Language users rely on their words and phrases having relatively fixed interpretations – semantics –  and they also rely on a common level of rationality when it comes to achieving communicative ends with language – pragmatics. At the very interface between semantics and pragmatics lies Scalar Implicature –inferences about alternative words or phrases to what the speaker actually used. While scalar implicature is seen as key to our understanding of several linguistic constructions across languages, a theory of how alternatives are computed remains elusive. Building on recent critical work, the project team will undertake an unprecedented experimental investigation into how alternatives are computed.

Scalar Implicature is the single most studied phenomenon in recent semantics/pragmatics research – not only in Linguistics but also psycholinguistics, developmental psychology, philosophy, computational linguistics and elsewhere. There is a growing recognition that Scalar Implicature plays a role in peoples’ understanding of the meaning of a wide range of linguistic constructions, across the world’s languages. Such constructions include the plural, questions, tense and temporal adverbials, quantifiers and connectives. It is thus a fundamental mechanism humans use to derive meaning from the languages they speak.

It is virtually universally agreed that deriving a Scalar Implicature involves an Alternative. It is also widely assumed, based on research on children’s pragmatic development, that an individual’s ability to recognise the correct alternative is a critical factor in their successfully deriving an Scalar Implicature, explaining developmental and other individual difference data.

In spite of the great interest in Scalar Implicature over decades and the recognition of the key role played by Alternatives, there has been no satisfactory answer to the question why some potential Alternatives are chosen and not others. Recent work by the applicants reviews two promising approaches to computing the correct alternative, highlighting weak points of both approaches. One of these takes the view that Alternatives are primarily constrained by linguistic structural factors; the other takes a Bayesian/probabilistic approach and is capable of incorporating several factors beyond purely structural considerations. To date, theoretical claims have been based on introspective judgements which lack the support of controlled experimental data. In spite of the existence of hundreds of studies on SI, only one or two previous studies have set out to systematically explore the facts about Alternative selection.