This book is the third volume in the new Oxford series of monographs about topics in the history of English, and like the earlier volumes, its topic is syntactic. Although there is a chapter on the rise of for. . . to infinitives and a chapter on participle clauses, so that the book's general title is fully justified, its main focus, and hence also the main focus of this review, is on the rise and spread of the most recent addition to the English inventory of non-finite clause types, the gerund. The mechanisms involved are not language-specific, which makes De Smet's findings generalizable to the spread of complements in many other languages. This makes this book essential reading for any investigation into complementation patterns, not just for English but for any language.

After a short introductory chapter, the second chapter discusses the corpora available for the study of complementation in the history of English, including the corpora compiled by De Smet himself: the Corpus of Early Modern English Texts (CEMET), spanning the period 1570–1640 and 1640–1710, and the Corpus of Late Modern English Texts (CLMET), which was later extended as the Corpus of Late Modern English Texts Extended Version (CLMETEV), spanning the periods 1710–1780, 1780–1850 and 1850–1920. Although De Smet does not explicitly discuss why existing corpora are not going to deliver the data he needs (beyond his explanation that existing resources focus too exclusively on the output of literate men and hence are bound to distort the picture of usage), the obvious reason is that complementation patterns have not only a syntactic but also a lexical component. The syntactical component includes changes to categorial status – gerunds derive from action nouns and are reanalysed as verbs at some stage in their history – and changes in the expression of the subject arguments ‘inherited’ from the verbal stem, from genitives/possessives to oblique case (cf. his insistence on the correct procedure/ his insisting on the correct procedure/ him insisting on the correct procedure); the lexical component includes the fact that while complementation patterns may show broad trends, with verbs with similar meanings taking similar complements, the question of whether any individual verb will appear with any particular complement not only depends on its meaning but on usage and entrenchment, which in turn depend on frequency – hence the need for large corpora. Consider these examples (from De Smet 2013: 5):

1. (1)

   1. (a) The examples here should set you thinking.

   2. (b) It made Euphrasia think.

   3. (c) If there are any defects likely to cause the house to fall down around your ears, they are not the inspector's concern.

All of these verbs can be argued to be causatives, but they each appear with a different non-finite clause, reflecting which particular complement was productive for that group of verbs at the time they first appeared with a causative meaning. The reason that they kept their original complements must be due to their high frequency.

In time, the regular association with certain groups of verbs lends some sort of meaning to the complements themselves, albeit a meaning that is highly abstract, like e.g. potentiality for to-infinitival clauses. Matching verbs to complements is a bi-directional process: the more verbs a complement appears with, the more general its meaning will become; and the more general its meaning becomes, the more verbs will appear with it. Synchronically, ragged edges, as in (1a–c), will always remain, because of entrenchment. Chapter 3 goes into some of these issues, and presents Construction Grammar as the author's framework of choice. It is clear from the discussion that the framework is not used to inform the hypotheses or the research questions (it is described in such all-encompassing terms that it is in fact difficult to see how it could); De Smet's research questions do not require any additional framework beyond traditional linguistic concepts. The added value of the framework could be its emphasis on usage, but for historical linguists, that is preaching to the converted, as historical data are by definition usage-based. What a Constructional analysis adds, I think, is a recognition that the semantics of a construction may depend on more than the sum of its parts (as is clear from the work on the relocation of meanings in constructions discussed in section 3.2), and a predisposition to a very fine-grained analysis of the way in which these meanings guide a speaker's selection of any particular complement. Construction Grammar in its earlier incarnations was seen to deliver hypotheses for data that straddled (morpho)syntax and the lexicon: constructions could contain open, or half-open, or fixed slots, which could explain both their productivity and idiomaticity (e.g. the one's way construction (Goldberg Reference Goldberg1995), the away construction (Jackendoff Reference Jackendoff1997), resultative complex predicates (Goldberg & Jackendoff Reference Goldberg and Jackendoff2004), impersonal verbs (Möhlig-Falke Reference Möhlig-Falke2012) or phrasal verbs (Los et al. Reference Los, Blom, Booij, Elenbaas and van Kemenade2012). I am not sure whether De Smet's claims that the framework is able to account for all aspects of linguistic behaviour (including extragrammatical factors) are particularly helpful for data that are less fine-grained, or phenomena like e.g. word order change that do not have a strong meaning component.

Much of the discussion in chapter 3 is to my mind reminiscent of the issues of storage versus computation in morphology, and its relevance to verb complementation can be explained by the fact that the relationship between a verb and a complementation type is much like that of a base and a derivational affix in morphology: there is the same idea of two conjuncts, each with a semantic load of its own, combined felicitously depending on (i) a reasonable semantic fit between the two, (ii) entrenchment and (iii) a speaker's communicative intent, which can of course be a desire to innovate (for comic effect, as a means to impress others, etc.). Chapter 4 does in fact contain a section on the relevance of blocking, a concept from derivational morphology, to the diffusion of complementation patterns, and the various mechanisms of analogy that are seen to drive the spread of complements are similarly concepts that naturally apply to morphology.

Chapters 5 and 6 present two excellent case studies of the rise and spread of for. . .to infinitives and participle clauses. Chapter 5 introduces the important concepts of semantic and paradigmatic analogy, which are also much to the fore in chapter 7, and as such will be discussed below. The investigation of integrated participle clauses as in (3) below in chapter 6 demonstrates that speakers recognize synchronic regularities that are highly local and only affect a small subset of a particular complement type (reminiscent, to my mind, of the islands of regularity in the irregular verbs, which may lead to an original weak verb like wear to become strong, on the pattern of bear, swear and tear). An important source for integrated participle clause complements is the adverbial clause as in (2):

1. (2) Up, and to the office betimes, and there all the morning very busy, causing papers to be entered and sorted, to put the office in order against the Parliament (1666, PPCEME (The Diary of Samuel Pepys), De Smet 2013: 115)

After some verbs and adjectives, this adverbial clause was reinterpreted as a complement, i.e. as a constituent that expressed an argument of the higher verb, or adjective, as in (3):

1. (3) He was busy sorting a sheaf of letters.

Predicates like busy, happy or tired acquire this extra argument because there is an additional semantic role lurking in the background: the reason (or source) for being busy, happy or tired. Although the present participle clause originally described the circumstances in which the state arose, which need not be the source, the implication must often have been that they were, and in time this led to the reinterpretation that the participle clause was a complement (De Smet 2013: 121):

1. (4) I am quite busy/happy/tired sorting this sheaf of letters for you.

This finding, too, is generalizable beyond the narrow confines of the history of the gerund. Adjuncts that come to be reinterpreted as complements are a frequent source of complement clauses, both finite and non-finite. An example is López-Couso's (Reference López Couso, Lenker and Meurman-Solin2007) account of the development of the conjunction lest (Old English þy læs (þe), Middle English the lesse the, thi les the, lest). This connective originally meant ‘so that not’, and introduced clauses of negative purpose. It was often used with verbs meaning ‘fear, dread’, and, as with busy, the inference that the clause following such verbs would explain what people were afraid of meant that lest-clauses started to be used interchangeably with that-clauses after such verbs (López-Couso Reference López Couso, Lenker and Meurman-Solin2007: 21):

1. (5) but bycause this texte of sayncte Paule is in latyn, and husbandes commonely can but lyttell laten, I fere leaste they can-not vnderstande it. (HC 1534 Fitzherbert, The Book of Husbandry, 99; López-Couso Reference López Couso, Lenker and Meurman-Solin2007: 14). (Cf.: I fear that they cannot understand it.)

The final case study in chapter 7, on gerunds, is clearly the centrepiece, as it takes up nearly half of the pages of the entire book, and this is the one I will discuss in detail. The first gerunds that appear as verb complements are bare gerunds as in (6); they derive from the Old English -ung/-ing suffix that builds nouns from verbal stems. These early gerunds do not have any modifiers or complements (De Smet calls them ‘bare gerunds’), and hence do not show clear signs of their category, nominal or already verbal:

1. (6) and halde þe in chastite, and iuil langingis do away; luue fasting and hold yourself in chastitie and evil longings do away love fasting (PPCME. a1425; De Smet 2013: 162)

   ‘and keep yourself chaste, and get rid of evil desires; love fasting’

Luue ‘love’ in (6) is one of the first verbs attested with a gerund complement. This verb and the other early gerund-taking verbs share another complement besides the gerund: the abstract noun. Typical examples of such nouns are the vices listed in (7):

1. (7) Jake loves lechery, foul language, war, theft, whoredom and drunkenness.

Present-day English (PDE) examples of bare gerunds after verbs like love usually force subject control, but this is not what we find with these early bare gerunds. They denote generic rather than specific acts, events or situations, and like the bare abstract nouns in (7), the control relations depend on the context. It is probably for the same reason that gerunds do not at first appear in a passive construction with be (as in Jake fears being captured) – instead, we get gerunds that are active in form but passive in sense – Jake fears capturing – by analogy with Jake fears capture. It is from this tiny niche of bare abstract nouns that the gerund takes off. De Smet calls this first stage of the diffusion of the gerund complement narrow paradigmatic analogy. His first examples date from c. 1225.

The second stage involves semantic analogy, in which verbs of Emotion, Avoidance, Necessity and Endurance start to occur with the gerund. The model here is still the bare abstract noun, although indefinite nouns with a generic interpretation are also found, as the (a) examples show; the gerund is still voice-neutral, as shown by the (b) examples.

1. (8)

   1. (a) Jake avoids/escapes/fears/risks capture/punishment/shipwreck

   2. (b) [He] escaped drowning verye narrowely (OED, 1560; De Smet 2013: 174)

2. (9)

   1. (a) In somych (=inasmuch) . . . as an vlcere (=ulcer) is an vlcere, it requireth desiccacion. . . (MED, ?a. 1425; De Smet 2013: 180)

   2. (b) Those who wanted a church consecrating, or a meeting to be held. (OED, 1868)

Endurance verbs are found with bare gerunds in a construction with cannot or could not; note that the conditional in (b) implies a negative: ‘. . . but it could not bear recapitulating’.

1. (10)

   1. (a) He cannot endure/bear criticism/banishment

   2. (b) I would summ up the Particulars of this Second Head, if the Examiner's Performance could bear recapitulating (OED, 1699; De Smet 2013: 195)

Some kind of threshold appears to be reached at this stage: so many verbs appear with gerund complements that users have started to identify coherent groups that share the same semantics, and the bare gerund is gradually extended to verbs that did not themselves collocate with a bare abstract noun, but had similar meanings to these established gerund ‘families’. Verbs of negative implication, which share a meaning component with the endurance verbs in the previous section but do and did not take bare abstract nouns, now start to appear with gerund complements. The gerund is being extended beyond its original model. A typical PDE example is (11):

1. (11) I could not help laughing.

Only one verb of this group provides a link with bare abstract noun complements: the now obsolete verb forbear ‘refrain from’:

1. (12) Quen þaim biheld at kinges here, was nan þat lahuter miht forbere when they beheld the king's army was none that laughter might forbear (MED, a1400, De Smet 2013: 173)

   ‘When they beheld the king's army, none of them could abstain from laughter’

Note that (12) shows that it is crucial to do a diachronic investigation, as the trajectory of the gerund cannot be recaptured on the basis of synchronic data alone.

The extension to new groups of verbs has consequences for the gerund complement itself. Another member of this new group, defer, did not collocate with bare abstract or indefinite nouns but with definite nouns: the search, the journey, the visit, probably because of its basic meaning of ‘postpone’; what gets postponed is usually a plan that was made earlier and is hence identifiable (De Smet 2013: 186). The remaining members of this group – decline, help, omit – do not collocate with abstract nouns, but appear with gerund complements in Early Modern English on the basis of their meaning only. Help is a relative newcomer to this group as it did not have the relevant meaning of negative implication when the group was first formed. This stage takes off at around 1500.

The third stage finds Restrospective and Proposal verbs taking gerund complements. These groups do not include a single member that ever collocated with bare abstract nouns, and the gerund did not appear here on the strength of their meanings either. They represent a significant departure from the original model: these new verbs collocate with definite NPs, and we find here the first non-bare gerunds, especially the type with a possessive, as in (13a). An oblique case appears in (13b). The bare gerund, as in (13c), is a secondary development from the non-bare gerund rather than the original model (as it was for the verbs in stages I and II), a process De Smet terms indirect paradigmatic analogy. His first examples date from the 1660s.

1. (13)

   1. (a) I cannot but remember my Lord's equinimity in all these affairs with admiration. (OED, 1663, S. Pepys Diary 8 Mar. (1971) IV. 69)

   2. (b) I remember/recollect/recall his mother asking him that.

   3. (c) I remember/recollect/recall asking him that.

Proposal verbs, which take definite NPs, as in (14a), now also start to appear with gerunds (14b, c). Note the definite article in (14b) and the possessive marker in (14c). The bare gerund in (14d) is a secondary development:

1. (14)

   1. (a) he was the man that did propose the removal of the Chancellor (1667, CEMET; De Smet 2013: 203)

   2. (b) I to the office, whither Creed come by my desire, and he and I to my wife, to whom I now propose the going to Chetham. (The Diary of Samuel Pepys 1667, CEMET; De Smet 2013: 201)

   3. (c) Mr Warren proposed my getting of £100 to get him a protection for a ship to go out, which I think I shall do. (The Diary of Samuel Pepys 1665, CEMET, De Smet 2013: 201)

   4. (d) I am so sick of it all, that if we are victorious or not, I propose leaving England in the spring (1741, CLMETEV; De Smet 2013: 200)

There is at this stage a broad association between gerund complements and noun phrases in general, not just between bare gerunds and bare abstract nouns: broad paradigmatic analogy. Note that the gerunds at this stage have achieved functional equivalence with a finite clause expressing e.g. propositions. This stage really takes off at around 1740.

Chapter 8 wraps up the conclusions of the book and highlights two questions that emerge from the case studies in chapters 5, 6 and 7: what gives diffusional change its diffusional character, i.e. why does a new pattern not arise in different environments simultaneously? And why do new patterns spread at all – what makes them so unstable? De Smet provides almost a dozen different answers to these questions, but what is at the bottom of all of them, to my mind, is the fact that variation is inherent in the system: there is never a single type of complement that fits all predicates, just as there is never a single derivational affix that fits an entire category of stems. Such variation naturally invites speakers to classify and hypothesize as to what motivates it: ‘a speaker's recognition of one regularity invites changes that give rise to new regularities’ (p. 255). Here, too, parallels with derivational morphology are lurking in the shadows, as De Smet's answers are just as appropriate to the waxing and waning fortunes of derivational affixes. Where complementation patterns differ lies in the fact that the match is not between a stem and an affix, but between items that have an internal complexity that is far more sensitive to additional production constraints like priming or horror aequi. These are highly context dependent, so that ‘a pattern is never sanctioned in the same lexicogrammatical environment to the same degree. This means that the moment a pattern is minimally sanctioned in some environment through the grammar of the language, some occasions are bound to occur when functional and pragmatic factors conspire to give the additional sanctioning boost necessary for the pattern to get selected as actual linguistic output in the environment in question’ (p. 257).

This short review cannot hope to do justice to this excellent book. It is not only meticulously researched but also well written and well argued, and required reading for anyone with an interest in verb complementation.