The role of scalar diversity and question under discussion in deriving impli- catures with embedded scales

Abstract

We investigate experimentally the role of scalar diversity and question under discussion for implicature rates of sentences with multiple scalar terms such as 'Some meals are adequate' with embedding scale ⟨'all', 'some'⟩ and embedded scale ⟨'good', 'adequate'⟩. These sentences can trigger different types of scalar implicatures. We modified the inference task by van Tiel et al. (2016) and tested the 43 scales studied by them in a position embedded under 'some' and 'possible'. We were particularly interested in whether implicatures involving embedded scales can be boosted if made relevant by 'Questions under Discussion' (QUDs). Our results showed that all tested types of implicatures are sensitive to QUDs. Most interestingly, the contrast between 'bounded' and 'unbounded' scales, which was a strong predictor in previous studies, no longer correlates with rates of implicatures once a QUD is added. We argue that our findings support a version of the Alternatives-based Account (the Contextual Alternatives and Scalar Distinctness Account) where contextual availability of alternatives is more important than lexical availability, and where, additionally, the (lexical/contextual) distinctness of the scales plays a role.