Syntactic Relations. A critical survey
P. H. Matthews

Cambridge Studies in Linguistics 114.Cambridge: CUP, 2007.
£22.99, 222 pp., ISBN-13: 9780521608299.

Reviewed by Olivier Simonin

What is a phrase? What is it made of? What is its internal structure? Do all syntactic relations have to be phrasal? P. H. Matthews addresses these and like issues in his latest book on syntax, which is a critical re-examination of widely shared assumptions on the subject. His survey remains exploratory in nature. It does not aim at providing a general theory of syntax, though viable treatments for the difficult problems encountered are always suggested. In fact, its relative non-technicality makes it even accessible to the general reader with an interest in syntax. Yet the publication of such a book, written by a distinguished syntax specialist, in a well-known CUP series (Cambridge Studies in Linguistics), might well be a landmark in the field. It is all the more important since it will lend new respectability to criticism directed at the dominant syntactic paradigm. Linguists of every persuasion will soon find that they can ill afford to ignore its contents.

The “Introduction” (Chapter 1) begins with delineating the field of syntax. Sentences are made up of smaller units which do not combine randomly, and it is the job of syntacticians to account for this. The main difficulty lies in that there are many different views on offer, as to how the combination of atomic syntactic units should be studied. This certainly depends on the way their groupings into syntactic categories are conceived (whether distributionally, semantically, etc.), but the crucial factor is how relations between constituents are understood. For instance, a clause that is subordinate, like that Mary is coming in They said that Mary is coming [3], can be understood as embedded within the matrix clause, or/and dependent on a lexical verb (said here). Subordination thus has two distinct senses. Similarly, in The house in Cambridge, though, they did not build [8], the initial NP has a special, complex, relationship with the rest of the predicate. It is both a (fronted) topic, and an object.

Chapter 2, “What beginners are told”, critically expounds current syntactic orthodoxy, drawing primarily on two pro-generative introductory textbooks: Radford et al. 1999 (Linguistics. An introduction, CUP), and Fromkin 2000 (Linguistics. An introduction to linguistic theory, Blackwell). Nowadays, most syntacticians would agree on a few working assumptions: “every sentence is constructed of a hierarchy of phrases” [11], whose internal structures are deeply symmetrical. Specifically, they always contain one head element which gives the whole phrase its identity, and to which all other elements relate—in accordance with X-bar phrase structure rules. In order to accommodate these tenets, auxiliary hypotheses are called for. A case in point is the existence of “empty” heads, which lack any physical realization. Postulating them makes it possible to say that these can be an NP on its own (these Ø: ‘these things/people/…’), or rather, to follow the analysis now favoured, that people, say, can be a DP with a Ø determiner. Yet positing a universal template for all phrases might suspiciously look like straight-jacketing them.

The next three chapters are concerned with the question of headship. Headship implies the dominance of one element over others, on which they might directly depend. A series of criteria, by and large borrowed (mostly from Zwicky 1985), are put forward for establishing whether a unit qualifies as a head. Some may fail to apply to individual cases, and several might point different ways. Here is a short, compendious list of these criteria, which appear somewhat scattered in the course of discussion:

1. “X heads a unit… if it stands in a specific relation to some Y which is part of a larger unit” [28]. In particular, X may be marked for agreement or government, with respect to elements outside the phrase to which it belongs. In They like coffee, the head of the VP like coffee is like (X), since this verb agrees in number with they, the subject (Y).

2. Another requirement, building on Bloomfield’s observation that a phrase might have the same distribution as just one of its constituents (e.g. tall girls and girls), is that a head must play the primary role in determining a unit (like girls in tall girls).

3. “If a part X depends on a part Y, then, all things being equal, Y heads a whole which includes both of them.” [34] In They said that Mary is coming, the subordinate clause that Mary is coming depends on the verb said. One can then surmise that those two constituents belong to a common unit, headed by said.

4. A dependent presupposes the existence of a head, while a head does not require dependents. This can be used as an argument to suggest that DPs/NPs are headed by nouns, rather than determiners. Whereas determiners such as articles always require nouns, even countable nouns like girl can occur without determiners (especially in the plural form).

5. Omitting a head creates an elliptical effect. In I said tall girls, not short!, the adjective short is not followed by a noun, though the general impression is that girls is implicitly understood.

6. Heads correlate with functors in semantics, i.e. elements calling for arguments. Prepositions and verbs are heads, according to this criterion.

“Heads and dependents” (Chapter 3) introduces these basic concepts, and examines how they relate to three familiar categories: CP, PP, and VP. Tests and observation support the view that syntactic accounts discordant with current phrasal orthodoxy might be as viable, and that the latter might conceal rather than reveal actual problems. Although prepositions are classified as heads, some exhibit an unusual property: they seem to be selected by the phrase that follows. To express time reference, prepositions vary, according to the noun that follows. Hence They met at Easter, but They met on Sunday, or They met in Spring. Note that selecting other constituents is here potentially taken as another criterion for headship.

“Must phrases have heads” (Chapter 4) continues the same exploratory work, focusing on NPs. It extends criticism by impugning the universality of what could be termed unilateral single-headedness, calling especial attention to variation between English and other European languages. In the Latin phrase multas puellas (‘many girls’), both the determiner and the noun agree in case, gender, and number. The agreement relation is bilateral, and it is difficult to establish whether a verb would agree with one or the other. Moreover, puellas multas can occur just as well, or the two words be separated by a verb. Order does not seem to be helpful in determining headship. Overall, explanations become more illuminating once the one-size-fits-all dogma of phrases is rejected, and one recognizes that several elements within a phrase might display distinct aspects of dominance.

This leads to the conclusion that “what is common to the types described as headed may be simply that, in one way or another, the constructions are all asymmetrical” [93] (Chapter 5, “Asymmetries”). To illustrate, consider Huddleston and Pullum’s (2002) view on NPs, which is approvingly quoted, etc. and predicate, although it is usually taken for granted that a subject is part of a VP, Jespersen, for instance, assumed that subjects were heads. He believed that a phrase like a furiously barking dog, headed by dog, is structurally similar to the corresponding sentence A dog barks furiously, so that its subject should be regarded as head. It might be that the subject-predicate relation cannot helpfully be described as phrasal.

The two following chapters turn specifically to the kind of tree structure underlying phrases, in current X-bar syntactic models. The X-bar system ensures that constituents bind in pairs to form units at higher nodal levels, and these units might, in turn, combine with others in the same fashion. After a brief presentation of this system as well as a sketch of its history, Chapter 6 (“Constituents”) examines two main topics: layering or stacking, and the connection of all syntactic relations to constituency. One major problem for X-bar syntax arises with the binary combinations it calls for. When three elements are bound together in a higher unit, it predicts that there must be a layering: two elements must join into a unit first, before forming, with the remaining element, the higher unit. This naturally implies a stacking order. However, in the phrase [a] large house in London, it is not clear that house is first determined  by large, or in London. Perhaps it is determined by both at the same time. If this is right, then one can avoid postulating an invisible structural ambiguity in such cases, which are numerous. Tree structures in general represent relations between immediate constituents only, so that traditional functions like object and subject are no longer given precedence. In What did you eat?,to account for the role of what as object of eat, it is assumed that the object first appears next to the verb, and then moves to its visible position in the sentence, leaving in its wake an unrealized constituent (or more, depending on the model).

One might wonder if such a step is not too costly for a theory, even though it elegantly does away with non-phrasal relations. Indeed, simply stating that long-distance relations may obtain between constituents seems a much more parsimonious solution, since it does not rely on the double expediency of empty constituents and movement. Chapter 7 (“Must constituents reduce to tree structure?”) opens on such considerations. It makes a strong case for treating some syntactic relations as complex networks, which do not easily reduce to orthodox, binary representations. In That made me angry, me is object in the matrix clauseas well as subject in the small clause me angry, and angry depends on made as much as me does. Since every word is related to every other, a classic tree diagram is bound to fail to illuminate the construction. In addition to small clauses, other syntactic types are selected for discussion: group-verbs (take care of), complex predications (I was expected to help), and coordination.

Chapter 8 (“Simplicity”) provides a short overview of the ground previously covered, and underscores the retrospective assessment that “little about syntax is straightforward. Few constructions or types of construction can be represented in just one way, on which every linguist will agree.” [178] The book closes with a few reflections on Universal Grammar and simplicity. The value of a theory that appears to be impervious to counter-evidence, thanks to expedients such as null categories, deserves to be queried. The same holds for the simplicity it claims for itself, if it abstracts away from the details and complications of language, in such a way that it could still have applied to them, had they been quite different.

As may already be apparent from this summary, the discussion can be rambling at times. Presenting headship criteria once and for all would have been helpful. Periodically returning to considerations on syntactic types like NPs can also be frustrating, though perhaps inevitable, owing to the conceptual outline that the book follows. In some few cases, one has the impression that conclusions are reached too quickly. Consider, for instance, this passage [155]:

There is no convincing evidence, we would argue, that built houses by the sea does not have the construction of a verb, an object and a locative: thus [built] [houses] [by the sea]. Equally, we would argue, there is no convincing evidence by which it does not have the construction of a verb and an object: [built] [houses by the sea]. Therefore, we conclude, it has both.

The fact that two alternative solutions are invalid does not imply that the conjunction of both is right. Moreover, the possibility that the distinction between the two analyses is somewhat neutralized is not even mentioned, although neutralization is a venerable linguistic concept.

Professor Matthews’ rich historical knowledge of developments in syntax, especially in the United States, enables him to explain insightfully how the salient features of today’s leading models have emerged. Even though significant figures that are often unacknowledged, like Tesnière and Jespersen, are quoted, one would have wished to hear about relevant and rather recent theories, like construction grammar. This has perhaps been ruled out, because semantic structures are mapped onto syntax in this model, which puts it beyond the purview of syntax, taken in a strict sense. At least a few words on other, formal frameworks, might also have been in order.

Nonetheless, it may still be that the complexities of syntax are perhaps best understood in the light of other domains, like logic. Logical relations applying to one or several elements are usually mirrored by syntactic structure. Othello loves Desdemona expresses the relation love (Othello, Desdemona), to take up one of Russell’s famous examples. The verb indicates a kind of relation, and the two participants involved are realized as arguments. This is not to say that the connection is always straightforward, as the author points out when considering weather verbs. The subject of It rains does not correspond to any element in logical semantics—as P. H. Matthews points out [107]. Among other domains linked with syntax, information structure (Lambrecht 1994) deserves a special mention. Highly topical elements appear first, even when they fail to be subjects (The house in Cambridge, though, they did not build [8]). One might also remember Givon’s (1979) characterization of subjects as grammaticalized topics, a suggestion which goes some way towards accounting for the syntactic discrepancy between subjects and objects. If it is true that looking at domains outside syntax can shed light on constructions, then there is little wonder that syntactic evidence is apt to point in opposite directions, when several external domains come into play.

These, however, are personal observations, and despite all the reservations that I have just voiced, Syntactic Relations is a book that deserves to sit high on every linguist’s syntax shelf.