The word degrammaticalization was coined by Christian Lehmann in 1982, in his early work on grammaticalization (published as Lehmann Reference Lehmann1995). Having emphasized the unidirectionality of grammaticalization, Lehmann denied the existence of (non-unidirectional) degrammaticalization. Following Lehmann's claim, as Muriel Norde points out in Degrammaticalization, linguists searched for and found instances of degrammaticalization (although perhaps lacking the degree of theoretical justification of the search to fill the gaps in Mendeleev's table!). Norde's aims in the book under review are (a) ‘to explore the concept of unidirectionality’ (2), and (b) ‘to define and classify degrammaticalization’ (3). Such a book is very much needed.

Chapter 1, ‘Introduction’, introduces the phenomena of grammaticalization, degrammaticalization, and lexicalization. Norde reviews the standard literature on these topics, including the work by Meillet, Heine, Traugott, Fischer, Willis, and Janda, and considers a number of definitions of the three types of change. As a working definition of degrammaticalization, she adopts the following: ‘a single shift from right to left on [the usual grammaticalization] cline’ (8). Accordingly, rather than going from content item to grammatical word to clitic to inflectional affix to a zero stage, degrammaticalization proceeds in the opposite direction, although it does not involve a shift all the way up the cline. Moreover, items may ‘jump over’ an intermediate stage. Norde distinguishes three types of degrammaticalization: (i) degrammation (a function word is reanalyzed as a content item), (ii) deinflectionalization (an inflectional affix becomes less inflectional), and (iii) debonding (inflectional, derivational, and clitic elements become more independent), which is the most common type of degrammaticalization. These three types of degrammaticalization form the subject matter of chapters 4–6.

Unidirectionality and grammaticalization underlie the method of internal reconstruction. In chapter 1, Norde criticizes reconstruction by arguing that, in strongly isolating languages, today's morphology is not yesterday's syntax and that yesterday's syntax may be reorganized by the language learner or that morphemes may become ‘trapped’, so that inflections are reordered or lost (38). However, even in strongly isolating languages like Chinese, we find examples of heavily grammaticalized morphemes, such as the perfective le and the question marker ma, which can no longer appear on their own. Norde returns to the usefulness of unidirectionality and grammaticalization for reconstruction in chapter 2 (‘Unidirectionality’), and poses the question of ‘what would count as sufficient counterevidence to unidirectionality’ (51). Unsurprisingly, her answer is that degrammaticalization is ‘the exception that challenges the rule’ (52), with the inevitable consequence that reconstruction needs to be abandoned as a tool. Norde then discusses conceptualizations of language change, irreversibility, lateral shifts, alternative sources of grammatical markers, allegedly irreversible changes, and explanations of unidirectionality.

The section on conceptualizations of change briefly discusses clines and cycles, presumably as instances of unidirectional change. The author describes the grammaticalization cycle as ‘the loss of synthetic grammatical markers … with the rise of new analytic markers’ (56), using Jespersen's cycle as an example. The chapter also includes a section on lateral, non-directional shifts from noun to verb. It is not altogether transparent that a noun has more content than a verb or an adjective, and Norde's discussion sits somewhat uneasily within this chapter.

In contrast, I found the section on allegedly irreversible changes very helpful. Norde discusses unidirectional changes on various linguistic levels, including desemanticization, decategorialization, univerbation, phonological attrition, and subjectification. Let me focus on just some of these. Semantic bleaching is involved in numerous examples of ‘space to time’ shifts. Norde, using Haspelmath's (Reference Haspelmath1997) example of after, argues that the reverse also occurs, that is, temporal after gets used as spatial after. I have trouble with this argument since, according to the Oxford English Dictionary (OED), after started out with a spatial meaning, which hence could have remained in the language. The OED gives evidence for such continued use, which Norde acknowledges in a footnote (69, fn. 27).

Decategorialization involves loss of autonomy (thanks as a noun to thanks to as a preposition), and a loss of inflections and modifiers. Norde argues that there is a process of recategorialization, evident in cases such as English a must or the ins and outs, which involves the addition of inflectional endings. Norde's conclusion here is a bit lukewarm:

While it is often assumed that everyone knows what phonological attrition means, I think it may be the least well-defined part of grammaticalization. Does it mean a reduction in pitch or length, the loss of an intervocalic fricative, or any of the above and a myriad of other phenomena? For degrammaticalization to occur, the opposite of phonological attrition needs to happen, viz. phonological strengthening (84). However, the same problem of non-specificity is evident. Norde recognizes spelling pronunciation as a case of phonetic strengthening which goes hand in hand with semantic strengthening, giving the example of the definite article in Dutch, which can be phonologically and semantically strengthened.

The last section of chapter 2 concerns explanations for directional tendencies. Norde discusses usage-based, psycholinguistic, and formal approaches. Functional or usage-based approaches emphasize the language user, whether speaker or listener. Norde notes that an inclination towards ‘least-effort’ on the part of the speaker explains ‘the preponderance of grammaticalization changes as compared to degrammaticalization changes’ (93). She then goes on to discuss the psycholinguistic approach but does not make it clear why ‘ease’ could not be subsumed under such an approach. The section on the psycholinguistic approach instead focuses on priming, presenting the work by Rosenbach & Jäger (Reference Rosenbach and Jäger2008), which suggests that the unidirectionality of language change can be linked to the asymmetric priming results known from the psycholinguistic literature (e.g. space can prime time, but not vice versa). This appears to me to be very interesting and promising work and could be integrated with a more formal biolinguistic approach. While Norde, too, finds this approach ‘definitely worth exploring’ (94), she does not think that unidirectionality can be explained in formal approaches, unfortunately adding little to earlier criticism in her statement that ‘[b]ecause a generative view of language acquisition and change cannot accommodate changes that occur in actual language use, it fails to account for both variation … and gradience’ (100). Abrupt major change is no longer as much a staple of generative work on historical change as Norde perhaps assumes (see some of the contributions in Traugott & Trousdale, to appear).

Chapter 3 discusses the question of ‘Defining degrammaticalization’. It recommends using degrammaticalization and not re-/anti-/countergrammaticalization. Most of this chapter is on what degrammaticalization does not involve; the last four pages present a classification of degrammaticalization that is based on Lehmann's (Reference Lehmann1995) six well-known criteria for grammaticalization, viz. integrity, paradigmaticity, paradigmatic variability, structural scope, bondedness, and syntagmatic variability. Instead of a decrease in weight, cohesion, and variability (as in the case of grammaticalization), degrammaticalization involves an increase. I think Norde could have expanded her discussion of Lehmann's criteria, rather than giving a one-page presentation of Andersen's (Reference Andersen2006) stages in the grammaticalization chain, complete with the remark that ‘[a]pplying Andersen's model to all types of grammaticalization is way beyond the scope of this paper [sic]’ (133).

The following three chapters finally turn to examples of actual degrammaticalization, considering ‘Degrammation’ (chapter 4), ‘Deinflectionalization’ (chapter 5), and ‘Debonding’ (6). Degrammation is defined as ‘a composite change whereby a function word in a specific linguistic context is reanalyzed as a member of a major word class, acquiring the morphosyntactic properties which are typical of that word class, and gaining in semantic substance’ (135). Norde surveys the instances of modals that become reanalyzed as full verbs and concludes that English dare and need, Swedish må ‘may’, and Danish turde ‘dare’ do not qualify as degrammaticalizations. In contrast, the Pennsylvania German case of welle ‘want to’ reanalyzed to wotte ‘wish’ does qualify since wotte gained morphosyntactic properties, such as acquiring verbal inflections and an infinitival form and the loss of the ability to take an infinitival complement (Burridge Reference Burridge, Hogg and Bergen1998). Norde's parameter analysis of Pennsylvania German wotte (141, table 4.1), based on Lehmann's six criteria, succinctly shows that there was an increase in weight, cohesion, and variability. Other data discussed here are (i) Chinese dei/de ‘obtain, permit’, which may be developing the new lexical meaning of ‘need, take, require’; (ii) Bulgarian nešto ‘thing’, which is derived from an indefinite pronominal; (iii) Welsh eiddo ‘property’, which stems from a possessive pronoun; and (iv) the development of Welsh yn ol ‘after’ to nôl ‘to fetch’. These data are taken from careful studies by Ziegeler (Reference Ziegeler, Fischer, Norde and Perridon2004) and Willis (Reference Willis2007), with Norde supplying case-by-case scenarios for each of these changes.

The topic of chapter 5, deinflectionalization, is defined as ‘a composite change whereby an inflectional affix in a specific linguistic context gains a new function, while shifting to a less bound morpheme type’ (152). Norde first gives us an overview of what counts as an inflection affix, derivational affix, clitic, and compound. She then turns to the perhaps most famous case of degrammaticalization, the development of the s-genitive from the morphological genitive in English, Danish, Swedish, and Bokmål Norwegian. Norde herself is very well known for her work on the Swedish s-genitive and for using these data to argue for the existence of degrammaticalization. Other lesser known examples of shifts from inflectional to derivational affixes that are discussed in this chapter are the Swedish nominative masculine case marker -er being reanalyzed as a nominalizer (of adjectives), the Swedish number affix -on being reanalyzed as a ‘berry-name suffix’, and the Kwaza quotative. The last of these is particularly interesting, as the author notes, since Kwaza is a polysynthetic language. Basing her discussion on van der Voort (2004), Norde shows how an exhortative (modal) inflectional morpheme -ni has become degrammaticalized to a derivational causative -nĩ. It is a nice example, although Norde's erroneous description of -nĩ as a derivational modality – instead of causative – marker initially leads the reader up the garden path.

Chapter 6 discusses debonding as ‘a composite change whereby a bound morpheme in a specific linguistic context becomes a free morpheme’ (186). Debonding affects clitics and derivational and inflectional affixes in different ways. Clitics and inflectional affixes do not gain new functions or meanings, whereas derivational affixes may become semantically or morphosyntactically enriched. The examples of debonding discussed here are among the best known examples of degrammaticalization: the Irish first-person plural suffix -muid reanalyzing as an independent first-person plural pronoun; the Japanese subordinating conjunction -ga ‘although’ reanalyzing as the independent connective ga ‘but’; the Saami abessive clitic taga ‘without’ reanalyzing as a postposition; the Estonian question suffix -s reanalyzing as the free particle es; the English infinitive prefix to reanalyzing as a proclitic; the Greek prefix ksana- ‘again’ reanalyzing as a free adverb; and the Latin rigid prefix re- ‘again’ reanalyzing as the Italian flexible prefix ri-. Other examples of debonding given in this book are the Norwegian infinitival marker å freeing itself from clitichood, the Hup evidential affix h reanalyzing as a clitic, the Dutch/Frisian/German suffixes -tig/-tich/-zig ‘-ty (as in twenty)’ changing to independent quantifiers, Swedish bö ‘need’ developing into a verb from a prefix, and the English derivational suffix -ish acquiring the status of a free morpheme.

In conclusion, this book provides careful definitions of a variety of connected changes, and gives ample examples of degrammaticalization, for which the author provides thorough analyses. Norde develops a set of criteria for what counts as degrammaticalization and applies these criteria to her case studies in chapters 4–6. I very much enjoyed both the descriptions and the analyses of the degrammaticalization changes.

Degrammaticalization changes are given case-based explanations rather than the more systematic explanations that are possible for grammaticalization, but this observation should not be taken as a criticism of the book as it may reflect a simple necessity. Finally, as a minor point, let me mention that the book has benefited from unusually careful proofreading.