This book arose out of the New Reflections on Grammaticalization 5 conference, held at the University of Edinburgh in 2012. The editors have grouped the chapters into a thematically coherent order, and their introduction provides an informative discussion of the chapters and the issues that they reflect. The introduction is preceded by an interesting historical preface (‘Meillet and grammaticalisation’) by John E. Joseph. The chapters themselves deal with a wide range of spoken and signed languages using a variety of theoretical frameworks; most are well-written and provide detailed descriptions of theoretically interesting phenomena. Each of the contributions is discussed below, taking a thematic approach wherever possible.

Roland Pfau (‘The grammaticalization of headshakes: From head movement to negative head’) argues that negative headshakes in sign languages have linguistic functions rather than being merely ‘co-speech’ gestures. Pfau distinguishes two kinds of headshakes in signed discourse: gestural headshakes, which indicate intensification or uncertainty, and linguistic headshakes, which indicate negation. Following Zeshan (Reference Zeshan2004), Pfau then discusses ‘manual dominant’ sign languages, in which negation expressions are obligatorily manual and a negative headshake only accompanies the manual sign, and ‘non-manual dominant’ sign languages, which allow negation to be expressed solely by a non-manual sign. In addition to this typological division in the function of headshakes, there are robust, language-specific differences in the syntactic scope of headshakes (that is, the manual elements which a headshake can or must accompany). Pfau suggests that in a number of manual dominant sign languages the negative headshake functions as a suprasegmental feature, similar to tone in spoken languages; when the manual negator is obligatorily accompanied by a headshake, they together constitute the lexical sign. When, as in German Sign Language, the headshake accompanies the negated verb (with or without a manual negator), the headshake is treated as a negative affix. Pfau proposes a version of Jespersen’s Cycle for sign languages in which ‘sign languages started out with a purely manual system, where sentences are negated by a manual sign alone’ (41). In this initial system, the headshake functioned as a reinforcer (compare pas in Early French). In the next stage of the cycle, the headshake became phonologically reanalyzed as a component of the negator. Following this, the headshake spread to the negated verb in the form of a bound affix, but still in combination with the manual negator (including headshake as a suprasegmental phonological feature). In the final (predicted) stage the manual negator will disappear completely, leaving the headshake accompanying the verb as the sole marker of negation in the language. Although clearly presented and supported with parallels from Jespersen’s Cycle in spoken languages, the main weakness of Pfau’s proposal is that the end points of his cycle are not currently attested. A crucial feature of Jespersen’s Cycle is precisely that, as a cycle, it may be renewed any number of times. We should therefore expect to find languages exemplifying all the proposed stages. However, Pfau fails to provide examples of either a sign language in which negation is indicated purely through a manual sign and in which negative headshakes – if present – are purely gestural or a sign language in which a headshake is the only negator and no manual negators exist. Aside from this criticism, this paper is a welcome addition to the growing body of work on grammaticalization in sign languages, and is notable for the number of distinct sign languages (twenty one) that are discussed.

The following four chapters deal with discourse markers, but exploit different theoretical frameworks. Gudrun Rawoens (‘The Swedish connective så att ‘so that’: From subordinator to discourse marker’) and Karin Beijering (‘The lexicalization–grammaticalization–pragmaticalization interface: The case of Mainland Scandinavian jeg tror’ [‘I think/believe’]) use synchronic corpus data from written sources (newspapers, plus blogs in Rawoens’ chapter) to investigate syntactic and semantic-pragmatic variation in these expressions. They both analyze the development of these discourse markers as cases of pragmaticalization (Aijmer Reference Aijmer, Swan and Westvik1997) based on their extra-positional (peripheral) syntactic status and subjective communicational functions. María José López-Couso & Belén Méndez-Naya (‘Epistemic/evidential markers of the type verb $+$ complementizer: Some parallels from English and Romance’) also use corpus data to investigate Latin American Spanish epistemic/evidential marker dizque and English like-parentheticals (it looks/sounds/seems like). As with så att and jeg tror, these expressions are peripheral to the clause and convey subjective meanings, and hence can be characterized as discourse markers; however, López-Couso & Méndez-Naya treat the development of these expressions as core cases of grammaticalization (following Thompson and Mulac Reference Thompson, Mulac, Traugott and Heine1991) rather than as cases of pragmaticalization, since, they argue, they can be viewed as intermediate stages in the development of fully grammaticalized evidential systems. Thompson & Mulac’s (Reference Thompson, Mulac, Traugott and Heine1991) analysis is also developed in Tetsuharu Moriya & Kaoru Horie (‘The Neg-Raising Phenomenon as a product of grammaticalization’), which is both the shortest and, for me, the most satisfying of these four chapters. Moriya & Horie investigate the Neg-Raising Phenomenon (NRP), in which negation in a lower clause can be interpreted as negating a higher clause; for example, the NRP reading of (1a) is the same as (1b):

Moriya & Horie note a number of similarities between NRP and parenthetical expressions such as I think and I believe (compare Beijering’s analysis of jeg tror). They note that not only individual subject–verb combinations (I think) but also variant forms (I would think, I thought, etc.) exhibit evidence of grammaticalization, and suggest that this can be explained through Van Bogaert’s (Reference Van Bogaert2011) constructional approach, according to which frequent use of expressions like I think creates a more schematic construction which licenses such variations. They propose that NRP expressions also form part of such schematic constructions, and that their behaviour is therefore linked to the grammaticalization of the schematic construction. This leads to the prediction that, in any given language, ‘the most typical parenthetical predicates will be recruited for NRP predicates’ (130). They test this against the Japanese predicates omou ‘think’ and kangaeru ‘believe’; only omou has grammaticalized a parenthetical, epistemic (subjective) use and, as predicted, it is omou rather than kangaeru that functions as an NRP predicate.

Grammaticalization in the left and right periphery of clauses is the topic of the following two papers, by Yuko Higashiizumi (‘Periphery of utterances and (inter) subjectification in Modern Japanese’) and Alexander Haselow (‘Left vs. right periphery in grammaticalization: The case of anyway’). Higashiizumi traces the grammaticalization of constructions involving the causal conjunctive particles kara and node in the left and right peripheries using a corpus consisting of conversational parts of novels dating from around 1900 and a conversational corpus recorded around 2000. She notes in a footnote that these corpora consist of different genres, but does not indicate whether or how this may have affected the analysis. The corpora are also quite small, and contain no examples of one of the four items under investigation (left peripheral nanode ‘therefore/so’). Higashiizumi invokes ‘(inter) subjectification’ and the development of ‘procedural’ functions in her analysis but without adequately defining these terms, and ultimately she fails to distinguish between the functions of the particles when they occur in the left and right peripheries. Like Higashiizumi, Haselow faces the problem that only written genres are available for historical corpora whereas his modern corpus consisted of spoken language, but he provides an extensive discussion of the steps taken to mitigate the mismatch (159–160). He also invokes ‘procedural’ meaning in his account, but defines this following Blakemore’s (Reference Blakemore1987) Relevance-theoretic formulation (175). His account of the grammaticalization of anyway is based on detailed analyses of contemporary and historical examples, including careful identification of critical contexts and scope relations. Haselow shows that in the left periphery, anyway has a discourse-organizing function of indicating ‘a major reorientation in discourse with respect to ongoing conversational activities or topics’ (182) and enters into a paradigmatic contrast with other transitional expressions such as actually, well and so. In the right periphery, anyway has a retrospective function of indicating the type of relation that exists between the propositional content of the associated discourse unit and the unit preceding it, and contrasts paradigmatically with other retrospective ‘final particles’ such as then, though and actually (181).

The importance of paradigmatic relations is developed further in the final chapter, Jens Nørgård-Sørensen & Lars Heltoft’s ‘Grammaticalisation as paradigmatisation’. The authors view paradigmatic organization as a defining feature of grammatical systems and as encompassing constructional syntax, information structure, and prosody. They argue that many linguists who emphasize the gradualness of grammatical change ‘fail to draw an important distinction, namely that between describing changes in the language of an individual speaker and in the language of an entire speech community’ (268). Only the latter is truly gradual, but this describes the spreading of a change (that is, ‘actualization’, Andersen Reference Andersen2001), rather than the change itself; reanalysis – including the transition from one paradigm to another – occurs in the mind of an individual speaker and must be abrupt.

Grammaticalization is treated as a gradient phenomenon in the preceding chapter, Maria Mazzoli’s ‘Complexity in gradience: The serial verb take in Nigerian Pidgin’. The item take (or tek) functions as a main verb, a serial verb expressing instrumental meaning, a serial verb in three-verb constructions expressing purpose, and (with sufficient contextual support) a modal expressing ability. Mazzoli terms this ‘synchronic gradience’ and lists a number of ‘micro-changes’ (257) that must have occurred to license these historically related functions; it is presumably the gradual adoption of these changes within the community of speakers, resulting in their increased frequency, that allows her to treat this as a gradient phenomenon.

Mazzoli suggests that the serial verb uses of take originated during creolization as a calque from the substrate languages (verb serialization being an areal feature in West Africa), and contact-induced grammaticalization is the focus of the remaining two chapters in the book. Bert Cornillie & Álvaro S. Octavio de Toledo y Huerta (‘The diachrony of subjective amenazar ‘threaten’: On Latin-induced grammaticalization in Spanish’) provide corpus evidence that the Spanish lexical verb amenazar developed subjective meanings owing to the influence of similar expressions in Latin humanist texts. In contrast, Theodore Markopoulos (‘Contact-induced grammaticalization in older texts: The Medieval Greek analytic comparatives’) argues that the grammaticalization of Greek analytic comparatives during the late Middle Ages (12th–16th centuries) resulted from extensive contact between speakers of Medieval Greek and various Romance languages, rather than from the influence of a textual tradition alone. Markopoulos demonstrates the importance of taking extra-linguistic and sociolinguistic factors into account, by building up a coherent scenario for contact-induced grammaticalization driven by first and second language interference in, respectively, Romance and Greek bilinguals.

In summary, the book represents a welcome contribution to the growing body of literature on grammaticalization. Although the contributions are of varying quality, a number (notably those by Pfau, Moriya & Horie, Haselow, Markopoulos, and Nørgård-Sørensen & Heltoft) do succeed in indicating new directions in grammaticalization research.