Investigating the Language of Uncertainty - experimental data, formal semantics & probabilistic pragmatics

Abstract

This dissertation reports a series of studies about the language of uncertainty in English. We investigate what we call “uncertainty expressions”, which include both verbal probabilities such as probable and likely and epistemic modals such as might and possible. Moreover, we look at complex, or nested, uncertainty expressions such as certainly likely and might be possible. The issues investigated in this work lie at the interface between semantics and pragmatics, and we attempt to answer questions such as: How do speakers communicate under uncertainty and about uncertainty? And what is the uncertainty intuitively expressed by uncertainty expressions? What do uncertainty expressions actually mean? And what is the role of context when we use them in a conversation? Almost as frequent as logical connectives and quantifiers, uncertainty expressions are ubiquitous in everyday conversations. Unsurprisingly, uncertainty expressions have been extensively investigated by philosophers, logicians, linguists and cognitive scientists. One of the goals of this dissertation is to bring closer together the different traditional approaches to the study of uncertainty expressions. In our investigation, we strive for interdisciplinarity. In doing so, we integrate methods coming from formal semantics and pragmatics with experimental data and computational modeling. The focal point of the dissertation is a novel data-driven probabilistic model of the use and the interpretation of simple and complex uncertainty expressions.