Is 'good' better than 'excellent'? An experimental investigation on scalar implicatures and gradable adjectives

Abstract

In this study we look at the computation of scalar implicatures with adjectives and modals, and show that scalar inferences are not equally triggered by these scales. In Experiment 1, we test implicatures on three-place adjective scales such as <decent, good, excellent> and find that only adjectives at the low end of the scale generate implicatures, while those in the middle fail to do so. In Experiment 2 we test implicatures on the modal scale <possible, likely, certain>, and show that, contrary to what happens to adjectives, both modals at the low end and in the middle of the scale generate implicatures. We interpret these results as suggesting that adjective scales are qualitatively different from modal scales. What distinguishes them, in particular, is the nature of the extreme term on the scale. While extreme adjectives do not constitute a real scalar maximum and make use of different semantic representations from their non-extreme counterparts, extreme modals are real scalar maximums and are semantically homogenous with respect to the other terms on the scale. We conclude by providing two alternative explanations of how the properties of extreme adjectives might inhibit the calculation of the inferences.