Gradience in linguistic data

Abstract

This paper provides a survey of the theoretical and experimental findings on degrees of grammaticality, with a special focus on gradience in syntax. We first discuss the theoretical relevance of gradient data, and argue that such data should be elicited experimentally in order to be reliable. We then review a set of experimental findings on gradience, which lead to the hypothesis that linguistic constraints come in two types: hard constraints whose violations trigger strong unacceptability, and soft constraints that lead to only mild unacceptability. There is experimental evidence that both types of constraints are subject to constraint ranking and show cumulativity and ganging up effects. However, soft and hard constraints differ with respect to context effects, crosslinguistic variation, and developmental optionality. This set of properties forms the basis for an evaluation of existing proposals for models of gradience in grammar, with a particular focus on models that extend Optimality Theory.