In the 1980s, Chomsky argued that syntactic theory should focus on ‘core grammar’, setting aside ‘a periphery of marked elements and constructions’ (Chomsky Reference Chomsky1981: 8). This view was quite influential at one time, but it has been apparent for some time that there is no clear distinction between the core of grammar and the periphery. Moreover, it has been clear since Culicover (Reference Culicover1999) that there are many constructions with idiosyncratic properties which might be consigned to the periphery. One such construction, discussed in Culicover (Reference Culicover1999) and more fully in Culicover & Jackendoff (Reference Culicover and Jackendoff1999), is the comparative correlative (CC), exemplified in English by sentences like The more you read about it, the more interesting it becomes. The book under review provides the first book-length discussion of the construction. Adopting the usage-based approach of Bybee (Reference Bybee2006) and others, and drawing on a variety of corpora, it investigates the Modern English CC construction, its Old English and Middle English predecessors, and its counterparts in German and various World Englishes. It presents a lot of data in all these areas and provides an interesting discussion of the analytic issues. Anyone who has doubts about the usage-based approach will have reservations about some aspects of the book, but it should be of interest to a variety of researchers.
The book consists of six main chapters and a brief conclusion. After a general introduction in Chapter 1, which outlines the usage-based theoretical framework and the corpus-based methodology, Thomas Hoffman looks, in Chapter 2, at previous synchronic research on both English and other languages, and relevant diachronic work. Among other things, he is quite critical here of the Minimalist proposals of den Dikken (Reference den Dikken2005).
In Chapter 3, Hoffman focuses on the diachronic evolution of the construction. He notes that Old English had a variety of constructions expressing the CC meaning and that the modern construction only emerged in the Middle English period, and he emphasizes the importance of analogy in this process. He also notes that complementizers declined in frequency in the first clause of the construction (C1) and that there was a preference for subject–verb order and not verb-second order in the second clause (C2). In other words, the two clauses came to look quite similar.
In Chapter 4, Hoffman turns to the Standard British and American English CC construction. He shows that complementizers are rare in C1 and possible in C2 (125), and that subject–auxiliary inversion is rare in C2 (126). Thus, the two clauses look very similar. He shows, however, that the C1 comparative phrase typically has a rising tone and the C2 comparative phrase a falling tone. He also shows that the two clauses tend to have parallel structures. In particular, the comparative phrases are typically the same part of speech, and the copula is commonly present or missing in both clauses.
In chapter 5, Hoffman considers the German CC construction. Whereas the English construction has the in both clauses, the German construction has je in the first clause but je, desto or umso in the second clause, as the following illustrates:
Hoffman notes that verb-placement shows clearly that C1 is a subordinate clause and C2 a main clause. He also notes that whereas copula omission (as in e.g. The better the students, the better the grades) is preferred in English, it is dispreferred in German (181) and that there is less of a preference for parallelism between the two clauses in German (182).
In Chapter 6, Hoffman looks at the CC construction in World Englishes. Finally, in Chapter 7, he summarizes the main conclusions of the book.
Arguably, the main question about the CC construction is whether it can be analysed within a framework such as Minimalism/Principles and Parameters theory or Categorial Grammar in which the grammar is just a few general mechanisms and constructions are epiphenomena or whether it requires a system of constructions of some kind. Den Dikken (Reference den Dikken2005: 498) argues for the first view, claiming that the construction is ‘analyzable in keeping with the principles and parameters of UG’. However, as Abeillé & Borsley (Reference Abeillé and Borsley2008) note, he does not provide an analysis and ignores key facts. Most work on the construction has assumed the second view, and Hoffman is no exception. However, his usage-based approach is different from some other versions of construction grammar.
It is not entirely clear to me how Hoffman’s approach views constructions. For the Sign-Based Construction Grammar framework of Sag (Reference Sag2010), they are constraints on linguistic expressions. One might suppose they have the same status for Hoffman. However, in a number of places (105, 169), he talks about the possibility of adopting a constraint-based approach as if this is one among a number of possibilities. It looks, then, as if they might be constraints but might also have a different status.
One thing that is clear is that Hoffman’s constructions are non-local. Whereas, constructions in Sag (Reference Sag2010) and related work only refer to an expression and its daughters, Hoffman’s refer to indefinitely large structures. It is worth noting that it is not easy to employ such constructions in parsing unlike the purely local constructions of Sag and others.
Hoffman proposes (140) that English has a CC construction with the following form (where 
$\uparrow$ indicates a rising tone and 
$\downarrow$ a falling tone):
This makes it look as if the is not a constituent of the comparative phrase. (The representations provided for German suggest similarly that je, desto and umso are not constituents of the comparative phrase.) Coordination examples such as The more articles and the more books I read, … suggest that the is a constituent of the comparative phrase. So do examples where the and the following comparative are not clause-initial, e.g. the following (103):
A further important fact about (2) is that the two clauses are identical apart from the tone of the comparative phrase. This suggests that there is a generalization to be captured. Sag (Reference Sag2010), building on earlier work, captures the similarities between the two clauses with a the-clause construction, which they both instantiate. Hoffman criticizes this on the grounds that the two clauses differ, specifically that the fillers have different tones. But treating them as instantiations of a single construction does not say that they are identical, just that they share some properties. One might treat verb phrases and prepositional phrases as instantiations of a single head–complement construction, but obviously they differ in various ways. One would expect the two clauses of the CC construction to differ if the first is a subordinate clause and the second a main clause, as Sag and others have proposed. However, Hoffman rejects this view (138–139). He suggests that the fact that extraction is possible from C1, as illustrated in (4), from Culicover & Jackendoff (Reference Culicover and Jackendoff1999: 564), is evidence that it is not a subordinate clause.
But while it has sometimes been claimed that extraction is not possible from adjunct clauses, Chaves (Reference Chaves2012: 468) shows with examples like the following that it is perfectly possible:
One might also suppose that it would be easier to extract from an adjunct clause that precedes the main clause. Hoffman also doubts whether C2 is a main clause. However, in Chapter 2, he notes evidence from Culicover & Jackendoff (Reference Culicover and Jackendoff1999) for its main clause status, e.g. the following contrast (42):
Of course, the CC construction is not a standard example of an adjunct clause + main clause structure. C2 is unlike a typical adjunct clause in being obligatory, and C2 is unlike a typical main clause in sometimes allowing a complementizer. However, it does not follow that it is not an adjunct clause + main clause structure at all. It seems to me that the English CC construction, like the related if–then construction, is a non-standard adjunct clause + main clause structure (Borsley Reference Borsley, Bonami and Hofherr2011).
Hoffman’s discussion suggests that the usage-based approach, unlike most work in syntactic theory, is not really concerned to capture generalizations. This is also highlighted by the fact that there is little discussion of other constructions which are similar in certain respects to the CC construction or its component clauses. If–then clauses are mentioned, but they are not discussed in any detail, and there is no discussion of exhaustive conditionals, which are like the two clauses of the CC construction in allowing copula omission:
The idea seems to be that the generalizations that syntacticians notice do not necessarily form a part of speakers’ mental grammars. This may be true, but it seems unwise to abandon the traditional syntactic practice of seeking to capture generalizations at the present time.
It also appears that the usage-based approach differs from a lot of work in syntactic theory in being concerned not just with what is possible but also with what is likely. Hoffman seeks to account for the fact that the clauses of the English CC construction tend to have parallel structures by positing three ‘meso-constructions’, subtypes of the CC construction with similar fillers in the two clauses. These essentially highlight possibilities that are allowed by the general CC construction. It is not really clear how they are supposed to account for the parallelism tendency. Moreover, it is not obvious that it should be seen as a matter of grammar. Gricean considerations may lead speakers to make the two clauses of the CC construction as similar as possible just as they may lead them to make the two conjuncts of a coordinate structure as similar as possible. It may well be, then, that this is not something that the grammar should be expected to deal with.
There is much more that could be said both about the virtues of the book, and about features that some would question. I think it has some important weaknesses. However, it is of considerable interest and deserves the attention of anyone interested in the CC construction or peripheral constructions more generally.
