Evidentiality is most often defined as a linguistic category indicating the ‘source of information’ for what is being said (cf. Aikhenvald 2004). The example from Cuzco Quechua (Quechuan, Peru, Faller 2002: 122) demonstrates how it can work in a language where this category is expressed by dedicated morphemes:

Parashanmi ‘It is raining’ [the speaker sees the rain]

Parashanchá ‘It is raining’ [the speakers conjectures it without observing the rain]

Parashansi ‘It is raining’ [the speaker was told by another person]

As these examples suggest, the use of evidentials does not change the propositional content of the utterance. Rather, it adds an additional layer of meaning (cf. Faller 2002; Boye 2012), the precise nature of which varies between languages and remains an object of study and debate.

Nonetheless, in line with the definition of evidentiality given above, this additional layer is widely assumed to indicate ‘how the speaker knows’ in all languages with evidential markers. However, a growing body of descriptive research shows that evidentials do more than that.

When we analyse interaction rather than isolated sentences, we find that, across the languages where they are attested, evidentials signal not so much the type of evidence, as the basis on which the proposition should be integrated with what is already known. Source of evidence is relevant, but not key for how they are used and interpreted. In interaction, evidentials appear to be structuring knowledge and providing interpretative cues to make communication more effective, encoding meanings related to the distribution of knowledge in interaction, epistemic stance and status, discursive and social roles of the interlocutors, etc.

This talk has three main objectives. Firstly, I will demonstrate how interactional data from under-described languages supports the observation that evidentials vary in terms of their meaning and their interactional functions. Secondly, I will show that, in line with the above, interactional uses of grammatical evidentials attested in under-described languages, and those of ‘evidential strategies’ (cf. Aikhenvald 2004) attested e.g. in standard average Indo-European languages, have much more in common than is acknowledged by the current scholarship. Finally, I will discuss the implications of these observations for comparative research on evidentiality, focusing on the issue related to comparing evidential data across spoken corpora of different languages, especially when these corpora differ in size, levels of annotation, represented genres, or the numbers of represented speakers.