Toward a Unified Socio-Cognitive Framework for Salience in Language

Surprisingly, linguists have actually relied on at least three of these four scenarios for defining the notion of salience (see also Bowman et al., 2013, for a psychological perspective). Scenario (1) lies at the heart of Giora’s idea of salience as what is “foremost on one’s mind [...] stored and coded in the mental lexicon” (Giora, 2003, p. 15). Scenario (2) accords with Geeraerts’ view of onomasiological salience in terms of “the relative frequency with which a signifiant is associated with a given signifié” (Geeraerts, forthcoming), i.e., the frequency with which a word is used to denote a given piece of experience. Scenario (3) corresponds to understanding salience in terms of surprisal, as, e.g., proposed by Rácz: “A segment is cognitively salient if it has a large surprisal value when compared to an array of language input” (Rácz, 2013, p. 37; see also Friston, 2010; Clark, 2013; Fine et al., 2013; Divjak, 2016). Scenario (4) represents an extreme variant of type (3) which builds onmemory-based novelty rather than context-based surprise (see Barto et al., 2013, for this distinction). The four scenarios can be summarized systematically by a cross-tabulation of two types of sources of expectations, viz. long-termmemory and current context, with two types of mechanisms of salience, viz. confirmation and violation of expectations: