For more than thirty years, Hawkins has been developing a research program of ‘functional typology’, in which psycholinguistically motivated principles of linguistic complexity and processing efficiency are argued to account for aspects of linguistic competence and typological variation. This book is the latest installment in the program, following in the footsteps of Hawkins (Reference Hawkins1994, Reference Hawkins2004). In the present book, Hawkins restates his basic theoretical principles, reviews the main results of previous research and extends the theory to empirical phenomena not covered in the earlier work. The book is well-written and well-edited, lucid, nicely organized, and carefully researched.

Hawkins' central thesis is that other things being equal, language users tend to avoid processing complexity and maximize efficiency, and this tendency is reflected both in the choices that speakers make when there are grammatical alternatives, and in the grammars of natural languages. This intuition is expressed as the Performance–GrammarCorrespondenceHypothesis (PGCH): ‘Grammars have conventionalized syntactic structures in proportion to their degree of preference in performance, as evidenced by patterns of selection in corpora and by ease of processing in psycholinguistic experiments’ (3).

Chapter 1 exemplifies this hypothesis and reviews some predictions and evidence that supports it. A typical case is the Accessibility Hierarchy (Keenan & Comrie Reference Keenan and Comrie1977). On the one hand, the AH accounts for cross-linguistic variation in the relativizability of arguments: for example, if a language permits relativization of direct object but not indirect object, it also permits relativization of subject, which is higher in the hierarchy, but not arguments that are lower, such as possessor. On the other hand, where relativization is permitted by a grammar, it is more difficult to process as we go further down in the hierarchy (Gibson Reference Gibson1998). Such a hierarchy is motivated, in turn, by the assumption that integrating information over more embedded structures is more complex from a processing perspective, and by a principle (Minimize Domains – MiD) that seeks to minimize this complexity.

Chapter 2 articulates three efficiency principles, including MiD, and shows how they can be applied to account for cross-linguistic variation. These principles are updates of Hawkins' earlier proposals. MiD accounts not only for the AH, but for linear ordering tendencies in languages with fixed positions for the verb. Placing a short constituent closer to the verb minimizes the domain in which the processor can identify the phrasal heads and thereby project the argument structure. Hawkins cites corpus data that shows such an ordering preference not only in a verb-initial language like English, but a verb-final language like Japanese. A second principle, Minimize Forms (MiF), predicts that forms themselves will be less complex, and even zero, to the extent that their properties are otherwise predictable. Such a principle predicts, for example, that complementizer that will be omitted in English with a frequency that corresponds to its predictability (Jaeger Reference Jaeger2006). The third principle, Maximize Online Processing (MaOP), holds that ‘[t]he human processor prefers to maximize the set of properties that are assignable to each item X as X is processed, thereby increasing O(nline) P(roperty) to U(ltimate) P(roperty) ratios' (28). This principle accounts for such phenomena as the fact that displaced wh-phrases precede their gaps, wide-scope quantifiers are preferred before narrow-scope quantifiers, and topics precede comments.

In Chapter 3 Hawkins takes on a range of implicit and explicit challenges and alternatives to his program. The major implicit challenge is this: if we have no direct measure of the complexity of a linguistic expression or the efficiency of the processor, how can we make predictions on the basis of complexity and efficiency, and how do we know which (if either) of these is reflected in the data? Moreover, what about the possible differences between what is efficient for the speaker and the hearer? Hawkins' response (in this chapter and throughout the book) is that the number of words and the branching structure that is projected from a string can serve as useful proxies for the complexity of a linguistic expression, even if we do not know the exact processing mechanisms. The major explicit challenge is Mobbs (Reference Mobbs2008: 12), who argues that Hawkins' principles are not principles of performance efficiency but ‘general computational constraints’ that are applicable to constraining accounts in the Minimalist Program. Hawkins argues that in the end, such constraints must be rooted in performance.

Chapter 4 takes up the question of how preferences get translated into rules of grammar. Chapter 5 explores in more detail the implications of MiD for harmonic and disharmonic branching direction. MiD predicts that branching will be harmonic, while disharmonic branching will be relatively rare and avoided through such devices as extraposition. Hawkins shows that disharmonic branching is a feature of non-rigid OV languages, and that it is in just such languages where reordering of constituents is employed to avoid the most serious disharmonies.

Chapter 6 reviews a number of predictions about the form of NPs in VO and OV languages. Most significantly, it is predicted, and the data appear to show, that in VO languages there will be more elements that signal the onset of the NP (e.g. definite articles) than in OV languages.

Chapter 7 is concerned with differences between VO and OV languages. While VO languages maximize efficiency both from the perspective of both MiD and MaOP, OV languages appear to pose a conflict. MiD favors ordering of complements before the head, while MaOP favors the opposite ordering. Both MaOP and MiD favor wh-fronting in V-initial languages, while only MaOP favors it in V-final languages, leading to the correct prediction that wh-fronting will be rarer in V-final languages.

Chapter 8 explores a number of predictions about the properties of NP arguments that follow from the efficiency principles, especially how arguments are organized in terms of grammatical, thematic and case hierarchies.

Chapter 9 explores the interactions between the principles, arguing that certain patterns can be understood as the consequence of several principles preferring the same ordering, while others are the product of principles pushing in different directions.

Chapter 10 concludes with a review and discussion of broader theoretical issues.

Throughout, Hawkins takes pains to contrast his approach to linguistic variation with the conventional one taken in mainstream (i.e. derivational) generative grammar (MGG). There are basically three kinds of variation that Hawkins accounts for in terms of the efficiency principles: linear ordering, overt marking and grammatical hierarchies.

In MGG, linear ordering is accounted for by triggered movement.Footnote ¹ One might say that the availability of alternative orders, for example of the complements of V in English, is a matter of competence, while the weight-sensitivity preference for one ordering over another is, as Hawkins says, a matter of performance (MiD). But Hawkins wants to go further, and explain the competence in terms of the performance. In the English VP, for example, the preference for placing an argument adjacent to the head even when it is heavier than an adjunct follows from MaOP, given that arguments are more central to the interpretation than adjuncts. Only if the difference becomes sufficiently large does MiD win out and push the argument further to the end of the phrase. So, critically, in such a case, a performance principle like MaOP supplants a competence principle that requires the argument to be a sister of the head (e.g. case assignment under adjacency).

Similarly, in Hawkins' performance theory, wh-fronting is not explained in terms of the discharge of features of functional heads, but in terms of MaOP. Languages lack wh-fronting not because they lack the relevant feature (or because the feature is ‘weak’ or somehow ‘inert’), but because it is more efficient in such languages to process wh in situ.

In these cases, and in many other cases involving linear ordering, the architecture of Hawkins' theory predicts rich variability both within and across languages, while an MGG account (of pretty much any vintage) is categorical, contrary to the facts. Such an account can perhaps be augmented with explicit stipulations about the strength or probability of triggering mechanisms, in order to capture the variability, but such a step falls short of genuine explanation. And it is not clear how psycholinguistically grounded weight effects could be incorporated into a competence grammar without effectively obliterating the competence–performance distinction, especially given that the effects appear to have to do with relative weight (Stallings & Macdonald Reference Stallings and MacDonald2011).

Turning to overt marking, it is generally viewed in MGG as a peripheral phenomenon, although zero realization has occasionally fallen under accounts invoking some form of government relation (Rizzi Reference Rizzi and Haegeman1997). And hierarchies, as Hawkins points out (228–229), have not been accounted for at all in MGG, beyond what can be formulated in terms of c-command and islands. It is particularly noteworthy, Hawkins argues (191), that the distribution of gaps in relative clauses follows the same hierarchy as the distribution of resumptive pronouns in relative clauses: forming filler–gap chains is more difficult further down in the hierarchy, filler–gap chains are grammatically excluded in languages following the hierarchy, resumptives occur spontaneously to save more complex structures in languages where they are optional, and are required further down in the hierarchy in languages where they have become grammaticalized.

What is a reasonable conclusion to draw from such observations? Hawkins' conclusion is at first glance a modest one (103): ‘the research done by formal grammarians … is most welcome and the online processing considerations … will hopefully complement their work and allow deeper questions of causality to be resolved in the future … typologists and formal grammarians need to work together … in order to identify the precise cross-linguistic patterns. The formal grammarian gives depth to the syntactic analysis, the typologist gives breadth’. But later on (230), he makes it clear what kind of formal grammar he has in mind: ‘Formal models that stand the greatest chance of being descriptively adequate, according to [the PGCH] view, are those that can accommodate functional principles as grammatical conventions, such as Simpler Syntax (Culicover & Jackendoff Reference Culicover and Jackendoff2005) and those variants of Optimality Theory that seek some functional grounding and explanation for their constraints'.

The implications are thus significant, although of course there are three big ‘ifs’. If the PGCH is on the right track, if the performance principles can be grounded in sound psycholinguistic theory, and if the distributions of linguistic properties found in corpora and in grammars can be sustained, then much of the theoretical apparatus of MGG can be dispensed with as irrelevant to explaining a substantial number of the facts of language.

In summary, Hawkins is on to something with potentially very far-reaching implications for linguistic theory. His work is clearly relevant to typology, given that he offers a psycholinguistically motivated framework for cross-linguistic variation. But the work is not just about typology, it is about grammatical theory. What Hawkins is proposing should be taken very seriously by any linguist who is concerned with explaining any phenomena involving linear order in syntax, the marking of constituents to represent argument structure and thematic roles, and restrictions on long-distance dependencies such as island phenomena and asymmetries. Which is to say, essentially all of us syntacticians. The general perspective that Hawkins is offering calls into question many widely accepted views about grammar, and raises very demanding challenges for conventional approaches to natural language grammar.