Evidentials and dubitatives in Finnish: perspective shift in questions and embedded contexts

2.1 Reportative evidentials outside root declaratives: distribution and interpretation

There exists considerable crosslinguistic variation in the distribution of evidentials outside of root declaratives. In some languages, evidentials can occur in embedded contexts, including complement clauses embedded under verbs such as say and think, but there is variation in terms of whether the speaker or the matrix subject is construed as the evidence holder. Next, I provide a brief overview of typological variation regarding the distribution and interpretation of reportative evidentials in questions and embedded contexts in particular, before turning to the Finnish data.

3.1 Finnish reportative evidentials in root clauses

Before looking at questions and embedding under attitude verbs, let us consider kuulemma in root declaratives (see Kaiser 2022 for an in-depth analysis). In essence, kuulemma expresses that the speaker received the information from someone else (10–11) (e.g. Kittilä and Sandman 2013) – i.e., kuulemma indicates that the current speaker is reporting second-hand information from someone else.

In this paper, I use constructed examples, examples from prior work, as well as naturally-occurring examples from the internet that I identified through google searches and using the corpora hosted by the Finnish language bank (KORP, Borin et al. 2012), marked with ‘www.’ Using naturally-occurring examples from present-day Finnish provides insights into the rich contextual factors that guide the use and interpretation of reportative kuulemma and dubitative muka.

The reportative kuulemma is a free-standing trisyllabic word and does not inflect. It is in all likelihood derived from the verb ‘to hear’ combined with the first-person pronoun in non-canonical VS order (kuulen ma ‘hear I’) (see Sadeniemi and Vesikansa 1989).

In earlier work (Kaiser 2022), I show that with kuulemma, the speaker (in the terminology of Goffman 1979, the ‘Animator’) is not committed to the truth of the reported proposition p (see also Pancheva and Rudin 2019). In this regard, kuulemma resembles reportative evidentials in other languages (e.g. AnderBois 2014; Faller 2002, 2019).

Furthermore, for kuulemma to be felicitous, the information in the reported proposition p must have been previously communicated linguistically – i.e. spoken or written – by someone else (the ‘Principal,’ to use Goffman’s term). In other words, the Principal (but not the Animator) must be discourse-committed to p. Kuulemma cannot be used if p is merely inferable from someone’s behavior or other non-linguistic inferential evidence (see e.g. Kuiri 1984: 33, Hakulinen et al. 2004, Section 1606, p. 1524).

In addition, as illustrated in (12), the reported proposition p is at-issue information – i.e. p is put on the table – as shown by the fact that it can be used to answer the ‘question under discussion’ (QUD, see e.g. Simons et al. 2010; Tonhauser 2012 on the QUD test). Thus, kuulemma resembles reportative evidentials in other languages (e.g. AnderBois 2014; Faller 2019; Murray 2014) which also put p on the Table.

However, the information expressed by kuulemma itself is not-at-issue, as shown by it not being directly deniable (13). (See e.g. Tonhauser 2012 on the direct deniability test.) Speaker B’s response in (13) seeks to target only the meaning of kuulemma, just like B’s response in (12) targets p – but the response is infelicitous in (13), in contrast to (12).^[3]

Given that kuulemma is neutral regarding to the speaker’s commitment to the reported proposition p, but p constitutes at-issue information, we may wonder whether p can be added to Common Ground, like assertions can. In Kaiser (2022) I show that p can be, but does not need to be, added to Common Ground. In this regard, kuulemma resembles Cuzco Quechua reportatives, regarding which Faller (2019) notes the following: “a reported proposition can update the discourse in ways similar to regularly asserted propositions, but only if the speaker intends this and if the addressee recognizes this intention” (Faller 2019: 8).

In sum, when kuulemma is used in root declaratives, the speaker is neutral about the truth of the reported proposition p, p must have been communicated linguistically, which entails that there exists another person who is publicly committed to p, p is at-issue information (i.e. put on the Table) that can potentially be added to the Common Ground, whereas kuulemma itself conveys not-at-issue information.

3.2 Finnish reportative evidentials in questions

Let us now look beyond root declaratives, starting with questions. Although many languages allow reportative evidentials in polar and wh-questions, in Finnish kuulemma seems to be odd in regular polar (yes/no) questions (14). However, kuulemma can felicitously occur in polar questions with negation (15) as well as in wh-questions (16).

In polar questions with negation, kuulemma is speaker-anchored – in other words, the speaker is the relevant evidence holder. For example, in (15), by using kuulemma the speaker signals that they have heard that the person being talked about lives in Tampere. Thus, there is no interrogative flip and kuulemma is not tapping into whether the addressee has heard this. Rather, kuulemma makes reference to reportative evidence possessed by the speaker who asks the question.

Similarly, kuulemma is speaker-anchored in wh-questions and does not appear to exhibit interrogative flip. In (16), the speaker signals that they have been told that their breath smells bad (i.e., the speaker heard this from someone else) and are trying to find out the cause of their halitosis.

In sum, kuulemma seems to be infelicitous in standard polar questions, but can occur in polar questions with negation and in wh-questions, where it has a speaker-anchored interpretation (does not exhibit interrogative flip). Why is the distribution of kuulemma constrained in this way? In the rest of this section, I argue that these distributional properties of kuulemma can be derived from independent properties of wh-questions and polar questions with negation (vs. standard polar questions), coupled with the observation that Finnish is a language where reportative evidentials do not exhibit interrogative flip, i.e., do not shift to the addressee in questions.

3.3 Finnish reportative evidentials in embedded contexts

In the preceding section I presented evidence suggesting that kuulemma can be used in (some) questions, but that it remains speaker-anchored and can only be used in contexts where the question conveys a proposition p that the speaker has reportative evidence for (i.e. that the speaker heard from someone else). Thus, unlike what is often reported in the literature for other languages, we see no signs of interrogative flip with the Finnish reportative.

Ultimately, the question data fits well with what Kaiser (2022) proposed for declaratives: In questions – just like in declaratives – when kuulemma is used, it indicates that the target proposition was previously expressed by another speaker (the Principal) who is discourse-committed to the proposition, unlike the current speaker (Animator). I propose that this explains why kuulemma is felicitous in biased polar questions and (some) wh-questions but not in regular polar questions: simply put, kuulemma can only be used when the utterance expresses a proposition p that the speaker (the evidence holder) heard from someone else.

Next, we consider another context with two potential evidence holders: kuulemma embedded under verbs of speaking (e.g. say). Here, the two possible evidence holders are the speaker and the matrix subject of the verb, rather than the speaker and the addressee. The next section explores the discourse profile of kuulemma in this context.

3.4 Taking stock: Finnish reportative evidentials

The data we have considered so far indicates that whether the Finnish reportative evidential can shift away from the speaker to another evidence holder depends on who the other possible evidence holder is. In the case of questions, there seems to be no shifting to the addressee: kuulemma fails to show interrogative flip and maintains a speaker-oriented reading. However, when embedded under attitude verbs (more specifically, the basic verb of saying, sanoa), kuulemma can optionally shift from the speaker to the matrix subject. This suggests that whatever mechanism is responsible for shifting under say – for an example some kind of shifting operator that applies to kuulemma, though here I remain agnostic about the details of this mechanism – can optionally occur in embedding contexts but not in questions.

4.1 Dubitatives in questions and embedded contexts

The empirical profile of dubitatives in questions and embedded contexts crosslinguistically is not yet well-investigated. In particular, there seems to be surprisingly little systematic crosslinguistic comparative work on dubitatives in questions. Broadly speaking, a speaker expressing doubt about the truth of a proposition is highly compatible with that speaker asking a question about that proposition. Thus, it seems reasonable to speculate that dubitatives can occur in questions.

Furthermore, there does not seem to be any reason to expect dubitatives in question to show interrogative flip, i.e. to be construed as tapping into the addressee’s doubts – in fact, quite the opposite. In declaratives, the speaker is the one who has doubts about the proposition, and this is expected to also hold in questions. Unlike reportative evidentials, there’s no communicatively plausible reason to expect dubitatives in question to be interpreted as signaling that the addressee is the one who doubts the proposition. It’s communicatively meaningful for someone who doubts a proposition to use a question to ask others if it’s true, but seems much less communicative meaningful for a speaker to pose a question to an addressee if the speaker believes the addressee has doubts about the truth of the proposition to start with.

In sum, it seems that the a priori expectation is that dubitative elements can occur in questions and that they will be speaker-anchored. However, the behavior of dubitatives in the second context we are concerned with, embedding under attitude verbs, is less clear.

Sauerland and Schenner (2007) investigate the behavior of the dubitative in Bulgarian, which is a periphrastic verbal form consisting of two occurrences of perfect morphology. It expresses “indirect evidence and doubt” (Sauerland and Schenner 2007:533). Sauerland and Schenner analyze the two occurrences of perfect morphology in dubitative sentences like (27) as constituting two components, ‘REP’ and ‘DUB’. Thus, on their analysis, the “dubitative in Bulgarian is inherently embedded under assertive REP” (p. 146).

Using (27), they show that the dubitative, when embedded under ‘to say,’ is obligatorily speaker-oriented and cannot shift to the attitude subject.

According to Sauerland and Schenner, (27) receives a reading where the speaker of the sentence doubts the information provided by Maria (i.e., that Todor has red hair). Thus, the dubitative is speaker-oriented. Sauerland and Schenner also argue that the dubitative cannot receive a shifted (non-speaker-oriented) reading, i.e., that (27) would not be felicitous in a context where Maria doubts the claim that Todor has red hair but the speaker knows it to be correct.

In contrast to Bulgarian, in Finnish the dubitative and the reportative are fully independent morphemes. While the Finnish dubitative and reportative can co-occur (Kaiser 2022), they do not need to do so (ex. 1–4); dubitative muka can occur entirely independently of reportative kuulemma and vice versa. Thus, Sauerland and Schenner’s analysis of Bulgarian cannot be extended directly to Finnish.

5.1 Finnish dubitatives in root clauses

Before turning to the behavior of the dubitative in questions and under attitude verbs, let us briefly consider root declaratives (see Kaiser 2022 for an in-depth analysis). Historically, muka evolved from a reportative evidential, but in present-day use it “indicates a speaker’s dubitative stance” (Nordlund and Pekkarinen 2014: 53).^[6] Kangasniemi (1992) describes muka as “an expressive adverb which enables the speaker or writer to indicate that he or she has got the information from someone else and personally doubts whether it is true or does not agree with it” (Kangasniemi 1992: 207).

In many cases where muka signals the speaker’s doubt, the speaker received the information via a prior linguistic communicative event. However, unlike the reportative kuulemma, muka can also be used to express doubt about propositions that have not been directly linguistically expressed by another person in a prior speech (or writing) event. Thus, in some contexts the existence of the Principal is rooted in inference, rather than explicit commitment (see Kaiser 2022).

In terms of its discourse update potential, muka resembles kuulemma, in that the proposition p in the scope of muka contributes at-issue meaning as shown by the fact that p can be used to answer the QUD (i.e., is put on the Table), as shown in (28).

However, muka itself (like kuulemma) contributes not-at-issue meaning. This is shown by the direct deniability test (29) which indicates that information conveyed by muka cannot be directly challenged or denied. Assuming that direct deniability/challengeability is a diagnostic of at-issueness, this indicates that the doubt meaning of muka is not-at-issue.^[7]

Although muka resembles kuulemma in that the target proposition p is at-issue but muka itself is not, the two forms differ in terms of whether p can be added to Common Ground. With kuulemma, Kaiser (2022) shows that the reported proposition p can be added to Common Ground if the speaker intends to do this and the addressee accepts this (see also Faller 2019). This differs from muka: Kaiser (2022) shows that the proposition p in the scope of muka cannot be added to Common Ground, although it has been put on the Table. In this regard, a proposition in the scope of muka patterns unlike asserted propositions and also unlike propositions modified by kuulemma.

5.2 Finnish dubitatives in questions

The dubitative muka can occur in ‘regular’ as well as negated polar questions (30–31), in addition to wh-questions (32). In all cases, use of muka signals the speaker’s doubt. For example, in (30), the speaker doubts the proposition that ‘microwave popcorn is more convenient than normal popcorn’, and in (31) the speaker doubts/does not believe the proposition that ‘one can’t find actors in Finland.’ Indeed, presence of muka is associated with a preference to interpret the question as having the illocutionary force of an assertion that has the opposite polarity that what is apparently being asked – i.e., use of muka is associated with rhetorical questions (see e.g. Sadock 1971, 1974 for an early discussion of rhetorical questions). However, rhetorical questions in Finnish do not require presence of muka.

At this stage, I leave a detailed analysis of the relation between muka and rhetorical questions for future work. What is most relevant for our purposes here is that in questions like (30–32), the doubter is clearly the question-asker, not the addressee. In other words, muka is anchored to the speaker and does not shift.

5.3 Finnish dubitatives in embedded contexts

The dubitative muka can be embedded under sanoa ‘say’, as in (33–34). Native speaker judgments suggest a preference for a speaker-oriented interpretation,^[8] such that the actual speaker of the sentence (not Mari) is the one who doubts whether Liisa has moved to Pori in (33). The same pattern is illustrated by the corpus example in (34). Here, the one doubting the customer service representative’s statement is clearly the speaker, not the customer service representative. (This example uses a ‘lauseenvastike’-construction, a participial complement clause.)

5.4 Taking stock: Finnish dubitatives

The examples considered so far suggest that, more than the reportative kuulemma, the dubitative muka has a strong preference for speaker-anchored interpretations. In questions, echoing what we saw with kuulemma, muka maintains a speaker-oriented reading. When embedded under sanoa ‘to say’, the dubitative muka seems to be more strongly speaker-oriented than the reportative kuulemma.

Why would muka have a stronger speaker-orientation than kuulemma? I suggest that this may be due to the relatively more subjective nature of muka. As discussed above, kuulemma requires the presence of a particular communicative configuration – namely that the current speaker heard p from another source during a prior linguistic communicative event. Thus, kuulemma provides objective information: whether a prior communicative event occurred is a factual matter. In contrast, muka seems to be arguably a more subjective and affective expression. It indicates the doxastic state of a particular speaker – i.e., signals that, for some reason, this individual doubts (or disbelieves) p. Other people in the same context may not experience the same level of doubt. Furthermore, as Nordlund (2011) points out, muka often has an affective/expressive component and can signal irony, surprisal or disapproval (“Sen sisältävään lausumaan liittyy lisäksi usein affektiivinen ironian, hämmästelyn tai paheksunnan sävy,” Nordlund 2011: 485). Thus, muka and kuulemma differ fundamentally in their level of subjectivity, and it may be that the stronger speaker orientation of muka relative to kuulemma stems from the former being a more subjective expression.