving at their house and are framed by the generic ‘say’ verb.

Different kinds of directives can also be quoted, e.g., the imperatives in (8d) above and (11) below. Declarative, interrogative, and directive sentence types are recognizable by finite inflectional morphology on the verb in the quoted speech.

Kotiria and Wa’ikhana speakers also regularly use quotative constructions for different types of internal talk. This may be the speakers’ own thinking, wondering, wishing, etc., as seen in (12), or the thoughts of protagonists in narrator speech, as is the case in (13). Parallels between external/internal speech are cross linguistically common, as is the fact that quoted thoughts may be identified by a generic speech verb, such as the ~di employed in both examples.

Interactional data reveal a further use of quotative constructions to demonstrate how someone else should understand or think about something. This is illustrated in (14), where a Wa’ikhana man (MC) was giving instructions to a group completing language portraits during a sociolinguistic interview (Figure 2).³ Part of the portrait task is to color in a silhouette of a human body using different colors to represent the languages in one’s linguistic repertoire and then identify each language in a key at the bottom (see insert in Figure 2). Some of the participants were unclear about what to do, but instead of using a directive, such as “Here is where you list all the languages you know,” or providing an external example, such as “In my case, here I listed … ”, MC used direct speech to enact the internal thoughts of someone remembering the languages they use in different situations, such as when flirting with a sweetheart. Indeed, the language used to romance one’s partner reflects an important facet of the multilingual makeup of Vaupés society and norms prescribing exogamic marriages (Epps & Stenzel 2013; Stenzel 2005). A potential mate must come from a different ethnolinguistic group; thus, successful romancing may entail feats of multilingual prowess! A final interesting element in (14c) is the term used for ‘woo’, a serial verb construction ya’u-sa-dʉhka, literally ‘to talk your way into’ someone’s heart.

We note that no verbal framers are employed in this sequence of sentences; the shift to direct speech is understood by MC’s use of first-person pronominals and inflectional markers. This kind of practice, sometimes called animation, has been studied from a range of perspectives (see Cantarutti 2020; Goffman 1981).

In addition to direct quotation, Kotiria and Wa’ikhana speakers can also use reported evidentials to identify talk as a (non-firsthand) source of information.4 Both languages have two reported evidential suffixes, shown in Table 2. The difference in forms codes whether the speaker is referencing a specific or diffuse secondhand source; in either case the referent is not explicitly identified.

Utterances with reported speech morphemes are difficult to elicit and are rare in narrative discourse. However, a diffuse reported marker was spontaneously used in (15) by a Wa’ikhana consultant during elicitation for this article, and the Kotiria quotative form in (16) was used early in a narrative-telling session at a community language workshop. Reported evidentials are much more frequent in conversational contexts, such as (17). We should note that in (15), the reported evidential is used within the quoted speech, as the finite marking on the verb ~kadi ‘sleep’; thus, it is not used as a tool for quoting talk, but instead implies that the subject, Roberto, is making the statement based on information he heard from others. In contrast, in (16) and (17), the -yu’ka suffixes occur on the main verbs, but only in (17) is this a speech verb with a quoted speech complement. Consultants affirm that the -yu’ka/-aye forms evoke a kind of “quotative” understanding: that the speaker is speaking for someone else or giving voice to an entity—such as the duckling in (17)—that cannot speak for itself. The boy’s choice of reported rather than visual evidential (~dira) on the speech verb ‘say’ is likely because he is clearly animating the thoughts/speech of a non-human.

As with other forms of quotation in Kotiria and Wa’ikhana, reported evidentials do not yield any kind of indirect speech interpretation or shift in perspective; indeed, the first-person inflection in Ducky’s speech in (17) clearly establishes it as directly quoted, albeit by proxy. These examples moreover demonstrate that reported evidentials may, but do not necessarily, reference speech events themselves; rather they infer speech as a source of information when talking about things for which one has no direct evidence.

Extract (18) is from a multilingual Kotiria/Tukano conversation in which the choices of languages used in quotation demonstrate differential orientation to two purported norms of multilingual speech in the Vaupés. The first is that while code-switching is generally eschewed, it is allowed in quotation (including self-quotation) with the expectation that it will faithfully reproduce the original speaker’s language choice. The second is that in conversations among people with overlapping language repertoires, each person should be loyal to their own language and use it preferentially, a practice known as “receptive” multilingualism (cf. Chernela 2013; Gomez-Imbert 1996; Stenzel & Williams 2021). Both practices occur in (18), but apparently in competition, with speakers orienting to them in different ways.

As people in the Kotiria village of Carurú-Cachoeira were gathering to watch a film, ES (ethnic Tariana, speaking Tukano, in the middle in Figure 3), her husband EG (ethnic Kotiria, speaking Kotiria, left), and FB (ethnic Desano, speaking Kotiria, right) look across the open plaza and ES points to someone (line 1), prompting the ensuing exchange.6

The first practice mentioned above occurs in lines 3–11, with FB reproducing a receptive multilingual exchange he once had with this person, who claimed to be called “Bastard” chi’nape kõakʉ, lit: ‘abandoned child’ (in Tukano). The quoted multilingual conversation consists of FB’s question in Kotiria (line 4), B’s answer in Tukano (line 6), and FB’s comment in Kotiria (lines 10–11). Embedding of multilingual quotation is clear in FB’s lines 6 and 16, where Bastard’s quoted response in Tukano is the clausal complement of the speech verb nire ‘he said’, in Kotiria. The conversation FB quotes in these lines is itself embedded into an ongoing receptive multilingual conversation in which FB and EG are speaking in Kotiria and ES is speaking in Tukano.

Competition between the two practices is triggered by EG’s request for clarification in line 12, to which both ES and FB respond, in overlap, in lines 14–16. Both employ quoted speech structures, but while FB’s line 16 continues to follow the norm of faithful multilingual quotation, ES’s version in lines 14–15 is entirely in her own language, Tukano, as the norm of language loyalty in receptive multilingual conversations would dictate.

Thus, in highly multilingual contexts such as the Vaupés, quotation in conversation entails an additional layer of complexity: the choice of which language to place in the mouth of another. While there may be additional interactional factors motivating FB and ES’s language choices in this extract, competition between ideological norms governing multilingual code-switching, e.g., faithful reproduction of quoted speech vs. loyalty in receptive multilingual talk, seem also to be at play.

Extract (19) demonstrates another interesting facet of quoted speech use in Kotiria and Wa’ikhana conversation: it is a go-to tool in storytelling contexts involving socially sensitive or taboo content. The sequence comes from a multilingual conversation around a communal breakfast during a Wa’ikhana language workshop (see Figure 4). In line 1, DD launches a story that will turn out to be an off-color joke involving a play on words in Portuguese.

Here, quoted speech in Portuguese is attributed to both the main protagonist, a “Japanese man”, in lines 4, 9, and 13 (the punchline), and an unidentified interlocutor (line 11). However, the quotative structures in lines 4, 11, and 13 have sequences of the inflected Wa’ikhana speech verb. This double structural embedding both frames the direct speech in the joke and allows DD to attribute the original joke-telling itself to some other speaker (whose identity MV suggests in line 7).

Given the joke format and the fact that it involves a play on words, we can easily understand both DD’s use of quotation and choice of quoted language. It is less clear why DD’s retelling involves structures with the observed double embedding, rather than, for example, use of a “diffuse” reported evidential (see Section 2.2 above). We suggest the choice is interactionally motivated in that it provides DD with an extra layer of distance from responsibility for the joke’s bawdy content, such distancing from taboo/delicate topics being a recognized function of quoted speech cross linguistically (cf. Bergmann 1993 on quoted speech and gossip, quoted in Holt & Clift 2007: 13). Indeed, there is evidence that DD’s interlocutors pick up on what might be coming in the story and aid his navigation of socially sensitive waters by encouraging and collaborating in the story’s production, e.g., MR’s “continuer/go-ahead” particles in lines 3, 5, and 12 (cf. Williams et al. 2020) and partial repetition in line 10.

Negotiating responsibility, albeit of a different nature, also factors in to how quotation is employed in extract (20), a short story told during a conversation between a group of Kotiria men who were transferring gasoline from a large tank to smaller receptacles for distribution (see Figure 5). The men had just concluded that the bucket they were using did not measure gasoline very accurately, and this prompted CS to recount how he once loaned gas to his neighbor, Senhor M. (Sr. M), using a similar, inaccurate bucket. From an interactional perspective, the interesting twist in the story is that it describes a morally tricky situation that benefited the teller, CS, in the end. Thus, ensuring his own role in the events is not viewed negatively is undoubtedly one of CS’s interactional goals in this telling. Indeed, this goal is masterfully accomplished through CS’s construction of what will emerge as the story’s main theses: (i) that everyone knows (or should know) such buckets are “bad”, (ii) that Sr. M did a dumb thing in choosing to use this kind of bucket, and (iii) that although CS profited from Sr. M’s choice, it was not his fault.

CS leads his interlocutors to align with his own rendition of events through skillful use of declarative and quotative constructions that support his theses. Recall that realis declaratives in Kotiria and Wa’ikhana are obligatorily coded by evidentials, and that the epistemic overtones these markers entail reflect the speaker’s perspective and relationship to events. Quotation, in contrast, allows the speaker to present events from someone else’s perspective, the finite marking in the quoted utterances duly shifting the epistemic stance and any entailed responsibility to the other person.7

When well employed, these are powerful interactional tools. For example, CS’s telling begins with a statement marked with a first-person visual evidential, -i in line 1, which ‘packages’ the story as a faithful representation of CS’s own (epistemologically unchallengeable) experience. His immediate shift to quotation of Sr. M with the first-person ‘intention’ marker -tha in line 2 (and again in line 9), on the other hand, effectively establishes the choice of which bucket to use as Sr. M’s responsibility. The visual marking in lines 3–8 shifts perspective back to CS, reinforcing the crucial point that he measured and loaned what he believed to be 25 liters of gas (but which was actually less because he was using the unreliable bucket) at Sr. M’s bidding.

As the storyline advances in line 10, however, CS states—again with visual evidential marking—that Sr. M didn’t use the same bucket to pay back the gas, the result being a windfall of extra gas for CS. Yet CS’s construction of the story continues to downplay any responsibility on his part. Note that CS now refers to Sr. M with the pejorative term ñariro ‘bad/dumb/pitiful guy’, and that both his return to quotation in line 11 and his accompanying gesture while speaking as Sr. M both underscore Sr. M’s agency in determining payback using a different bucket. CS’s self-quoted8 response (line 12) moreover has the assertion evidential, -ka, used for reasoned speculations. This strategic marking change underscores CS’s acquiescence to Sr. M’s decision and unawareness that the switch might result in anything amiss. CS’s incremental turns in lines 13–14 continue the work of diminishing his responsibility for receiving more than he was owed. Strategically, his use of the inferential evidential construction (-ri hi-re) in these lines indicates that CS only realized after the fact that a mistake (in his favor and never rectified) had occurred. The story concludes in line 16, a statement of fact (vis-a-vis visual evidential inflection) that directly attributes responsibility for Sr. M’s use of a different bucket, and indirectly absolves CS of any wrongdoing.

Yet the end of CS’s story is still not the end of our story. The close of the telling now triggers CS’s interlocutors to produce different expressions of alignment. JM’s input in lines 17–18 is particularly interesting for our discussion here. He first states a conclusion that Sr. M was not “able to see” the difference between the buckets. This statement is marked by the epistemically strong assertion evidential -a, which frames statements as ‘collectively known’ rather than ‘personally experienced’. Thus, JM essentially promotes ‘what CS said about what happened’ to ‘something we all know/agree happened’. JM follows this up with line 18, his own quotation of what Sr. M must have been thinking at the time he made his regretful choice of bucket. CS immediately acknowledges and confirms JM’s interpretation in line 19, a statement with both explicit epistemic and visual evidential inflection.