The study of evidentiality within interactional linguistics has been influenced by work in CA on epistemics (Pomerantz 1980, Kamio 1994, Heritage/Raymond 2005, Stivers, Mondada, Steensig 2011), which is discussed in evidential typology (cf. e.g. Mushin 2013, Bergqvist/Kittilä 2020, Sandman/Grzech 2022) and applied in descriptive studies of evidential markers (e.g. Jacquin 2022, Miecznikowski/Battaglia/Geddo in press). I sketch a complementary approach that starts out from sequential categories and investigates the associated evidential resources, including grammatical/lexical means, argumentation, embodied conduct and implicit meanings (cf. also Miecznikowski 2020). It offers different opportunities for a dialogue between linguistics and CA, as will be shown here discussing the example of noticings.

Schegloff (2007:219) defines noticings as sequence-openers that speakers present as being an “outcome” of a “source” element in the setting, which typically was not attended to previously and which they retrospectively make relevant and categorize, while projecting further actions. The “outcome” relation appears to be, partly, of an evidential sort, since a verbalized noticing is typically “occasioned by a perceptual/cognitive one” (Schegloff 2007: 87, fn. 17). Multimodal analyses of noticings show that the public display of direct perception plays an important role (Kääntä 2014, Mondada 2014:382-383, Haddington/Kamunen/Rautiainen 2022); examples categorized as noticings further suggest that these actions are possible sites for markers of inference based on perceptual cues (e.g.Kärkkäinen 2007:193-194, Kendrick 2019:258, Mondada 2014:382-383).

I look at a collection of noticings taken from the videorecorded TIGR corpus of spoken Italian (SNSF grant no. 192771), consider the role of information source type and engagement in the formation of this action type, and point out theoretical implications for evidential typology. The analysis shows that noticings are compatible with various direct and indirect information sources, as long as the acquisition of knowledge occurs during the on-going interaction. It thus suggests the emic relevance of a distinction between in situ vs. past sources (cf. also Geddo, in preparation), to be interpreted within the current typological debate about evidentiality and mirativity (Aikhenvald 2021). As to verbal evidential means, a frequent construction is guarda ‘look’ (cf. Ghezzi 2012), a polyfunctional booster and promoter of intersubjective visual access. Other constructions help speakers perform evidential fine-tuning within the inferential domain (cf. Dendale & Miecznikowski in press); they add semantic specificity that distinguishes them from multimodal and deictic means, which attract the participants’ attention to perceptual cues, but cannot index the kinds of reasonings to be performed on the basis of those clues.