Universal meaning extensions of perception verbs are grounded in interaction

1.1 Where is polysemy?
Synchronic studies of polysemy that focus on a single language or several closely related languages employ a wide range of data-gathering techniques, for example, dictionary searches, elicitation, and participant observation, as well as examining contextualized written or spoken examples (see, e.g., Alm-Arvius 1993;Enfield 2003; Aikhenvald and Storch 2013). For languages with a long tradition of literacy and established broadcast media (and often thanks to committed individuals who work towards creating accessible electronic corpora), polysemy can be approached using quantitative corpus linguistics techniques (e.g.,Gries 2006; Glynn and Robinson 2014), parallel translation (e.g., Enghels and Jansegers 2013; Wälchli 2016), and with a growing range of tools and resources (e.g., FrameNet; Mapping Metaphor 2015). Lexicographic, elicitation, introspective, and usage-based work can further be combined with psycholinguistic and neurolinguistic inquiry (see, e.g., Gries 2015), and semantic mapping approaches (e.g., Haspelmath 2003).
Informal dyadic or multi-party talk is rarely the sole focus of polysemy studies. An alternative approach to the multi-functionality of words and particles — which relies almost exclusively on interactional data — comes from conversation analysis (e.g., Clift 2001; Heritage 2015). This approach generally emphasises sequential action (e.g., what is being done in this turn in the conversation, and how does this relate to prior and subsequent turns?) rather than lexical semantics; and sometimes rejects the notion of “poly-semy” as enshrining a view of decontextualised meaning at the expense of interactional practice (Mazeland and Huiskes 2001: 165; see also Depperman 2011 concerning semantic study in conversational analysis more generally). Several studies in the conversation analytic tradition (or loosely based upon it) have examined discourse uses of perception verbs, discussed below. In a similarly interaction-focused vein, Nerlich and Clarke (2001) argue that polysemy is itself an important resource in conversation, while first language development researchers have examined meanings and functions of perception verbs in conversations involving children (see San Roque and Schieffelin in press).
Broader cross-linguistic inquiries into polysemy have by tradition and necessity largely focused on data from dictionaries and questionnaires. For example, for each language in a study, a linguist might look up (and cross-check) a word that relates to the target concept (e.g., ‘hear’) in a dictionary, and/or ask native speakers or other language experts whether this word can also have such-and-such a meaning (e.g.,Vanhove 2008; Youn et al. 2015). Recent projects have also made use of automated extraction of data from multiple dictionaries and databases (e.g., the CLICS database of cross-linguistic colexifications, List et al. 2014). However, dictionary definitions cannot always capture all current senses of a word (e.g., Fillmore and Beryl 2000), and meta-linguistic intuition varies across speakers. These difficulties can be exacerbated when working with smaller, under-studied languages, as the coverage of available materials varies widely (e.g., Ross et al. 2016: 7–8), and the number of native speakers who are available to advise on their language may be limited. Some parallel corpora may be available for many languages, but relevant words and contexts may be absent from such texts (Wälchli 2016), and differing translation practices and versions bring their own problems (Christodoulouplos and Steedman 2015). Corpora of interactional language are difficult and time-consuming to create, and as such make only rare appearances in cross-linguistic studies of lexical polysemy. Yet we know that informal conversation can differ from other linguistic genres in various ways (e.g., Biber 1986; Dahl 2000), and that differences across genres can relate specifically to perception and related domains (e.g., the frequency of perception verbs as discourse markers, Romero Trillo 1997; differences in the expression of stance and evidentiality, Zariquiey 2015; the frequency of lexical items that relate to certain sensory modalities, Divjak 2015; Winter et al. 2018).
Here, we complement existing work by examining perception verb polysemies as found in conversation, the primary site of language use and language change (Levinson 2006). Through work with a specialized corpus (created by the authors and the researchers named in Figure 1) we enhance the ecological validity and linguistic diversity of polysemy studies. Conversation is the main form of discourse that is shared by all language-culture groups, and gives the opportunity to ground claims in people’s day-to-day linguistic experience. Moreover, by studying informal talk we can gather contextualized examples in comparable discourse that represent both well-known languages and smaller under-studied languages for which sizeable written and broadcast media-based corpora do not exist, and, in many cases, probably never will.