Reviewed by Heiko Narrog, Tohoku University, Japan

This empirical study, based on Anna Wärnsby’s doctoral dissertation at Lund University, compares the English modals must and may with their Swedish counterparts måste and kan. Specifically, W is interested in the polyfunctionality of Germanic modal verbs (e.g. deontic must, which indicates an obligation, in contrast to epistemic must, which indicates a conclusion) as well as elements in the syntactic, semantic, and pragmatic context that help to resolve this ambiguity.

W explores these issues in six chapters. In Ch. 1 (10–47), W discusses the subtypes of modality and introduces the corpus (i.e. the English-Swedish Parallel Corpus, which contains about 2.8 million words). In Ch. 2 (48–59), W presents the findings of a similar study by Jennifer Coates (The semantics of modal auxiliaries, London: Croom Helm, 1983) and identifies a number of contextual factors, including aspect, verb type, and the nature of the grammatical subject, that disambiguate polyfunctional modals.

Ch. 3 (60–112) presents the corpus study. W examines the impact of perfect and progressive aspect, of introductory there subjects, of state verbs, and of inanimate subjects on the interpretation of each of the four modals. In Ch. 4 (113–49), W proposes a refined analysis of the contextual factors discussed by Coates. Crucially, W suggests that all the features that contribute to deontic or epistemic interpretation are related to one overarching concept: controllability. In essence, if a situation described in the proposition can be controlled by an agent with intentions, it is likely to receive a deontic interpretation, and if not, it is likely to receive an epistemic interpretation.

In Ch. 5 (150–207), W presents an analysis of the modals using data mining software. This software allows her to propose a decision tree for each modal and to combine the decision trees of two or more modals. For example, in the decision tree for must, the criterion in the top node of the tree is the presence or absence of marked perfect or progressive aspect: presence leads to an epistemic interpretation, absence leads further down the tree. Although must and måste have very similar decision trees, the trees for may and kan are dissimilar, which W attributes to different degrees of grammaticalization. The final remarks in Ch. 6 (208–215) are followed by references, appendices, and name and subject indexes.

This volume is a welcome addition to the ever-growing literature on modality. W has skillfully refined the information presented in Coates’s important study.