Exclusives vary in strength and scale structure: experimental evidence

Abstract

This paper is an investigation of parameters of variation across English exclusive modifiers. We report on three experiments that test the robustness of exclusionary inference calculation (e.g., merely intelligent → not brilliant) across a large number of different lexical scales. Our findings reveal that 1) just excludes less robustly than only, 2) while only allows both complement-exclusion and rank-order readings, merely prefers rank-order ones, and 3) just and only are equally QUD-sensitive. We discuss these results in light of existing theoretical proposals about the semantics of exclusives.