This book takes a diachronic approach to a variety of forms commonly subsumed under the notion of comment clause (CC). It explicitly excludes some of the most frequent, well-known and widely studied CCs, for example I think and I guess, which have been discussed elsewhere by the same author (Brinton Reference Brinton1996). At the same time, it includes some less widely studied examples, such as CCs with see, say and look among others. The book resembles Brinton's (1996) study in organization and spirit. It is made up of eleven chapters, the first three of which establish the theoretical background and review the relevant literature. The first chapter serves as an introduction to the central notion of the study – the comment clause – and delineates it from related categories. Chapter 2 focuses on the semantic and syntactic development of pragmatic markers. Chapter 3 addresses the processes of change ‘that have been variously seen as accounting for the development of pragmatic markers’ (p. 22), among them grammaticalization. These first three chapters are followed by seven chapters (chapters 4 to 10) devoted to case studies of (sets of) individual CCs. Each of these seven chapters begins by detailing the use and semantic–pragmatic functions of the relevant CCs in Present-day English. Second, starting from Old English, the historical development of their semantic–pragmatic functions and of their syntactic forms is studied, in order to ‘determine when and in what syntactic contexts these functions first arose and how they were related to one another’ (p. 19). Third, the development is situated in the context of grammaticalization and related concepts. The empirical work reported in these chapters is based on a variety of online and electronic corpora for the periods of English from Old English (OE) to Present-day English (PDE) (see p. 20). While ‘frequency and distribution of comment clauses across corpora [is taken as] evidence for their semantic and syntactic development’ (p. 19), the approach is qualitative rather than quantitative. Several case studies are based on previous work by Brinton, including the chapters on (I) say (chapter 4), I mean (chapter 5), CCs with look (chapter 8), and what's more (first half of chapter 9).

Chapter 11 of the present monograph serves as a summary, conclusion and outlook. I will now look in more detail at the analysis.

Chapter 1, ‘Introduction: comment clauses, parentheticals, and pragmatic markers’, begins with a definition of pragmatic marker as ‘a phonologically short item that is not syntactically connected to the rest of the clause (i.e. is parenthetical), and has little or no referential meaning but serves pragmatic or procedural purposes’ (p. 1). CCs are situated ‘among the larger categories of sentence adverbial and disjunct’ (p. 2). The chapter also includes a review of the literature on the syntax, semantics and prosody of parentheticals (chapter 1.5) and of the literature on pragmatic markers (chapter 1.6). Brinton concludes that CCs are best understood as epistemic/evidential parentheticals which are clausal in origin and which function as pragmatic markers.

Chapter 2 reviews the literature on the semantic development of pragmatic markers before it studies their syntactic development in some detail. Based on much previous work, the chapter outlines the paths of development of CCs and many other members of the class of pragmatic markers (e.g. well, right, like, indeed, in fact) from numerous sources, such as single words (adverbs, prepositions), phrases and clauses (matrix clauses, adverbial clauses, relative clauses). This chapter also introduces the influential matrix clause hypothesis, which holds that sentence-initial CCs ‘are grammaticized forms of subjects and verbs introducing complement clauses’ starting with that (Thompson & Mulac Reference Thompson, Mulac, Traugott and Heine1991: 317). This hypothesis, being ‘the only one that has been proposed in the diachronic context’ (p. 6), plays a central role in the discussions of the syntactic pathways of the CCs in the subsequent case studies, where it is shown that in the light of historical data, the matrix clause hypothesis often fails to account for the facts. The essential conclusion of chapter 2 is that despite the variety of sources and pathways in the development of pragmatic markers, the common feature is the widening of scope in a unidirectional way. The semantic–pragmatic development and the syntactic change go hand in hand. With respect to meaning, the scope of the element in question goes from scope within the proposition, to scope over the proposition, to scope over discourse. In a parallel way, the syntactic scope develops from scope within a clausal element, to scope over clausal elements, to scope over more global discourse elements.

Chapter 3, ‘Processes of change’, sets out ‘to reassess whether grammaticalization . . . is indeed the process that underlies the development of pragmatic markers’ (p. 8), as has been argued in much of the literature. Essentially, this chapter is a very useful review of the literature in the relevant framework. It provides the reader with a background on the theory of grammaticalization, including milestones such as Lehmann's parameters of grammaticalization, Hopper's principles of grammaticalization, Heine's context generalization and Himmelmann's context expansion. It sets pragmatic markers in general and CCs in particular in the context of the theory of grammaticalization. Taking account of the suggestion that pragmatic markers undergo processes of pragmaticalization, lexicalization, idiomatization, and subjectification and intersubjectification, rather than grammaticalization, the chapter also introduces the reader to relevant aspects of these processes. Having reviewed the theoretical background, Brinton establishes a major claim for the case studies: that CCs are ‘the result of grammaticalization, in large part because they undergo decategorialization, lose referential meaning, and acquire functional (pragmatic) roles’ (p. 72).

The subsequent chapters report on systematic case studies of the following CCs: CCs with say, specifically (I) say, I daresay, (as) you say and that is (to say) in chapter 4; I mean in chapter 5; CCs with see, specifically you see, as/so you see, and see in chapter 6; if you will and as it were in chapter 7; CCs with look in chapter 8; what's more and what else in chapter 9; and I gather and I find as representatives of first-person epistemic/evidential parentheticals in chapter 10. In chapter 11, ‘Concluding remarks’, the various case studies are brought together with a focus on the various syntactic pathways.

The case studies proceed from the three basic assumptions outlined in the first three chapters and reviewed in chapter 11: (i) CCs are parenthetical pragmatic markers; (ii) CCs are the result of a process of grammaticalization; and (iii) the development of CCs cannot be fully and invariably accounted for in terms of the matrix clause hypothesis. In what follows, I will focus on these three hypotheses and the evidence provided.

(i) CCs are parenthetical pragmatic markers. Brinton maintains that ‘a well-recognized quality of pragmatic markers is their multifunctionality’ (p. 18). Features shared by CCs and other pragmatic markers include the lack of propositional and referential content, pragmatic and procedural functions, their scope over the entire clause, textual and interpersonal functions, their optionality and syntactic freedom, frequent occurrence in initial position, and their occurrence in oral and speech-based discourse (p. 241).

The following features are considered essential to parentheticals (chapter 1.5): they are syntactically unintegrated and thus mobile with regard to their syntactic position, they are marked by comma intonation, they function as comments, and they have non-parenthetical uses. Brinton also claims that ‘[i]n non-initial position, forms such as I think or you know are unambiguously parenthetical’ (p. 12; see also pp. 240f). While I agree in principle that CCs are parentheticals, I would like to address two of these features. First, as has repeatedly been argued at least since Crystal (Reference Crystal1969), certain types of relatively short parentheticals such as CCs, reporting verbs, question tags and vocatives may be prosodically integrated into an adjacent domain rather than separated from the host by comma intonation. Thus, while comma intonation may be a salient feature of other (longer) types of parentheticals, it is not a salient feature of CCs (see also Dehé Reference Dehé2009). Second, I do not agree that non-initial (i.e. sentence-medial and sentence-final) parentheticals are unambiguously parenthetical. If they are preceded by a full constituent (e.g. a subject, a full clause, or a wh-phrase in a relative clause) the word order may be the result of syntactic operations such as topicalization. Syntactic tests may reveal their true status. While I do not believe that this caveat would affect the overall analysis, the possibility that even the status of non-initial CCs may be ambiguous should in my view not be ignored in the data analysis.

(ii) CCs are the result of a process of grammaticalization. Brinton follows earlier work on pragmatic markers by Traugott (Reference Traugott1995) and Brinton (Reference Brinton1996) and on lexicalization vs grammaticalization by Brinton & Traugott (Reference Brinton and Traugott2005) in that she considers CCs the result of a process of grammaticalization, which encompasses pragmaticalization, idiomaticization and subjectification, but not lexicalization. This is systematically studied for each CC under investigation. To take but one example, Brinton argues that I mean (chapter 5), like other CCs, undergoes many of the changes identified with grammaticalization: it is decategorialized and desemanticized such that mean loses the properties of a full lexical verb, including its full semantic meaning and develops ‘from major to minor word class’ (p. 130); I mean undergoes pragmaticalization such that it takes on a number of discourse functions. It further undergoes formal freezing, resulting in an invariable first-person singular present tense form, and also some degree of coalescence and phonological reduction. Over time, it acquires non-referential meanings, and it is subject to subjectification and intersubjectification. Finally, I mean exhibits Hopper's principles of divergence (i.e., it continues to be used with literal meaning), persistence and layering (p. 131).

Decategorialization is the feature which according to Brinton & Traugott (Reference Brinton and Traugott2005) most clearly distinguishes between grammaticalization and lexicalization. While category changes in lexicalization may go both from minor to major and from major to minor, category changes in grammaticalization are unidirectional from major to minor. Changes from major to minor are also characteristic of the development of CCs, for example from verb to an adverbial- or particle-like element.

Among the features of grammaticalization consistently not shared by CCs (or any other pragmatic marker) is Lehmann's parameter of coalescence, or scope reduction. CCs typically increase their scope from scope over just a part of the sentence to scope over the whole sentence, to scope over larger discoursal units. This is also a central feature of subjectification and intersubjectification. Instead of ‘rejecting the view of comment clauses and pragmatic markers generally as undergoing grammaticalization’, Brinton, based also on previous research, suggests ‘that Lehmann's parameters are salient but not necessary features of grammaticalization’ (p. 244).

(iii) The development of CCs cannot be fully and invariably accounted for in terms of the matrix clause hypothesis. The case studies show – convincingly in my view – that the matrix clause hypothesis, according to which CCs develop from matrix clauses via deletion of the complementizer that (e.g. I think that the world is flat > I think the world is flat > The world is, I think, flat/ The world is flat, I think; p. 246) is not plausible in light of much of the historical evidence. The main reason is that the source construction, a matrix clause with a following that-clause, is often rare in earlier stages of the language (p. 249), and if it occurs in earlier periods, it is often a minority form. Furthermore, the occurrence of that-deletion does not necessarily increase over time, as would be expected if that-deletion indicates grammaticalization in progress. In some cases, parentheticals pre-date matrix clauses, thus making it implausible that parentheticals develop from matrix clauses. For some CCs, other types of verbal complements (interrogative, imperative, phrasal) may outnumber that-complements.

One alternative to the matrix clause hypothesis is what Brinton refers to as the parenthetical hypothesis, which holds that ‘the parenthetical begins as a syntactically independent utterance and is then inserted into the anchor clause with which it bears no syntactic relationship’ (p. 11). Brinton rejects this hypothesis on independent grounds. She argues that it fails to account for ‘the syntactic incompleteness of comment clauses, or the absence of obligatory complement structures in cases such as I say or I mean’ (p. 250). A few comments are in order. First, the problem of the syntactic incompleteness of CCs has long been recognized in the literature and has been accounted for along various lines, some of which are in fact outlined in chapter 1 (see pp. 11–12). Moreover, syntactic incompleteness is not a feature unique to CCs and related reporting verbs. Elliptical parenthetical clauses are not uncommon. In the example For those of us who remember nineteen sixty-five one or two of our listeners may Tory party uh leadership contests used to be uh as the cardinals in Rome and leaders would emerge (ICE-GB: S1B-024 001) the parenthetical is clearly clausal, yet it lacks a verb phrase. This can be recovered from the host clause. Similarly, the missing complements of CCs such as I find, I mean and I think can be recovered from the host. This may be inconsistent with the assumption that parentheticals are outside the syntactic structure of the host (although suggestions have been made in previous literature to remedy this inconsistency), but it is not unique to CCs. Second, it seems implausible that the syntactic incompleteness should pose a problem for the parenthetical hypothesis, but not for any other hypothesis which considers CCs parentheticals and which considers parentheticals outside the syntactic structure of the host. Whether a CC develops from an independent utterance or whether it starts out as a subordinated clause or as a phrase, the fact remains that it ends up as incomplete and independent, a structure which must ultimately be accounted for. In other words, if anything, the syntactic incompleteness of CCs and other parentheticals is a problem for the assumption that parentheticals are outside the syntactic structure of their host; it is not a problem for the parenthetical hypothesis as defined above.

What is more convincing – especially as an argument against the matrix clause hypothesis – and more important, is Brinton's finding that the syntactic source constructions for CCs are numerous and diverse. They are conveniently summarized, along with the time lines for their development, in figure 11.2 (p. 248). For example, imperative matrix clauses are the syntactic source of a number of bare forms such as say (chapter 4), see (chapter 6) and look (chapter 8), among others. Adverbial clauses are the syntactic source of CCs such as if you will and as it were (chapter 7), so you see (chapter 6), and of first-person epistemic/evidential parentheticals such as I think (Brinton Reference Brinton1996: ch. 8). (Non-restrictive) relative clauses are the syntactic source of CCs such as (as) you say (chapter 4), (as) you see (chapter 6), as I find (chapter 10) and what is more (chapter 9). Needless to say, the CCs vary in the exact pathways from source construction to PDE-CC. Other potential sources for CCs include interrogative clauses and interrogative tags (p. 253). CCs whose development can possibly be accounted for along the lines of the matrix clause hypothesis include I find, I gather, I say, I daresay, I see, I mean, you see, say, see, look. However, note that for a number of these forms Brinton also identifies alternative sources (e.g. for the bare forms say, see, look and you see). Moreover, the evidence in favour of the matrix clause hypothesis for I gather is not convincing. In chapter 10.3.3 (pp. 228–30), Brinton maintains that ‘[t]he syntactic origin of parenthetical I gather is somewhat unclear’ (p. 229). Both the adverbial clause as I gather and the matrix clause I gather (that . . .) are argued to be unlikely sources on the grounds of empirical evidence. She then states similarities between sentential relative clauses with gather and parenthetical I gather, before declaring that the matrix clause hypothesis might be ‘the best account’ (p. 230). Given that other first-person epistemic/evidential parentheticals such as I think and I guess have been argued not to be derived from a matrix clause, it is hard to see why the author resorts to the matrix clause hypothesis for I gather without solid evidence.

When I first picked up the book I was disappointed to see that a large group of frequent (and frequently studied) CCs, above all I think and I guess, were not among the case studies discussed here. It is true that these CCs have been discussed elsewhere by the same author (Brinton Reference Brinton1996: ch. 8), but it is also true that they are seen, by many researchers, as prototypical CCs, and therefore simply belong in a full-length monograph on English CCs. Moreover, given that other chapters in this book have also been based on previous work by the same author, this alone does not seem a good reason to exclude these elements from the present study. At the very least, it would have been helpful to include I think and I guess in the overview of source constructions and pathways in figure 11.2 (p. 248), in order to provide the reader with an even more complete picture of the development of CCs and to allow for an even more thorough comparison between the various CCs.

In my view, the merits of this study are as follows. First, to the best of my knowledge, it is the first full-length diachronic approach to CCs. In the past, the key research questions relating to the semantic–pragmatic functions, the syntax and the historical development of CCs have mostly been approached from a synchronic perspective. Using a variety of historical and present-day corpora, Brinton systematically studies the development of a number of CCs. Second, the analysis includes CCs that have only rarely been studied in previous literature. Third, it establishes the development of CCs as a case of grammaticalization. While this assumption has been around for a relatively long time (e.g. Thompson & Mulac Reference Thompson, Mulac, Traugott and Heine1991), what is new here is that Brinton explicitly argues for grammaticalization – encompassing pragmaticalization, idiomatization and subjectification/intersubjectification – as opposed to lexicalization, and that the present discussion is based on solid historical evidence for a large variety of CCs. Here, the value of the book goes beyond its contribution to the discussion of CCs. The systematic study of CCs for each parameter and principle in a theory of grammaticalization and related processes of language change makes it a valuable contribution to this field in general. Fourth, from a syntactic perspective, and again based on solid historical evidence, the study scrutinizes the influential matrix clause hypothesis for the development of CCs (e.g. Thompson & Mulac Reference Thompson, Mulac, Traugott and Heine1991). Brinton's findings suggest that only relatively few forms can safely and unambiguously be accounted for in terms of the matrix clause hypothesis, thus arguing against an established, but never historically well-grounded, analysis.

Overall, I consider this book an important addition to the literature on CCs. Researchers working on the phenomenon of CCs will value its thorough historical approach and the fact that it includes a number of CCs which have to date not been very well studied. Researchers working on grammaticalization will value the book as a contribution to the discussion of the various processes of change and their delimitations.