Workshop: Implicational relations and the distribution of lexical properties
- Géraldine Walther - University of Zürich
- Michael Ramscar - University of Tübingen
- Benoît Sagot space - Inria
Work on lexical properties in theoretical linguistics has been largely dominated by two opposing views. One influential tradition, based on the Bloomfieldian view of linguistic units as identifiable regular form-meaning pairings, focuses on the combinatorics of identifiable minimal linguistic units (Chomsky (1957); Lieber (1992); Halle and Marantz (1993), among others). This tradition considers patterns that deviate from regular form-meaning correspondences to be unsystematic ‘lexical’ exceptions that are irrelevant to understanding the associative structure of grammar (Halle, 1973; Di Sciullo and Williams, 1987). A second approach focuses almost exclusively on the diversity of lexical word properties (Gross, 1975; Fillmore and Atkins, 1992; Joshi, 1987), seeing the distribution of forms as being the blueprint of a particular linguistic system. This line of work thus treats the ‘licitness’ of lexical combinations in utterances as a secondary concern (see also, albeit to a lesser extent Bresnan (1982); Gazdar et al. (1985); Pollard and Sag (1994)). The two approaches share a common view of language as a combinatoric system operating over associative form-meaning units, with the different traditions focusing on inventories of lexical properties at different granularity levels.
This workshop focuses on a radically different view of human communication: Languages are discriminative systems. Implicational patterns in language are seen as systems of contrasts that structure both the distribution of individual lexical properties and their paradigmatic organisation. A discriminative perspective has recently been elaborated in learning models (Ramscar et al., 2010, 2013) that incorporate the discriminative learning rule of Rescorla and Wagner (1972). This work has clarified how the non-uniform distribution of lexical properties favours the learnability of linguistic systems, while facilitating the discriminability of forms during linguistic processing (Ackerman et al., 2009; Baayen et al., 2011; Ramscar, 2013; Blevins et al., 2016; Dye et al., 2016; Baayen et al., 2016). Work presented in this workshop will show how investigations based on large-scale data sources reveal a systematic interplay of implicational relations across all levels of a linguistic system and how contrast-based paradigmatic linguistic structure emerges from distributional properties of forms in context. A particular focus will be placed on issues connected with learnability and cognitive processing as well as on possible diachronic pathways.