Asked about the newly postulated particle muon, Nobel laureate Isaac Rabi replied, “Who ordered that?” This remark could well be applied to the linguistic domain, where the expanding range of features, and the ease with which they are postulated, reminds one of the ‘particle zoo’ in physics.

There are books that say more thanks to what they do not contain than to what they offer. From my own, admittedly idiosyncratic perspective, the volume under review falls into this category. Anna Kibort & Greville G. Corbett have assembled a wide array of contributions, offering diverse ‘perspectives on a key notion in linguistics’, as described in the book's subtitle. Indeed, features appear to be key in virtually all domains of linguistic inquiry. They are perhaps the key notion in linguistics. But, I wonder, at what cost?

The answer to this question, of course, depends on what one takes the goal of linguistics to be. If the aim is to describe the character of grammars of specific languages, then features of the sort routinely posited by linguists (‘Aspect’, ‘Definiteness’, etc.) strike me as just about right. But if the goal is to explain properties of the human language faculty, as a biological organ, then, features of the standard sort acquire a more dubious status, though one that is rarely recognized as problematic. This is so mainly due to two reasons. First, the features posited as atomic clearly lack the elementary character that would allow them to count as plausible primitives. All too often, the labels given to these features are clearly the result of compositional interpretation, the product of interface operations, and yet features are where the analysis starts in virtually all frameworks. A second and related problem is that features are often assumed to combine (or ‘bundle’) to form syntactic atoms (and hence categories are understood as bundles of features), but the nature of the very operation of feature-combination is never clarified. To make matter worse, the structures formed by bundling are reminiscent of the structures formed by syntactic combinations (‘Merge’). The feature structures of the sort one finds in the Kibort & Corbett volume are little syntactic trees. But no one knows or, worse, seems to care about how these little trees are formed. Even more puzzlingly, in many syntactic frameworks (come to think of it, virtually all existing syntactic frameworks), such feature structures are taken to provide essential instructions or constraints for the formation of the bigger syntactic trees (phrases, etc.). This clearly introduces an intolerable degree of redundancy and stipulation. The specter of preformationism looms very large indeed.

Only David Adger's contribution to the volume under review, ‘A Minimalist theory of feature structure’, addresses the issues just raised. All the other contributions take features and feature bundles for granted, and build their analyses on them like one builds houses on sand. Perhaps this is why Adger's chapter is one of my all-time favorite reads. To the editors' credit, it should be pointed out that Adger's chapter, and thereby the volume at large, is unique in the current linguistic literature in addressing the concerns I have raised above. Unfortunately, readers of Adger's chapter will soon discover that syntactic analysis, as it is standardly practiced, requires a fair amount of feature structure that must be stipulated (‘preformed’) before any opportunity arises to construct that structure by applying Merge. This is the lexicalist residue in all current frameworks: the structures of the signs in Head-driven Phrase Structure Grammar, the feature bundles in Minimalism, and, of course, the functional hierarchies among cartographers. All of these illustrate how much there is left to understand before our theories of language will rest on a firm foundation, how much lexicocentrism there is to eliminate. Adger himself recognizes this explanatory gap in his concluding section, and I want to stress that his is a valiant attempt. He has tackled what most people fear to contemplate, and I wish more studies of this ilk could be found in the literature. (For a more extensive and technical discussion of the limits of Adger's attempt, see Boeckx Reference Boeckx2010.)

All chapters in Kibort & Corbett's volume are written by linguists, which is a missed opportunity to include an ‘outside’ viewpoint. In my experience, talk about features invariably induces breakdown of interdisciplinary communication on language. There is nothing more specific, more modular or more isolationist than a grammatical feature. I would have liked to see an attempt to derive some of our standard features from cognitive primitives of the sort that are able to facilitate evolutionary accounts emphasizing descent.

All of this is to say that there is no theory of features in linguistics. This is quite depressing, considering the prominent role features play in all domains of linguistic theory. Features (and their structures) are claimed to be what gives us categories, what provides the units of variation, what drives syntactic derivation, what captures locality constraints, what underlies semantic composition (e.g. semantic types, which are features), what triggers phonological rules, and so on – fulfilling a multitude of functions even if everyone knows that the buck must stop somewhere. But no one has a clue where, perhaps because all too often our attention is focused on what features buy us, when it should be on what buys us these features. The caricature of the elephant as a bundle of elephant features (bundled in an elephant manner) is not very far from many influential linguistic analyses. We should be wary of our use of features, lest we end up like the biologists who put all their (and our) money on genes and genome sequencing activities, and are now left picking up the pieces and wondering what they all mean. It is time that we recover from our severe case of featuritis.