The Blackwell companion to syntax

The Blackwell companion to syntax comprises around 4,100 pages, in five volumes, with a total of seventy-seven chapters (3,285 pages), cowritten by seventy-nine authors, many of them leading figures in generative grammar. Each chapter offers a case study in generative syntax. The chapters vary in length from eleven pages for the shortest (Ch. 10, 'Bridge phenomena', by NOMI ERTESCHIK-SHIR) to over 100 pages for the longest (Ch. 61, 'Specificational copular sentences and pseudoclefts', by MARCEL DEN DIKKEN, is 118 pages; Ch. 75, 'Verb clusters, verb raising, and restructuring', by SUSI WURMBRAND, runs to 115 pages). There is a five-page preface by the editors and a 113-page index to all volumes, both reproduced in each volume, and a consolidated reference section of 160 pages at the end of Vol. 5. The accompanying CD is compatible with PC and Macintosh computers.

Though the term 'generative' does not figure in the title of this collection, the very first line of the preface, 'Generative grammar has always set its aims high' (xix), makes it clear that the Companion focuses on generative syntax. The Companion brings together a vast store of empirical knowledge accumulated in the generative tradition during the last half-century. As the editors point out, in generative syntax empirical data and their analyses remain subordinate to the theory, or putting it differently, 'data and analyses are seen as what they are: tools to help us understand the structure and properties of the human language faculty' (xix). A potential drawback of this theoretical orientation is that with a change of the theoretical perspective, or of the theory as such, empirical generalizations that at some point had been the focal point of the theory may fade into the background. Even if the original papers dealing with the relevant phenomena remain available, changes in the theory mean that earlier papers become less easily accessible to a newer generation of researchers. Also, the knowledge accumulated on a particular empirical domain is not often recorded in one paper; rather it will be scattered over a number of papers, each with its own theoretical perspective. Thus, it may often be difficult to trace all the relevant empirical material. To quote from the introduction: 'So searching for the relevant materials . . . is often hard. For young researchers, such a search task is doubly difficult because very often the descriptive generalisations are presented in a jargon and formalisms that were characteristic of the theoretical framework at the time, sometimes decades ago' (xx). Moreover, 'the presentation of the data in these articles is subservient to the theoretical claims that are made and hence not presented in succinct, didactically useful ways' (xx). To make up for the observed deficit in accessible source material, the case studies in the Companion present empirical phenomena that have furnished the basis of much research. The aim is also to make this presentation didactically useful.

Each self-contained chapter deals with a particular phenomenon. For reasons of space I do not provide an exhaustive list of topics; suffice it to say that there are chapters on a widely diversified range of areas such as adjectives, adverbials, clitic climbing, free relatives, pied-piping, preposition stranding, verb-particle combinations, verb clusters, and so on.¹ Each chapter aims to provide an overview of the empirical generalizations relevant to the particular domain of analysis, an evaluation of the analyses proposed in the generative literature, and a discussion of the theoretical relevance of the data. As seventy-nine authors have contributed to these volumes, the chapters inevitably end up looking quite different in terms of length, scope, and organization.

The chapters are all written within the mainstream generative approach; they contain references to the relevant generative literature, but many chapters also contain references to lexical-functional grammar, relational grammar, head phrase structure grammar, and so on. In addition, there are references to philosophical writings and to nonformal or less formal approaches to linguistics.

The variation among the chapters is, among other things, determined by the way each author articulates the relation between data and theory. While adopting a generative approach, some authors give the data a central place: theoretical tools and formalisms are presented in a way compatible with different implementations of generative theory. In addition, authors also address theoretical issues that go beyond the empirical phenomenon investigated, and these discussions are an important reminder that the phenomena studied should not be seen in isolation but that they remain part of a larger research enterprise.

In this respect, Wurmbrand's Ch. 75 on 'Verb clusters' is a real tour de force. Verb clusters can be illustrated with data from West Flemish: the three verbs in 1-the finite modal zou 'should', the infinitival modal moeten 'must', and the infinitival lexical verb lezen 'read'-are lined up at the end of the clause as one cluster zou moeten lezen. The DP dienen boek 'that book', the object of the most deeply embedded verb lezen, precedes all verbs. Similarly, the PP van myn 'of me', which gives the source of the modality in moeten, precedes the verb cluster zou moeten lezen.

(1) da  Marie dienen boek van myn nie zou    moeten lezen

that Marie that    book of  me  not should must    read

'that, as far as I am concerned, Mary needn't read that book'

Wurmbrand's chapter offers a systematic survey of verb clusters in a range of Germanic languages and their dialects (German, a number of German dialects including those spoken in Austria and Switzerland, Dutch, West Flemish, Frisian, Afrikaans) focusing mainly on the sequencing of the verbs. She examines how the orders can or cannot be derived, testing a number of hypotheses that differ in terms of two general theoretical questions: (i) whether verb clusters are derived by head movement or by phrasal movement, and (ii) whether the Germanic head-final OV order is base-generated or derived. With respect to the latter question, Wurmbrand points out that 'the question of whether one of the two approaches . . . is superior cannot be determined on purely empirical grounds. . . . Comparing representative examples of both approaches, the conclusion this chapter will reach . . . is that the choice between the two approaches can only be made in conjunction with the choice of a particular syntactic framework or aspect of a syntactic theory' (Vol. 5:233). In other words, the data as such do not allow us to choose between an approach in which OV orders are base-generated or one in which they are derived; see Haegeman 2001 for a similar conclusion on the basis of a discussion of sentential negation.

In other chapters the discussion is tied to chronological stages in the development of generative grammar, illustrating the interplay between the development of the theory and the relevance of particular data. An excellent illustration of this approach is ANNA SZABOLCSI's chapter on 'Strong vs. weak islands' (Ch. 64). The author shows that the development of the theoretical treatment of this issue is paralleled by a gradual widening of the database: 'practically each theory comes with a significant new set of data' (Vol. 4:492). For instance, while the initial discussion of weak islands focused on the argument-adjunct asymmetry illustrated in 2, soon it became clear that, in order to be able to be extracted, the argument must be referential in some sense (3).

(2)

1. a. Which problem did John ask [how to phrase _____]?          (Vol. 4:494, ex. 82)

2. b. *How did John ask [which problem to phrase _____]?

(3)

1. a. *What did John ask [whether these pearls cost _____]?          (Vol. 4:495, ex. 87)

2. b. *How did John ask [whether to behave _____]?               (Vol. 4:495, ex. 88)

Szabolcsi discusses a number of additional factors that play a role in determining extractability, concluding that '[a]lthough we still do not have a fully unified theory at our disposal, it seems true beyond reasonable doubt that a substantial portion of this large phenomenon is genuinely semantic in nature' (Vol. 4:527).

Kriszta Szendrői's Ch. 26, 'Focus movement (with special reference to Hungarian)', surveys three types of analyses of focus movement: a first approach uses the adjacency requirement between the verb and a focused constituent to align focus with case. A second approach exploits the similarities between WH-movement and focus movement. Finally, a more recent line of research starts from the observation that what unites focal constituents is that they bear main stress. For each type of analysis, Szendrői provides the most important empirical generalizations and then offers a critical evaluation of the advantages and disadvantages of the analysis. Given that focus is related both to prosodic properties ('main accent') and to information packaging and discourse relations ('new information; theme vs. rheme', etc.), her discussion also raises the more general question as to what, if any, should be the place of focus in the narrow syntax.

A few authors have opted for a more personal approach, basing their contribution mainly on their own analyses of the data. This is the case, for instance, in JEAN-YVES POLLOCK's Ch. 67 on 'Subject clitics and complex inversion', which is a short version of the second chapter of Poletto & Pollock 2004. Though to a lesser extent, ANDREA MORO's contributions, 'Copular sentences' (Ch. 18) and 'Existential sentences and expletive there' (Ch. 24), also highlight his own analyses. Such chapters have the advantage of offering a coherent and unified discussion of the data with a fully worked-out analysis. What is lost, however, is the systematic inventory of data and references and the critical survey of theoretical approaches that we find with authors opting for a more neutral approach.

For syntacticians working in the generative framework this collection is a goldmine; it offers survey chapters for a host of phenomena and each of them will be a solid basis for research. The Companion, however, is not only a reference work: the very juxtaposition of these wide-ranging materials will encourage the inquisitive reader to look beyond particular chapters and discover relevant links among the chapters. Let me provide just one example for this, related to my own research interests. In Ch. 70, 'The syntax of modal auxiliaries', SJEF BARBIERS discusses the observation that expressions of epistemic necessity are incompatible with yes-no questions, as shown in 4.

(4) *Must John be a speaker of Finnish?

Intended meaning: 'Is it necessarily true that John is a native speaker of Finnish?'

He points out that, along the lines of McDowell 1987, one might propose that epistemic modals have to raise to a clause-initial position at LF to scope over the proposition. The incompatibility of epistemic modals with yes-no questions could be due to the epistemic operator and the left-peripheral question operator competing for the same position (Vol. 5:15-16; see, however, Barbiers for important qualifications (Vol. 5:16) and Nilsen 2004 for an alternative account).

Rajesh Bhatt and Roumyana Pancheva's Ch. 16, 'Conditionals', proposes that, like yes-no questions, antecedents of conditionals be derived by operator movement (Larson 1985). Conditional clauses constitute another environment in which expressions of epistemic modality are restricted (Haegeman 2003, Nilsen 2004), as given in 5a.

(5)

1. a. *If it probably rains, we won't go out.

2. b. If Le Pen will probably win, Jospin must be disappointed.

(Nilsen 2004:811, n. 5, ex. (ii))

The ungrammaticality of 5a could be accounted for if we assume that here too the epistemic modality and the conditional operator compete for the same position (pace Nilsen 2004). By contrast, 'factual' conditionals (Vol. 1:639-40) do allow epistemic modals (Haegeman 2003), as shown in 5b above. To differentiate 5a and 5b, one might propose either that the conditional in 5b is not derived by operator movement, or, alternatively, that the factual conditional operator and the epistemic operator have different landing sites. Obviously, this is just one example of how the juxtaposition of empirical domains may spark new ideas, to be confirmed or disconfirmed in future research.

In most cases, generative research will have explored a particular phenomenon especially well in terms of one language or one language group. Hence the range of languages discussed in the case studies inevitably varies. For instance, the phenomenon of 'Split topicalization' (Ch. 62) has received most attention with respect to German and Dutch, and it is therefore only natural that these languages figure prominently in the discussion. Toward the end of the chapter, however, HANNEKE VAN HOOF elegantly widens the scope of the discussion by including a range of languages for which split topicalization has also been studied, such as Modern Greek (Vol. 4:445-46), Slavic languages (Vol. 4:446-49), and Japanese (Vol. 4:449). This chapter, which I personally found impressive, ends with a brief discussion of Malagasy (Vol. 4:449-50) and Vata (Vol. 4:450), and a further discussion of 'nonconfigurational' languages in the conclusion section, thus once again moving from a narrow empirical focus on one case study into wider theoretical issues.

Occasionally, however, the discussion remains narrowly limited to one language or one language group. For instance, Ch. 1, 'The accusative plus infinitive construction in English' by JEFFREY T. RUNNER, is restricted to English (which the title makes clear), and Ch. 6, 'Analytic causatives' by MARIA TERESA GUASTI, deals exclusively with the Romance languages. Such limitations are unexpected-and to my mind regrettable-especially against the background of the other chapters that attempt to broaden their outlook.

The specialized researcher will certainly find these volumes useful in a number of different ways. The Companion also targets nonspecialist readers such as the (graduate) student who has some theoretical basis but who needs empirical underpinning for his or her research or/and who needs an overview of various theoretical approaches to a phenomenon, and the nonspecialist researcher (e.g. a linguist or a psycholinguist) who is only broadly familiar with the generative framework. For these readers, the Companion will constitute a basis for reference, and the didactic presentation should obviously be a major asset.

As mentioned, one of the important aspects of the collection is that by having these seventy-seven chapters presented together the reader is encouraged to explore relations between different kinds of phenomena. For the specialist it will be fairly easy to (re)discover relations across the chapters. For the nonspecialist, discovering links between the chapters may sometimes be harder. Though many chapters do provide cross-references to other relevant chapters in the volume (for instance, ADRIANA BELLETTI's Ch. 48, '(Past) participle agreement', offers systematic cross-references), other chapters do this to a lesser degree. To give just one example of many, DOMINIQUE SPORTICHE, in Ch. 54, 'Reconstruction, binding, and scope', points out that with respect to connectivity effects, 'one type of construction in particular stands out as significant: (specificational) cleft constructions' (Vol. 4:38). The construction is not illustrated at that point, nor is there a reference to Ch. 61, 'Specificational copular sentences and pseudoclefts' by Marcel den Dikken, in which connectivity effects in specificational pseudoclefts are discussed in great detail (Vol. 4:313-21). Obviously, for cross-references, the reader can turn to the reference section, but a reference in the chapter would have been useful. Interchapter cross-references are particularly handy in the CD version (which I discuss more later), in which a click on a cross-reference opens the relevant chapter, allowing the reader to quickly browse through the related material. In addition, systematic cross-referencing would emphasize more clearly that even if we may study individual linguistic phenomena, in the generative approach such phenomena are never viewed as isolated 'constructions' but remain part of a broader picture.

Overall this is a first-rate publication, but inevitably, with such a large number of contributors, there remain a few flaws. Though not substantive, they mean that, in spite of the overall effort to present the data in a didactically useful (xx, Vol. 2:231) or pedagogical (Vol. 4:627, 640) way, this praiseworthy goal is not always fully achieved.

A first observation is that there are quite a few typing errors. When these concern languages other than English, they are particularly unfortunate in that the reader may get the facts wrong. For instance, the Dutch auxiliary of the passive is not werden (Vol. 5:242), but worden. Ch. 24 has the ungrammatical French 6a instead of 6b.

(6)

1. a. [*]Il-y-a      beaucoup des    filles.                    (Vol. 2:211)

   there-are many      of.the girls

2. b. Il y a beaucoup de filles.

In the otherwise excellent Ch. 43 on 'Mittelfeld phenomena (scrambling in Germanic)' by HUBERT HAIDER, we find the following (my italics): 'This would incur a crossing violation in that the trace is not within the non-breaking hyphen domain of its antecedent' (Vol. 3:211). I suspect that 'non-breaking hyphen domain' is a residue of a copyediting instruction, but a student might be puzzled. In the extract below from Ch. 31, 'Grammatical verbs (with special reference to light verbs)' by TARA MOHANAN, the commentary does not seem to quite fit the examples.

[7b] is a causative (see chapter 4 [sic]) version of [7a]:

[7]

1. a. Pat combed Ellen's hair.

2. b. Pat made Jan comb Ellen's hair.

If the object Ellen in [7b] is replaced with the pronoun him, the pronoun can take Pat as its antecedent. On the other hand, if we replace it with the reflexive himself, the reflexive cannot take Pat as its antecedent.

(Vol. 2:472, ex. 24)

Following the instructions would lead to gibberish ('him hair', 'himself hair'). Presumably, what was intended was that the object Ellen's hair rather than just Ellen should be replaced.

There are other problems of presentation. In Ch. 3 on 'Adjectives: Order within DP and attributive APs', by DANIEL VALOIS, we find example 7 (Vol. 1:64), repeated here in 8. No glosses (or explanations) are given for oo in the third and fourth lines of 8a. For Kurdish (8b), neither the ending -a on the noun and the adjectives nor the final morpheme ka are glossed or commented on. The reader thus has no way of knowing if these particular morphemes play any role in allowing the apparent free ordering of the adjectives.

(8)

1. a. Somali gabadh fiican dheer madow

              girl        good tall    dark-skinned

              gabadh fiican oo     madow          oo dheer

              gabadh dheer oo    madow          oo fiican

              etc.

2. b. Kurdish top-a jwan-a   gawr-a sur-a-ka

                ball    beautiful big        red

                top-a-sur-a gawr-a jwan-a-ka

                etc.

When authors temporarily forget their stated didactic goals (Vol. 2:231) the discussion may become rather dense. This is, for instance, the case in Ch. 24 on 'Existential sentences and expletive there'. Based on, among others, the data in 9, the defining properties of existential sentences (Vol. 2:215) are listed in 10.

(9)

1. a. *Which girls do you think that there are [t in the garden]?

2. b. What do you think that there is [t in the garden]?

3. c. Which girls do you think that there are [pictures of t] in the garden?

4. d. [How many men] do you think that there were t in the room?

(Vol. 2:212, ex. 8a,b,c)

(10)

1. a. Movement of the subject is impossible.

2. b. Movement from within the subject is possible.

3. c. The determiner of the subject has an adjectival character.

But the subject what is extracted in the grammatical 9b and also in 9d (Vol. 2:221, ex. 32a), which seems to contradict 10b.

Concerning the extraction in 9d, footnote 23 says: 'As for the case involving how many illustrated in 22, we are led to conclude that how many is extracted from the subject and that men is pied-piped for locality reasons. For a detailed discussion see Moro (1997)'. For the nonspecialist this will be less than illuminating. First, the relevant text example with how many is not 22 (which is there is a girl in the room/there are girls in the room, Vol. 2:219), but rather 32a (i.e. my 9d). Second, without a more explicit account the nonspecialist will wonder why moving which and pied-piping girls is not equally possible in 9a. The global reference to Moro 1997 is less than charitable. Finally, the reader will remain puzzled by 9b, which is not discussed in detail. In n. 1 (Vol. 2:231) and on page 222 (Vol. 2), the reader is referred to Ch. 18 on 'Copular sentences', and this chapter contains highly relevant discussion of the difference between which and what in relation to extraction (see Vol. 2:15-16), but Ch. 24 does not give a precise reference to that discussion.

Omissions of (precise) cross-references and incomplete or dense accounts can be overcome by the specialists, who will supply missing information from their background knowledge, but they do make these chapters less user-friendly for the nonspecialist readers.

Another worry I have is that in certain cases the data themselves are not presented adequately. This has already been illustrated, but I add two more examples here. In Ch. 51, 'Preposition stranding', PAUL LAW proposes that 'P-stranding is related to the independent grammatical property of D incorporating into P . . . [P-stranding] is contingent on D not incorporating into P' (Vol. 3:633). He illustrates this for French: prepositions incorporate the determiner associated with their complement, which often leads to suppletive forms; in 11 the preposition de 'of' and the definite article le 'the' form du.

(11)

1. a. *J'ai parlé de le garçon.

   'I have talked of the boy.'

2. b. J'ai parlé du garçon.

Law's account of the ungrammaticality of P-stranding in French (12a) is as follows: 'the head D quel incorporates into the P de' (Vol. 3:649); as a result of the incorporation, DP cannot be fronted without pied-piping the preposition. D-to-P incorporation is said to be 'most evident in [12b] and [12c] with the suppletive form duquel' (Vol. 3:649).

(12)

1. a. *Quel  sujet_i    as-tu      parlé de    t_i?

   which subject have-you talked about

   'Which subject did you talk about?'

2. b. [*]Qui a        parlé  duquel    sujet?

   who have [sic] talked about.the subject²

   'Who talked about which subject?'

3. c. [*]Duquel      sujet_i    as-tu      parlé [t_i?]

   about.which subject have-you talked

   'About which subject did you talk?'

However, neither 12b nor 12c is acceptable in French: there is no suppletion. Duquel should be replaced by de quel. Though the argumentation as such is independent of these concrete examples, the availability of suppletive forms not being a necessary condition for incorporation (Vol. 3:647), it remains unfortunate that data presented here as 'most evident' turn out to be incorrect.

Sten Vikner's Ch. 46 on 'Object shift' offers an excellent survey of the empirical data in the Scandinavian languages and an evaluation of various analyses. The chapter also contains some discussion of scrambling as found in German and Dutch. In Dutch (13), for instance, the object of besproken 'discussed' is separated from the verb by the adverb nooit 'never'.

(13) dat Peter waarschijnlijk dit  probleem nooit besproken heeft

that Peter probably      this problem  never discussed has

It is said that Dutch scrambling 'may cross a preposition' (Vol. 3:411). No examples are given, but certainly for me and the Dutch speakers consulted, prepositions cannot normally be crossed, as shown by 14b. Leftward movement of the DP dit probleem 'this problem' pied-pipes the preposition over 'about' in 14c. I fear that, on the basis of Vikner's discussion, a reader unfamiliar with Dutch might incorrectly conclude that 14b is grammatical.

(14)

1. a. dat  Peter waarschijnlijk nooit [_PP over [_DP dit probleem]] gesproken heeft

   that Peter probably      never    about    this problem    talked      has

2. b. *dat Peter waarschijnlijk [_DP dit probleem] nooit [_PP over [_DP _____]] gesproken heeft

3. c. dat Peter waarschijnlijk [_PP over [_DP dit probleem]] nooit [_PP ______] gesproken heeft

4. d. dat Peter er      waarschijnlijk nooit over    gesproken heeft

   that Peter there probably        never about talked      has

Dutch prepositions can be stranded when a so-called R-pronoun is extracted as in 14d. But such extractions have special properties (see Ch. 51, 636-37, 650-53, van Riemsdijk 1978, Huybregts 1991, Broekhuis 2002:254-313, 2006:48-49).

It is only fair to say that given the size of this collection and the number of contributors, problems such as those illustrated above are probably inevitable. The additional presentational problems arising for the CD accompanying the Companion, however, should not have arisen. Text examples are presented in a quite unusual format using columns. For instance, the CD-ROM version of example 5a from Ch. 62, given here as 15a, is as shown in 15b.

(15)

1. a. [_CP [Grüne Bohnen]_i [_C' glaube   ich, [_CP dass er  gerne         mag [e]_i]]]

           green beens            believe I            that  he very much likes

2. [CP [Grüne	Bohnen]i [C'	glaube	ich, [CP	dass	er	gerne	mag [e] i]]].
   Green	beans	believe	I	that	he	very much	likes
   'Green beans, I think he likes very much.'

Ch. 46 on 'Object shift' contains representation 50, repeated here as 16a. In the CD this comes out rather mangled, as in 16b.

(16)

1. a. . . . [_AgrOP spec AgrOº [_V(aux)P spec Vº (aux) [_V(main)P spec Vº (main)]]]

2. [AgrOP

   spec

   AgrOº

   [V(aux) P

   spec

   Vº (aux)

   [V (main)P

   spec

   Vº main)]]]

The Somali example illustrated in 8a comes out as 17a and for Kurdish 8b we get 17b.

(17)

a.	Somali:	gabadh	fiican	dheer	madow
	girl	good	tall	dark- skinned
		gabadh	fiican	oo madow oo dheer
		gabadh	dheer	oo madow oo fiican
		etc.
b.	Kurdish:	top-a	jwan- a	gawr-a	sur-a- ka
	ball	beautiful	big	red
	gabadh	fiican	oo madow oo dheer
	top-a sur-a gawr-a jwan-a-ka
	etc.

If this type of presentation has been adopted to ensure correct alignments, then it has miserably failed. I add that in the printouts of the CD-files, such example tables often come out cut into two due to page cuts.

There are further presentational problems in the CD-ROM. For instance, in Ch. 35, 'Inalienable possession' by JACQUELINE GUÉRON, words were not always separated, neither in the screen version nor in the printout of the file, making the text hard to read (e.g. 'denoting the possessor . . . languagesand . . . Englishthe'). Fortunately, in the book this problem did not arise. Given the presentational problems with the CD-version of the Companion, it is surprising that, at least for the moment, the publisher did not opt for simply using pdf files.

My overall conclusion is definitely positive. The Companion is a first-class achievement. In these five volumes is stored a lasting legacy of generative research. One can use the Companion as a reference work, in order to check certain data, but if one is lucky enough to have a copy available on one's desk (or in the library), one can explore the volumes by roaming through the chapters and (re)discovering sets of data described in a clear and coherent way. For the syntax teacher, the Companion will be an endless source of inspiration and support for classes at all levels. Students and nonspecialists find gathered in one book research results covering fifty years of research. There remain a few, mainly presentational, glitches in the printed volumes, but neither these nor the badly produced CD-ROM can diminish the achievement: these five volumes constitute a milestone in the generative literature. The Companion has set its aims high and it has undeniably attained them.