This is a grammar of British English for those concerned with English as a Second Language (ESL). It results from a seven-year research project ‘informed by’ the Cambridge International Corpus of 700 million words, including the Cambridge and Nottingham Corpus of Discourse in English (CANCODE) containing 5 million words of spoken English.

The book contains an Introduction (19 pp.); an A–Z component dealing with over a hundred individual words, or groups of words (141 pp.); 33 unnumbered chapters grouped into sets named ‘Spoken language’ (78 pp.), ‘Grammar and discourse’ (54 pp.), ‘Word and phrase classes’ (22 pp.), ‘Nouns’ (77 pp.), ‘Verbs’ (43 pp.), ‘Adjectives and adverbs’ (24 pp.), ‘Prepositions and particles’ (9 pp.), ‘Word formation’ (15 pp.), ‘Sentence and clause patterns’ (112 pp.), ‘Time’ (40 pp.), ‘Notions and functions’ – modality, speech acts, etc. – (140 pp.), ‘Information packaging’ (48 pp.); nine appendices – ‘Punctuation’, ‘North American English grammar’, etc. – (64 pp.); Glossary (41 pp.); Bibliography (5 pp.); Index (37 pp.).Footnote ¹ The book can be bought alone or with a useful CD-ROM containing the text in searchable format, audio recordings of examples and links to The Cambridge Advanced Learner's Dictionary. The ESL orientation of the book is reflected in the inclusion of numerous helpful ‘error warnings’ drawing attention to mistakes commonly made by ESL learners.

The ordering of the material means that the basic description of the grammar does not begin until p. 295, though a good deal of it is inevitably presupposed in the early chapters. It seems likely that this order of presentation is motivated by the wish to give prominence to the discussion of spoken English, for Carter & McCarthy (henceforth C&M) regard it as a major feature of the book that the CANCODE corpus has enabled them to avoid the bias towards the written language that is found in most books on English grammar (p. 9). It is certainly good to have a large number of examples from a spoken corpus, and the Spoken Language set of chapters provide a useful account of features that are characteristic of speech rather than writing. However, C&M do not deal satisfactorily with the problem of dysfluency and performance errors in spontaneous speech.

In their discussion of different kinds or degrees of acceptability (pp. 5, 168, 235) they have a category ‘non-occurring and unacceptable in all varieties of British English’ but they do not include the indispensable category of ‘occurring but nevertheless unacceptable in all varieties’. They recognise clausal ‘blends’ such as In fact, that's why last year they rented a nice house, in er Spain, was it, is that it was near the airport, saying that they are ‘usually communicatively complete, effective and easily understood’ (p. 171), and go on to say that such recastings are ‘perfectly normal in spontaneous speech, and . . . not to be taken as a sign of sloppy or lazy performance’ (p. 173). That, however, doesn't make them ‘acceptable’ in the sense in which that term is standardly used in grammar. It won't do to put such examples in the category ‘acceptable in standard spoken English’, with the same status as, for example, The teacher with glasses, he seems very nice (p. 193). And in practice C&M's account of the grammar does not in fact cover such blends and recastings. Moreover, they explicitly mark as unacceptable certain structures which undoubtedly do occur. For example, they say that the ‘head noun of a subject noun phrase determines number [of the verb], not other nouns in the noun phrase’, thus excluding *The general quality of supermarket vegetables are very poor (p. 496). But it is well known that violations of this rule are quite often attested (in both writing and speech), as in *In this case a woman may continue to use both names provided the use of both commonly used names are disclosed (from a credit union's rules). The mere fact that someone wrote this doesn't suffice to make it acceptable: the writer simply made a mistake. Such examples thus do not invalidate C&M's rule, but they do show that we have to allow that occurrence is not a guarantee of acceptability.Footnote ²

I turn now to the core of the book, the grammatical description and analysis, and have to say at the outset that it is characterised by a quite amazing degree of inconsistency. Consider, as an initial example, the first two sentences of the section on tags (p. 547):

1. (1) Tags are a type of clause without a lexical verb but which relate to the verb in the main clause of a sentence.

   Tags consist of auxiliary be, do, have, lexical verb be or a modal verb and a subject (most typically a pronoun).

The first sentence says a tag has no lexical verb while the second says that it may contain the lexical verb be; six of the ten illustrations contain this be and hence are counterexamples to the first statement in (1). This example is untypical only in the immediate adjacency of the contradictory statements. I have never seen a large-scale work on English grammar with anything remotely comparable to the amount of inconsistency and confusion to be found in the present work. It seems evident that the authors have not first worked out and assimilated an outline grammatical account of the major constructions and classes of English which they could then systematically apply in writing up the various chapters of grammatical description. The work is clearly intended as a major publication in the ESL field, and the CANCODE project is still ongoing: the authors envisage subsequent editions of the book incorporating the results of further research. For these reasons I believe it is appropriate to substantiate the above harsh criticism by demonstrating in detail how serious and pervasive the deficiencies are, though limitations of space mean that the account must be very far from exhaustive.

Verb inflection. In the introductory discussion of verbs (p. 302) C&M distinguish six forms: base form, present form, -s form, -ing form, past form, and -ed participle; on the same page they use ‘-ed form’ as a variant of ‘-ed participle’. Apart from the issue of terminology, this six-form analysis is, I believe, the right one for all except a small number of verbs. But this is not the analysis that they follow in the rest of the text.

Most importantly, they present on p. 421 a table incorporating a quite different account. It distinguishes between ‘forms’ and ‘functions’, and instead of the earlier six forms, the table has just four: base form, -s form, -ing form, and -ed form. The -ed form is assigned four functions: past tense, -ed participle (as in I've looked everywhere), non-tensed in non-finite clauses (as in Looked at in that way, it seemed less problematic), and -ed form adjectives. Thus -ed form is here used more broadly than before, covering the past form as well as the -ed participle/-ed form of the earlier analysis (and also adjectives). Conversely, -ed participle (now a function) is used more narrowly, being restricted apparently to uses of traditional grammar's past participle that are preceded by the perfect or passive auxiliary; this restriction, however, is not maintained elsewhere: for example, buried is called an -ed participle in They want him buried there (p. 526). The fact that with many irregular verbs the past tense function of the -ed form is overtly distinct from the other functions is handled by confusingly using ‘form’ in a second sense, so that come,Footnote ³ for example, is said to ‘have the same form for the base and -ed form as -ed participle/adjective/non-tensed [namely come], but a different form for the -ed form as past tense [namely came]’ (p. 421).

The table gives only two functions for the base form, present tense (a distinct form in the earlier analysis) and the infinitive: unaccountably it omits the imperative and subjunctive uses. It is true that these are mentioned on p. 422, but they are again omitted in the glossary definition (p. 893) – and also on p. 398. Furthermore, the distinction between forms and functions is not maintained later. For example, while C&M say on p. 422 that the base form ‘functions as the subjunctive mood for all persons’, they elsewhere talk of the ‘subjunctive verb form’ (e.g. p. 509).

Two further problems arise in their treatment of the subjunctive. First, in the ‘Modality’ chapter they analyse the came of I'd rather she came on Tuesday than Monday as a ‘past subjunctive form’ (p. 669). This is of course traditional grammar's analysis, but C&M must have forgotten that the accounts of verb inflection they present elsewhere differ radically from that of traditional grammar and in particular do not recognise a past subjunctive form (or ‘function’): the analysis is thus inconsistent with the rest of the book, where traditional past subjunctives, other than were (and modals), are treated as ordinary (indicative) past tense forms. Secondly, C&M say that ‘the subjunctive form of the verb be may occur as the base form be or as hypothetical were’ (p. 308). But be and were are not variant realisations of a single ‘subjunctive form’; if (erroneously, in my view) they are both regarded as instances of the subjunctive mood, then it is necessary to distinguish two subcategories of subjunctive. C&M don't do this, but elsewhere generally forget about were, using ‘subjunctive’ to apply exclusively to the base-form type – and that is how it is defined in the glossary (p. 924).

The chapter on word structure and word formation contains a section on inflection less than a page long (p. 473), but there are two points in it that are inconsistent with the rest of the book. Firstly, they list -n't as a negative verb inflection, so that can't and won't are negative forms of can and will. This is fine, but elsewhere in the book such forms are treated as contractions, just like I'm or she'd – note, for example, the statement on p. 639 that modals have only one form. Secondly, they treat advice as an inflectional form of the verb advise: it is included, along with pairs like hung and hang, as an instance of inflection through vowel or consonant change. This is inconsistent with the statement on p. 297 that forms of a single lexeme must belong to the same word class and can only be regarded as a thoughtless mistake.

Finiteness and tensed verbs. A finite clause is defined as one which has a ‘tensed verb form’ (p. 532). There is, however, inconsistency and confusion over what counts as a tensed verb. There are two cases to consider: modals on the one hand, and base-form subjunctives and imperatives on the other.

C&M say repeatedly that modals don't inflect for tense (e.g. pp. 303, 398, 405, 896, 928), but having said on p. 398 that might break has no tense they go on, on the very next page, to cite They may get here by six o'clock as an example containing a tensed verb phrase – and to say that ‘core’ modals do not occur in non-tensed verb phrases! And they treat clauses containing such modals as finite. When they say on p. 303 that modals don't inflect for tense, they add that modals ‘only display historical remains of tense contrasts, such as can−could, will−would’. It is a serious mistake, however, to dismiss the relationship between pairs like can and could as merely ‘historical’, irrelevant to their inflectional status in Present-day English. Not only does this lead to the contradictions just noted concerning tensed verbs and finite clauses, it leads to further inconsistencies and complexities. Thus they say on p. 644 that could ‘is used as the past tense form of can when clauses with can are reported as past time events’, as in such examples as We could hear that dog barking all through the night. They are right in saying that this could is a past tense form, but they cannot simultaneously claim that the modals don't inflect for tense. Or take the glossary entry for ‘backshifting’, where they say that ‘when an indirect report is perceived as referring to the past, the tense in the reported clause usually changes to a past form of the tense of the direct report’. One of the examples given has can changing to could, which is compelling evidence that could is the past of can. All these contradictions involving the modals can be avoided if we simply recognise that they are indeed tensed verbs, with the past tense forms, however, displaying a quite considerable amount of irregularity in their meaning and use.

Consider next the case of the base-form subjunctive and the imperative. C&M treat these clauses as finite, but the verb forms are explicitly said to be non-tensed on p. 305. This time we can't remove the inconsistencies by saying that the verb forms should be treated as tensed, for there is compelling evidence against the traditional analysis of the base-form subjunctive as a present tense form. It does not undergo backshift, and it can occur in contexts where a genuine present tense form is inadmissible, as in They insisted that he finish the job before he went home. Thus on the assumption that subjunctive clauses are correctly classified as finite it will be necessary to refine the definition of finiteness so that it is not simply a matter of the presence or absence of a tensed verb (see H&P 2005: 36–7).

Verb phrase structure. There are a good number of different analyses found in the literature of expressions like that underlined in She was writing a letter, but for present purposes I will begin with the contrast shown in (2), where the bracketed constituent is a VP and its head is marked by underlining:

1. (2)

In (a) auxiliary was is a dependent of writing, the latter being a main verb functioning as head of the VP. In (b) was is head of the bracketed VP while writing belongs within its complement, writing a letter; within the latter writing is head, so each of the verbs is head of its respective VP. Analysis (b) tends to be favoured in more theoretically-oriented works, and I do not think it would be appropriate to criticise C&M for not adopting it in an ESL-oriented grammar. It is valid to criticise them, however, for further inconsistency and unclarity and for adopting an idiosyncratic analysis that is an untenable compromise between (a) and (b). According to the analysis presented on p. 304, the VP in my example will be just was writing, as in (a), yet the head is not writing but was, as in (b). Note first that this is contradicted by the glossary, where verb phrase is defined as ‘a phrase with a lexical verb as its head’, as in We [should have phoned] you (p. 928): the glossary follows analysis (a). Moreover, when presenting their analysis on p. 304, C&M introduce the concept of ‘main verb’, applying it to the last overtly expressed verb, so that the main verb is writing in She [was writing] a letter, but was in elliptical She [was]. No further explanation is given of what a main verb is; a few pages earlier (p. 299), however, ‘main word’ was used as an alternative to ‘head’ in describing the structure of the noun phrase, whereas we have just noted that they take was, not the main verb writing, as head of was writing. ‘Main verb’ does not appear in the index or glossary, but it is used in three other sections of the book: again with no explanation, but with three different interpretations. On p. 42 it must refer to the first verb in the verb phrase; on p. 583 it contrasts with ‘modal’ and ‘auxiliary’ and is equivalent to ‘lexical verb’; on p. 786 it refers to the verb of the main clause.

The reason C&M take the first verb as head of the VP is that it ‘determines the form of what may follow’. But in fact it only determines the form of the verb that immediately follows. Thus in their example must have been talking the form have is determined by must, but the forms been and talking are not, so there are no grounds for taking these as dependents of must. They assume a flat rather than a layered structure for the verb phrase, giving for this example a head followed by three dependents, but the argument from determining the form of what may follow implies a layering of binary structures: [[must] [[have] [[been] [talking]]]]. The problem now is that it is the last verb, not the first, that determines what may follow the whole, with the lexeme talk permitting, for example, a to phrase: must have been talking to Ed. The same reasoning would therefore require to Ed to be bracketed with talking: [[must] [[have] [[been] [[talking] [to Ed]]]]]. This, however, is what we have in analysis (2b), for a more detailed representation of the VP in (2b) is [[was] [[writing] [a letter]]]. If one wants a structure in which was writing forms a constituent, the only viable analysis is to take writing as head and was as dependent, as in (2a), not the other way round, as C&M propose.

‘NICE’ verbs. I use this acronym as a label for the small class of verbs distinguished from others by their well-known properties relating to negation (He can't swim), inversion with the subject (Can he swim?), code (or ellipsis, Yes, he can), emphatic polarity (He can swim). From the perspective of analysis (2a) the class comprises those verbs generally regarded as filling the dependent positions in VP structure, together with be and, in certain senses, have in head function – ‘lexical’ be and have in C&M's terminology (He isn't ready, Have you any money?). C&M don't discuss the NICE properties all together, and don't have a name for this class of verbs; they don't even have a name for the non-lexical verbs that function in VPs, for they use ‘auxiliary’ only for be, have, and do, calling can, may, will, etc. ‘modal verbs’ (or just ‘modals’), not ‘modal auxiliaries’. It follows that in the numerous places where they are dealing with constructions involving the NICE verbs they have to resort to a disjunctive description. But they don't do this in a consistent way. In the discussion of interrogatives (pp. 534−6) they deal with lexical have, but for the most part they omit it and talk just of auxiliary or modal or lexical verb be – and there are times when they even omit lexical be too (e.g. ‘When there is no auxiliary or modal verb in the main clause, auxiliary do/does/did is used in the tag’, p. 548, a statement which incorrectly excludes not only He hasn't any money, has he? but also He isn't here, is he?).

Modal verbs. These constitute one of ‘three main grammatical classes of verb’ (p. 303), the others being auxiliary and lexical verbs. They are defined at this point as ‘a closed class consisting of core modal verbs (can, could, shall, should, will, would, must, might, may), semi-modal verbs (dare, need, ought to, used to), and modal expressions (be able to, have (got) to)’. In the ‘Types of verb’ chapter, however, there is no mention of ‘modal expressions’ – and on p. 420 the semi-modals are included among the core members of the class of modal verbs. The ‘Modality’ chapter doesn't talk of ‘modal verbs’ and there is no glossary definition for this term. It is clear that modal verbs as defined above do not form a grammatical class at all: it is a semantic grouping of expressions. Thus most of the grammatical statements made in terms of ‘modal verbs’ are patently false if the term is understood as including the ‘modal expressions’. Note, for example, that on the page following the above definition C&M say that a VP can't contain more than one modal verb, whereas examples like We may have to postpone it clearly contain two, a core modal and a modal expression. Indeed the same statement is made on p. 396 and actually illustrated by the contrast between *He will can get a lift and He’ll be able to get a lift (which contrary to what is implied likewise contains two modal verbs). Again, the statement that ‘modal verbs do not show person or number’ (p. 398) doesn't apply to ‘modal expressions’; and so on. The grammatically relevant class is that subset of NICE verbs that have only tensed forms, show no person/number variation and take a following infinitival (in most cases without to): I will henceforth use the term ‘modal auxiliary’ for this class to make clear that it excludes C&M's ‘modal expressions’.

There is also inconsistency and confusion relating to semi-modal verbs. They are said to ‘behave in similar ways to core modal verbs but share some characteristics with lexical verbs’ (p. 638). Note, however, that with dare and need C&M recognise that there are (partially) homonymous lexemes to be distinguished (pp. 657−9): in I daren't ring Linda again and She needn't take the exam if she doesn't want to the underlined items are said to be semi-modal verbs, whereas in Nobody ever dares to say it and He needs to take more care they are lexical verbs. In the case of dare, however, they confuse the two lexemes; most importantly, they treat that in he didn't dare go out as semi-modal, whereas it is clearly lexical. If the distinction between lexical and non-lexical dare and need is drawn correctly, there is no need to call the non-lexical ones ‘semi-modal’: they can simply be classified as modal auxiliaries. So can ought; the glossary definition of semi-modals (p. 922) says that they ‘behave like lexical verbs as they allow an auxiliary to precede them’, but this is obviously not true for ought in Standard English. As for use, this is not syntactically a modal auxiliary – and the meaning it expresses is aspectual, not modal. The fact that there is no person/number variation in She/They used to like it is irrelevant because used is unquestionably a past tense form – one which, unlike that of the modal auxiliaries, invariably has the basic past tense meaning of past time reference. While it is found (though quite infrequently) with the first two of the NICE properties, it does not have the last two; in particular, the emphatic positive requires auxiliary do (as in She did use(d) to like it), so that it also differs from the modal auxiliaries in having a base form.

Multi-word verbs. C&M take over from Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 1150−68; henceforth Q et al.) the concept of multi-word verb, together with the subdivision into phrasal, prepositional and phrasal-prepositional verbs illustrated in sortout, approveof, lookforward to respectively. They claim that ‘[i]n all respects multi-word verbs behave in the same way as single-word verbs’ (p. 303), but that is clearly not true. In the only method [of which he approved], for example, of appears in front position along with the relative pronoun, not after approved: nothing remotely comparable to this fronting of the prepositional component of approveof is found with any single-word verb. The only viable analysis here has of as part of a PP rather than as part of a multi-word verb, and that should be generalised to clauses without fronting: He | approved | of that method. It is perhaps not surprising, therefore, that we find inconsistency in the description of clauses containing so-called prepositional verbs. According to the multi-word verb analysis given in the ‘Types of verb’ chapter (p. 434), for example, that method would be direct object of the verb approved of. But in the ‘Prepositions and prepositional phrases’ chapter (p. 467) they write: ‘Prepositional phrases . . . commonly function as the complements of verbs. Where a special preposition is required to introduce the complement of the verb, such verbs are called prepositional verbs.’ Applying this to our example, we get approved as a single-word verb and of that method as a PP functioning as its complement.

I should add that although they say there are three kinds of multi-word verbs (p. 911), there are numerous multi-word expressions that C&M call verbs at various places in the book which don't belong to any of the above three types. Many, such as be able to and ought to, are assigned to the class of modal verbs discussed above. But it is a mistake to say that such expressions belong to the syntactic class of verbs. It is ought, for example, that is a modal auxiliary, not oughtto, as evident from the fact that ought alone inverts with the subject: Ought we to tell them?, not *Ought to we tell them? It is certainly useful, particularly in an ESL context, to have some convenient way of giving information about the constructions verbs enter into – perhaps by using parentheses, ‘approve (of)’ or ‘ought (to)’ – but saying that approve takes an of PP as complement or that ought takes an infinitive with to is quite different from saying, mistakenly, that of combines with approve or to with ought to form syntactic constituents, multi-word verbs.

Functions in clause structure. The major functions in clause structure that C&M distinguish are S (subject), V (verb), O (object – direct or indirect), C (complement – subject complement or object complement) and A (adjunct). But, again, their treatment of these functions is riddled with inconsistency and error.

In the first place, the term ‘complement’ is used in two distinct senses, without attention being drawn to the dual usage. It is sometimes used in the sense of ‘predicative complement’ (a term which is sporadically used instead), and sometimes in the broader sense of an element involved in the complementation of a head – in this second sense it applies to elements in the structure of phrases as well as clauses. Moreover, neither ‘predicative complement’ nor ‘complement’ in the broad sense is satisfactorily defined. ‘Predicative’ is said in the glossary (p. 916) to refer to ‘a clause element that occurs in the predicate’, but that of course covers objects and adjuncts too, though these are not predicative complements. Another glossary entry says that a predicative complement ‘completes a clause with such verbs as be, become, look’ (p. 896); this applies, however, only to subject complements, whereas ‘predicative complement’ covers object complements too (e.g. p. 491). Complement in the broad sense is defined in the glossary (p. 896) as a ‘word or phrase which is required to complete the meaning of another word or phrase’. The first correction to be made here is to allow for a complement to have the form of a clause (not just a word or phrase), as in numerous examples in the ‘Verb complementation’ chapter. A more general weakness in the definition, however, concerns the claim that complements are ‘required’. This implies that they are obligatory elements – see, for example, p. 502, where adjuncts are contrasted with complements as optional, not required elements.Footnote ⁴ But it is not even approximately true that complements in English are obligatory: most verbs that take an object, for example, can also be used intransitively (as indeed C&M acknowledge on p. 505), and there do not appear to be any convincing examples at all of nouns that take a complement which is obligatory.Footnote ⁵ Furthermore, the idea of completing the meaning of the head word is too vague to provide a workable criterion. Note here that in the discussion of verbs that may be used transitively or intransitively C&M observe that ‘sometimes the meaning remains the same’ in the two constructions (p. 505): it doesn't make sense to say that the complement is required to complete the meaning of the verb if the meaning of the verb is the same whether the complement is present or not. To avoid these contradictions in the explanation of complementation we need to replace the criterion of being required by that of being licensed: verbs, nouns, etc., can be classified according to the dependents that they require or permit, and the dependents that are relevant to this classification are complements.Footnote ⁶

Secondly, C&M say on p. 502 that all elements in basic clauses which are not S, V, O, or C are A, but this is inconsistent with their recognition of other types of complement (in the broad sense) than O and C, namely prepositional complements and locative complements (e.g. pp. 497, 526). Both of these are involved in further inconsistencies. In examples like She gave it to me (p. 784) the PP to me is called a prepositional complement and in the ‘Verb complementation’ chapter such clauses are classified as ditransitive (with the PP now called ‘oblique complement’); in the glossary, however, a ditransitive verb is defined as one with two objects, indirect and direct. As for locative complement, this is illustrated in the ‘Verb complementation’ chapter with Did you put your stuff in our bedroom? (p. 527); on p. 468, however, the PP in Just put it on my desk, please is analysed not as a locative but as a predicative complement. In addition, the ‘Verb complementation’ chapter only allows for locative complements in transitive clauses, yet it is inconsistent to treat to school as a locative complement in I took her to school, but not in She went to school.

Thirdly, one of the things that shocks most in this book is the treatment of the relation between clause functions and semantic roles, such as agent. Thus C&M say that ‘the subject is a noun phrase or its equivalent . . . which indicates the doer or agent of an action, state or event’ (p. 486; the glossary entry for subject uses virtually the same wording, p. 924). This is the sort of gross oversimplification one finds in many school textbooks, but not what one would expect in a serious work in applied linguistics in the twenty-first century. It makes no sense to talk of the agent or doer in clauses like Grass is green, The problem looks serious, and so on – indeed the glossary definition of ‘agent’ omits ‘state or event’ and talks just of the performer or doer of an action. And of course the above description of the subject is inconsistent with their discussion of passives (e.g. pp. 784, 929), and various other constructions (e.g. p. 506), where the subject is explicitly said not to indicate the doer/agent. Surely it is not beyond the grasp of the intended audience to say that the subject is associated with the agent role in an important but limited range of cases (approximately active clauses expressing actions), and that a definition of subject that covers the full range of English clauses must be based on syntactic properties.

Dependent clauses. In the ‘Clause combination’ chapter C&M distinguish two kinds of dependent clause, embedded and subordinate. Embedded clauses are constituents of higher-level phrases or clauses (p. 902) and comprise relative clauses, nominal clauses, and some adverbial clauses, including comparative clauses. Subordinate clauses combine with main clauses to form sentences (p. 553); the examples discussed in this chapter are all adverbial clauses. Again there are inconsistencies in the use of these terms. In the preceding chapter, ‘subordinate clause’ is used in the sense of dependent clause, covering both the above subtypes (p. 545). When subordinate clauses are first introduced in the ‘Clause combination’ chapter itself, on p. 553 (before embedded clauses) they are said to include nominal and relative clauses as well as adverbial clauses, but certainly nominal clauses, by glossary definition (p. 911) and according to the treatment in this chapter, will always be embedded. The definition of embedded clauses given above is based on the glossary and the examples in this chapter (which include, for example, a nominal clause functioning as subject, p. 565), but the definition given in this chapter says they are constituents of phrases (pp. 553, 564), instead of phrases or clauses. Adverbial clauses are said to be ‘elements of sentence structure, rather than as adjuncts . . . in the clause structure’ (p. 579), but the glossary definition of adjunct says this function can be realised by adverbial clauses (p. 890) – and on p. 561 the underlined part of Your evening class tonight has had to be cancelled because the lecturer's ill is analysed as an adverbial clause functioning as adjunct of reason. Comparative clauses are defined as clauses expressing the second part of a comparison of degree, functioning as complements of comparative expressions (p. 773), making them embedded clauses, as above, but two pages later C&M apply the term to the main clause in examples like They are just too upset to speak about it. And while subordinate clauses are first said to combine with main clauses to form sentences, as reported above, it is later recognised that they can also be subordinate to another dependent clause rather than a main clause, as in She worked there for some time, although, as she herself has told you, she was not happy in her job (p. 555: I use double underlining for the lower subordinate clause, and single for the clause on which it is dependent).

This brings us to a major problem with C&M's concept of subordinate clause. When a subordinate clause combines with a main clause the result is a sentence, but C&M have no term for the combination of a subordinate clause with another dependent clause (e.g. for although, as she herself has told you, she was not happy in her job in the above). This means that there is no way of drawing a normal tree diagram – the one C&M provide for this example has a branch going from ‘subordinate clause’ above the lowest clause up to ‘(. . .)’ following although in the upper subordinate clause. And suppose you take a sentence consisting of main clause + subordinate clause, such as Roman soldiers could settle there with their families after they had retired from the army and embed it within an NP, as in Lindum was a town where Roman soldiers could settle with their familiesafter they had retired from the army. The doubly underlined clause is now classified as embedded rather than subordinate, and the tree they provide (p. 565) is uninterpretable. The relevant part is given in (3):

1. (3)

Branches in normal tree diagrams have function or class terms labelling the upper node, but in this tree the branch above ‘adverbial clause of time’ has an actual clause at the upper node and hence cannot be interpreted in the standard way. C&M give no reason for their decision not to treat adverbial clauses as adjuncts, but they pay a huge price for it in terms of unnecessary complexity and loss of consistency with the implicit principles that elsewhere govern permitted tree structures. It would be much better to abandon the distinction between subordinate and embedded clauses, treating all dependent clauses as embedded.

I should add that C&M's treatment of non-finite dependent clauses is seriously limited and unsystematic. There is one section labelled ‘Non-finite clauses’, located in the ‘Clause types chapter’, but it is little over half a page in length (p. 546), and deals only with the issue of their temporal interpretation.Footnote ⁷ Nowhere do they deal with the infinitival construction introduced by the subordinator for followed by the subject, as in It's important for the job to be finished this week. In the section ‘Central clause functions’ they say that the subject is ‘most typically a noun phrase, though it may sometimes be a wh-nominal clause, or, very occasionally, a prepositional phrase’ (p. 495): they fail to mention other possibilities more important than PPs, in particular non-finite clauses, as in To turn back now would be a mistake. In the ‘Verb complementation’ chapter they mainly deal in turn with monotransitive, ditransitive, complex-transitive, and copular constructions, and include lists of verbs which can take non-finite clauses as one of the complements in the first three of these, omitting mention of non-finite complements in copular clauses like His favourite pastime was reading novels or These changes are to increase security. For the three constructions where non-finite complements are said to occur, not one word of commentary is given concerning syntactic or semantic differences between them, and there are numerous aspects of the verb classification that are quite implausible. Consider the following examples containing verbs from the three lists given on pp. 516−17, 523, 524:

1. (4)

2. (5)

Why should choose differ from order and allow in having Kim as subject in the dependent clause rather than object in the main clause, and why should the function of to chair the meeting be different in (4b) and (c)? (In (c) it is object complement, whereas in (b) no functional label is assigned, but as the clause is ditransitive it could not be an object complement.) And why with intend in (5) is to win the match a direct object in (a) but an object complement in (b)? There's no point in classifying verbs with non-finite complements according to which of the three constructions they enter into unless some explanation is given of the differences between them: readers are in effect asked to accept the verb classification as an ex cathedra statement with no evident relevance to the way they construct or interpret sentences. It would be far more useful to deal with non-finite complementation of verbs separately rather than treating the complements as functionally equivalent to NPs. And again we find inconsistency between the analysis presented in this chapter and that found elsewhere in the book. Thus fail and manage are said to take infinitival direct objects (pp. 516−17), as in I failed/managed to solve the problem, but on p. 401 such examples are analysed as simple sentences with failed/managed to solve classified as ‘catenative verb phrases’.Footnote ⁸

Declaratives vs interrogatives. The major source of inconsistency in this area stems from a failure to distinguish clearly between the clause types on the one hand and their characteristic constituent order on the other. Thus C&M talk not only of declarative and interrogative clauses, but also of declarative and interrogative word order (e.g. p. 547). They say several times, without qualification, that declarative clauses have the order S – V – X (e.g. pp. 532, 899) and call this declarative word order. Interrogative word order is Modal/Aux – S – X (or variants allowing for lexical be or have to precede S: see the discussion of NICE verbs above). But of course the latter order can occur in declaratives that have fronting of various kinds, as in Not once did she thank me (p. 781), and wh-interrogatives like Who wants more coffee? (p. 538) have the former order. The inconsistency and confusion is particularly striking in the case of dependent interrogatives. In the ‘Clause types’ chapter they are called ‘indirect interrogatives’ (pp. 539−41), but elsewhere the term generally used is ‘wh-clause’, defined as ‘a declarative clause that is introduced by question words who(m) [etc.]’ (p. 929, emphasis added). This inconsistency should surely be resolved in favour of the former term, so that the contrast between declaratives and interrogatives in main clauses will apply also to their dependent clause counterparts – and so that we can avoid such anomalies as saying that who told him is interrogative when it stands alone but declarative when it occurs in contexts like I wonder who told him. Note also that in the rules concerning the distribution of polarity-sensitive items like yet (p. 159) ‘interrogative’ must cover dependent as well as main clauses: Has he finished yet?, I wonder if he has finished yet, *He has finished yet, *I know that he has finished yet. It is important to see that the major distinctive property of wh-interrogatives is that they contain an interrogative wh-word, not that they have a particular constituent order, for the order is variable, depending on whether the wh-constituent precedes the subject and whether the interrogative is a main clause. Confusion between clause categories and word order could be avoided by talking not of declarative word order but of basic word order, and not of interrogative word order but of inversion of subject and whatever term is used for NICE verbs.Footnote ⁹

Other inconsistencies in this area concern the term ‘polar’ and the classification of whether clauses. ‘Polar’ is sometimes used in contrast with ‘alternative’ (e.g. p. 805), sometimes (and inappropriately, given its relation to ‘polarity’) to cover both yes/no and alternative interrogatives (e.g. p. 534). A dependent interrogative with whether is classified as a wh-clause on pp. 329, 513, but excluded from this category by glossary definitions (pp. 907, 929) – and also excluded from the category of wh-interrogatives (pp. 539–40).

Wh-constructions. There is an important set of constructions in English marked by the presence of one (or in some cases more) of the wh-words who, whoever, whom, whomever, etc. For present purposes I will refer to them informally as wh-constructions. The set consists of wh-interrogatives, main and dependent; exclamatives, main and dependent; exhaustive conditionals (aka concessives or conditional-concessives), as in They'll go ahead, whatever you say; wh-relatives, as in the book which she was reading, and fused relatives (aka headless, free or nominal relatives), as in What you gave me was not enough. This set is of course much larger than C&M's wh-clause. The wh-word, or rather the phrase containing it, may have a variety of functions, including a function within a clause embedded in the wh-construction (compare Who said that?, subject; What do you want?, object; Who do you think will win?, subject of embedded clause), but C&M don't bring this out. And they devote little attention to comparing the structures of the different wh-constructions.

Nothing further need be said about wh-interrogatives. As for exclamatives, C&M distinguish this syntactic category from the speech act category of exclamations (pp. 531, 682), but the distinction is not applied consistently. Thus Oh God, was I exhausted by the time I got home! is correctly described as an interrogative clause used as an exclamation (p. 683), but Oh isn't that nice! is classified as exclamative rather than interrogative on p. 717. Dependent exclamatives are not mentioned as such – the ‘Verb complementation’ chapter is unsystematic in that it lists various verb classes that take interrogative complements but ignores exclamative complements, as in She remarked what a beautiful house it was. That example, however, is cited on p. 804 as a wh-clause reporting an exclamation, which is of course inconsistent with the definition of wh-clause cited above, as this what is not a question word.

Exhaustive conditionals are dealt with twice, as though there were two quite separate constructions here – there isn't even a cross-reference between the two sections concerned. On p. 562 they are included as examples of a larger class of adverbial clauses of concession, while on p. 752 they are analysed as wh-conditional clauses. There is no basis for this dual analysis: Whatever I do, I'm compromised (illustrating the former) and Whatever she says, don't believe it (illustrating the latter) belong to a single syntactic construction. Notice, moreover, that C&M classify the wh-words in the concessive construction as subordinators, like although (pp. 558, 562), but whatever in the above examples is clearly a pronoun, functioning as head of the object NP, just like what in What did she say?

C&M's treatment of relatives is one of the worst features of the book, involving as it does two inexcusable omissions. In the first place, for non-fused relatives (pp. 566−76) they allow only for relative pronouns (including that and zero). There is no mention of the relative determiner which, though it occurs in an example they cite on p. 40 (By which time, it will be too late . . .). Nor do they recognise traditional grammar's relative adverbs; two examples are in fact cited with an incorrect analysis. Thus the relative clause in That's part of the reason why we bought it (p. 329) is wrongly classified as a ‘defining wh-interrogative clause’ functioning as complement of a noun: its status as relative (modifier) rather than interrogative (complement) is evident from the fact that why can be replaced by that or omitted, and that reason can't take obvious interrogative complements such as what we bought. The other example I have already quoted in another connection: Lindum was a town where Roman soldiers could settle with their families . . . (p. 565), where the underlined clause is said to be an adverbial clause of place, as shown in (3). There's nothing ‘adverbial’ about this clause: traditional grammar would call it adjectival since its function is to modify the noun town (and, as we noted, C&M do in fact label it postmodifier). Its status as a relative can easily be seen from the fact that we can replace where by in which, where which couldn't be anything other than a relative pronoun. The second omission concerns fused relatives: this is not recognised as a distinct construction at all. A good number occur in the examples cited at various places in the book, but in the few cases where they are classified they are called ‘wh-clauses’ (e.g. It's not quite what I'm looking for, p. 128, She went back to where she used to work, p. 462 – and in ‘wh-clefts’, as in What you need is a telephone bank account, p. 786). This classification is inconsistent with the definition of ‘wh-clause’ quoted above, since the wh-words here are not question words. But it would not of course do to simply modify the definition (e.g. by replacing ‘question words’ by ‘wh-words’), for fused relatives must be distinguished from dependent interrogatives on well-known syntactic and semantic grounds.Footnote ¹⁰

Determiners, pronouns, and possessives pronouns. C&M's section on possessive pronouns gets off to a terrible start: ‘There are two classes of possessive pronoun: possessive determiners and possessive pronouns’ (p. 382). This is to treat ‘possessive pronoun’ as ambiguous, having one sense in which possessive pronouns contrast with possessive determiners and another, more general, sense in which they include possessive determiners as a subclass. This is inconsistent with the principle implicit elsewhere in the book that the parts of speech are mutually exclusive, i.e. that no word in an unambiguous sentence can belong simultaneously to more than one part of speech. What they presumably meant was something like this: ‘Possessive counterparts to the personal pronouns fall into two classes: possessive determiners and possessive pronouns.’

Inconsistencies abound in the discussion of these forms, as illustrated in C&M's analysis of the following examples, where the annotation on the right gives their classification of the underlined form:

1. (6)

Examples (a–b) illustrate the basic uses of possessive determiners and possessive pronouns respectively. C&M treat own as a dependent of a preceding possessive (though they are inconsistent as to what kind of dependent it is, treating it now as a complement (e.g. p. 383), now as a postmodifier (same page!)). This makes my in (c) head of the NP, and it is therefore treated as a pronoun (p. 359), but this is inconsistent with the table of forms given, where my appears only in the determiner column. I should add that the table of possessive determiners and pronouns is given twice, but in two mutually inconsistent versions. On p. 358 the pronoun entries corresponding to the determiners its and one's are ‘(no form)’, whereas on p. 382 they are given as its and one's, with a note saying that these pronouns occur ‘only when complemented by (very) own’.Footnote ¹¹ But if its and one's qualify as pronouns on the strength of their occurrence in the NPs its own and one's own, then my, your, etc. should likewise figure in the pronoun column. C&M's analysis of examples like (c) requires two sets of possessive pronouns, one of them homonymous with the determiners.

The analysis of expressions like the school's in (6d) as determiners is inconsistent with: (a) the account of determiners as a closed class of function words (p. 894); (b) the definition of genitive as a ‘case that denotes a possessive relationship with another noun phrase in a sentence’, as in the woman's husband (p. 905, emphasis added); (c) the statement that when ‘a possessive determiner and a possessive ’s form modify the same head noun, the possessive determiner normally comes second’, as in Jane's and my wedding (p. 383). It is of course a highly unorthodox analysis and can't have been thought through carefully. C&M take those as a determiner in those videos but as a pronoun in I'll need those (p. 357), which matches the classification of my as a determiner and mine as a pronoun in (6a–b), but they fail to mention the use of expressions like the school's without a following head, as in The college's reputation isn't as good as the school's. They couldn't say that the school's is a pronoun here but a determiner in (6d): the difference must be accounted for in terms of the structure of the topmost NP, not in terms of the class of the possessive expressions, for they are systematically identical in form. And of course they couldn't say that the school's was a determiner in both constructions.

The classification of relative whose in (6e) as a pronoun is inconsistent with the inclusion of whose in the list of determiners on p. 353, and with the analysis of all other possessive expressions as determiners when they occur as dependent to a following head noun.

The construction illustrated in (6f) is not mentioned in the ‘Pronouns’ chapter, and the discussion in the ‘Verb complementation’ chapter (p. 515) implies a different treatment of possessive forms from that in the earlier chapter: C&M say that if the subject of a verb in the -ing form is a pronoun it occurs in the object form, except that in formal styles it may occur ‘as a possessive form’. This suggests that his is considered as an inflectional form of the personal pronoun he, which matches the glossary entry for case, but not the account of personal pronouns in the text (pp. 376–82). The only other example of this construction that is discussed (pp. 466−7) likewise involves his: C&M don't point out that where we have an overt difference between my and mine, your and yours, etc., it is the former member of the pair that is required here, and the pronoun analysis is again inconsistent with the tables of possessive determiners and pronouns on pp. 358 and 382, mentioned above.

There are two things we need to do to get rid of these inconsistencies. The first is to apply the distinction between classes (or categories) and functions to the description of determiners. C&M are inconsistent as to whether ‘determiner’ is a class or a function term. In the glossary definition of class (p. 894), determiner is included among the closed word classes, as noted above, and in the Introduction to ‘Word classes and phrase classes’ they write: ‘A noun phrase usually consists of a head along with modifiers or a complement. . . . Premodifiers include determiners . . ., adjectives . . ., and other nouns’ (p. 299), implying that determiner is a class with the function premodifier. But in the next chapter they say that noun heads ‘may have three types of dependent element: determiners, modifiers and complements’ (p. 322), with ‘determiner’ now a function term. Importantly, however, to the extent that they are conscious of the distinction, they tacitly assume that there is a one-to-one relation between the determiner class and the determiner function. This is a mistake: we need to recognise that the determiner function can be filled by other kinds of expression than members of the determiner class.Footnote ¹² The most obvious are genitive NPs, like the school's in (6d):Footnote ¹³ the inconsistencies described above are avoided if we say this has determiner function, but belongs to the class NP. This solution extends naturally to NPs with pronouns as head. Thus in (6e) whose likewise belongs to the class of genitive NPs and has determiner function. With the personal pronouns there are two sets of genitive forms, dependent (my, your, etc.) and independent (mine, yours, etc.): the dependent genitives function as dependent (determiner), not head, in NP structure and as subject in clauses like his coming here in (6f) (subject being arguably also a dependent function, relative to the VP, which can be considered as head of the clause).Footnote ¹⁴

The second change we need to make concerns the analysis of the adjective own. In my own birthday it is a premodifier of birthday, not a postmodifier (or complement) of my. C&M claim (p. 125) that own can't occur alone (i.e. with no determiner) or after a or the, but although it usually follows a genitive it is not required to do so, as shown in these examples from Richard Dawkins’ The Selfish Gene, which provide evidence against taking own as dependent of a preceding genitive:

1. (7) When we talked about flukes and snails we accustomed ourselves to the idea that a parasite's genes could have phenotypic effects on the host's body, in the same way as any animal's genes have phenotypic effects on its ‘own’ body. We showed that the very idea of an ‘own’ body was a loaded assumption. (2nd edition, p. 250; see also p. 247)

Further evidence for the treatment of own as modifier of the following noun in NPs like my own birthday (which again is the traditional analysis) is that it explains why own cannot occur in (6f): coming here is a VP, not a noun, and hence cannot take an adjectival premodifier. C&M's my own in (6c) is thus like my youngest in My eldest son is twelve years older than my youngest, where there is no overt noun following, and my is a dependent genitive pronoun functioning as determiner in accordance with the rule given above. Their analysis of my own here may be due to their belief that noun phrases ‘consist minimally of a noun or pronoun, which acts as the head of the noun phrase’ (p. 318). This, however, is clearly not true, and it is another serious gap in the coverage of the book that it fails to deal with the structure of NPs not headed by a noun or pronoun, like my youngest, the latter, the very rich, etc.

I confess it is a mystery to me how so much inconsistency and confusion could remain in a book seven years in the making – especially as the research project involved a twelve-member ‘Reference Panel’ credited with having contributed to workshops and commented on draft chapters (six of them are singled out for their ‘most thorough and meticulously detailed reports on the various versions of the manuscript’, p. v). In his (very magnanimous) review of H&P, Leech (Reference Leech2004: 126) suggests a scale of abstraction for grammars ranging from the most data-oriented, through descriptive, to the most theoretical works, placing H&P towards the theory-oriented section of the descriptive area on the scale and Q et al. towards the data-oriented section of that area. C&M would then occupy a position somewhat further removed from the theory-oriented end of the scale than Q et al. I would emphasise, however, that the criticisms I have made here have not been made from a theoretically oriented position. This would not have been the place, for example, to argue for some of the more radical departures from traditional grammar that H&P adopt, and in fact some of the main criticisms I have made of C&M's grammatical analysis concern issues where traditional grammar's account is superior to theirs: modal auxiliaries as tensed verbs, dependent interrogatives, the part of speech classification of possessives like my, your, etc. Moreover, the serious gaps in coverage I have drawn attention to involve common constructions dealt with in Q et al.: infinitivals introduced by for, NPs without a noun or pronoun as head, and various relative constructions. And of course the bulk of my criticisms have been concerned with the extraordinary amount of inconsistency in C&M's account of the grammar, a failure to meet elementary standards that apply to any grammar, whatever its position on Leech's scale. Readers will undoubtedly be able to derive much of benefit from the book, but to a significant extent this will result from their ability to generalise for themselves from the great number of examples provided rather than from struggling to work out and assimilate what C&M say about the grammatical analysis. It is very much to be regretted that what is likely to be regarded as an important work in applied linguistics should have such a weak foundation in descriptive linguistics.

One final complaint concerns the index, which is nothing short of a disaster. Firstly, there are over sixty entries where the sole reference is to the glossary – case, gender, person, polarity, stem, tense, topic, etc. This is useless: what we need is information about where the concept is dealt with in the main text. Secondly, there are a good few terms that don't appear in the index at all: allomorph, commissive, conjunct, -ed form, -ed participle, hypothetical conditional, oblique complement, syncretism, tensed verb (phrase), and so on. Thirdly, where an item is listed and given references other than to the glossary, they are often very incomplete. For example, just two references are given for non-finite clauses but there are another sixteen sections where we find relevant mention of such clauses. Or take the form none: the index gives four references, but there are a further eight instances that should have been included. If the plans for a second edition materialise, there will need to be not only a large-scale overhaul of the grammatical description, but also a complete rethinking of how to compile the index for a book of this kind.