It is well known that Latin word order is variable, but the factors underlying this variation are not well understood. Spevak's important contribution, which is based in part on her habilitation, aims to shed light on this topic from a pragmatic perspective. The approach is that of Functional Grammar (with its notions Topic, Focus, Tail, Theme, contrast and emphasis), combined with aspects of Functional Sentence Perspective (contextual dependency). Despite the theoretical background, the author insists that the aim is ‘not to apply a theory to Latin constituent order but to try to understand more about it’ (12). Data are taken from Caesar, Cicero and Sallust, though other authors are used when they illustrate a point. S. asserts in the introduction that ‘constituent order is … determined by and indicative of (i) the role of a constituent within the discourse to which its sentence belongs …, (ii) the speaker or writer's estimation of what the addressee knows and expects, and (iii) how important the speaker and hearer consider a constituent within the overall communication’ (1). And this view is reiterated in the conclusion: ‘Latin constituent order obeys pragmatic rules of placement’ (285).

In some cases, S.'s remarks and analyses between the introduction and the conclusion agree with these assertions. For instance, she argues (225) that determinants for noun + adjective sequences are pragmatic. When the adjective is contrastive, emphatic, or part of a contextually dependent noun phrase, it precedes the noun. Adjectives that are unmarked for pragmatic function are usually found post-nominally. But in other cases, these assertions do not hold. For example, in the section on ad te litteras misi it is stated ‘we cannot decide about saliency of a constituent judging from the position it occupies in a sentence’ (140). At the end of the discussion of verbo-nominals (e.g. castra moueo, terga uerto), S. concludes ‘it would be inappropriate to try to establish a relative ordering of the verb and the noun of verbo-nominal constructions because their behaviour is in a direct relationship with their syntactic capacities and their semantic properties’ (131). At another point, S. states ‘there is no one-to-one correspondence between syntactic patterns and pragmatic values; in other words, one syntactic pattern can encompass several pragmatic values’ (116).

The range of grammatical structures studied is impressive, and as such the investigation is more ambitious and ostensibly more systematic than previous functional accounts. I point out a number of issues. First, the sheer number of topics covered means that relatively few Latin examples can be given as evidence (per structure). For example, in her discussion of the analytic passive, e.g. factus est, est factus, etc. (149–54), S. includes and discusses only 4.5 per cent (12/265) of her available data, and rather briefly at that. While it is impossible to discuss all tokens in the text, it would have been useful if the author had included the references for the absent tokens, so that we might scrutinize these for ourselves. Secondly, some topics are not discussed in sufficient detail to merit inclusion. The analysis of the word order in noun phrases exhibiting ‘Fixed Formulas’ (e.g. res gestae, nauis oneraria) is a case in point. It is only sixteen lines long (excluding her three examples), and the author vaguely concludes ‘there are different degrees of “fixedness” of lexical units, and I leave their typology for further research’ (229). This is hardly systematic, and it might instead have been consigned to a footnote or scrapped altogether to make way for more illuminating topics. Some of the material is extraneous. Ellipsis, which concerns not constituent order (the title of the book!) but rather constituent realization more generally construed, is given an analysis of approximately ten pages (96–106). Though the analysis is interesting, the space could instead have been given over to material that actually concerns word order.

The previous literature has been assimilated well for the most part. But ‘the common opinion … that Latin imperative sentences distinguish themselves by having the verb constrained to the initial position [my italics]’ (205) is not found in Devine and Stephens' Latin Word Order. Structured Meaning and Information (2006). They actually say that imperative verbs are ‘often initial’ (2006, 149), ‘but imperative sentences still show a lot of word order variation, which is due to variation in the pragmatic values of their constituents’ (2006, 150, with examples). In her discussion of noun phrases that are combined with a head noun and an adjective as a modifier, S. states (225) that she will base herself on observations made by de Jong in the 1983 volume Latin Linguistics and Linguistic Theory. This is rather curious, given that de Jong's study is not about adjectives, but about other types of modifier.

Finally, while I read this study I kept asking myself about its replicability, which is vital for investigations of this type: will other students of Latin word order be able to base themselves safely on S.'s investigation, using her definitions to identify Topic, Focus, contrast, and emphasis, using the exact same data, and reach virtually the same conclusions? I have my doubts here, because, as with most (qualitative) functional work, S.'s definitions of the pragmatic values are vague to the point of being operationally unstable. ‘[Contrast]’, it is asserted, ‘is a type of confrontation between two elements that are brought into contact with each other. These elements either share some property or differ in some property’ (45). But what exactly does S. mean here? How can one safely go through a text and pick out contrastive constituents objectively based on this definition? S.'s definitions unfortunately foster subjectivity, and not the objectivity required for a rigorous, replicable investigation.

Despite these criticisms, this book is an enjoyable read. It contains important information and insights, and there is a useful appendix containing pragmatic commentaries on three passages. Serious students of Latin syntax must not neglect to consult this rich contribution to the field.