Publishing Partner: Cambridge University Press CUP Extra Publisher Login

New from Oxford University Press!


Style, Mediation, and Change

Edited by Janus Mortensen, Nikolas Coupland, and Jacob Thogersen

Style, Mediation, and Change "Offers a coherent view of style as a unifying concept for the sociolinguistics of talking media."

New from Cambridge University Press!


Intonation and Prosodic Structure

By Caroline Féry

Intonation and Prosodic Structure "provides a state-of-the-art survey of intonation and prosodic structure."

The LINGUIST List is dedicated to providing information on language and language analysis, and to providing the discipline of linguistics with the infrastructure necessary to function in the digital world. LINGUIST is a free resource, run by linguistics students and faculty, and supported by your donations. Please support LINGUIST List during the 2017 Fund Drive.

Review of  The Oxford Handbook of Linguistic Analysis

Reviewer: Geoffrey Sampson
Book Title: The Oxford Handbook of Linguistic Analysis
Book Author: Bernd Heine Heiko Narrog
Publisher: Oxford University Press
Linguistic Field(s): Computational Linguistics
General Linguistics
Linguistic Theories
Cognitive Science
Issue Number: 26.5216

Discuss this Review
Help on Posting
Reviews Editor: Helen Aristar-Dry


When I was a teacher of linguistics, in the 1970s and 1980s, our students would hear about transformational–generative grammar from the theoreticians, about Michael Halliday’s systemic–functional grammar from the language-teaching experts, and maybe about one or two other approaches. Sometimes they would ask us to dispel their resulting confusion by providing some kind of comparative guide to linguistic theories, perhaps a document which showed how one or a few specimens of English would be treated by the respective theories. I don’t think we ever did what they asked, and it seems to me that we probably could not have compared the theories by reference to common examples, because different theories tend to be interested in different kinds of example. But, far too late for my own students, the book under review could be seen as an attempt to satisfy that request. (Though, at more than a thousand pages before the bibliography is reached, I doubt if the students would have thanked us if we had been able to refer them to it.)

Linguistic theories have multiplied since the 1980s. After the editorial introduction, the book contains 39 chapters each presenting a different approach to language description and analysis. (In a few cases, two or three chapters are devoted to separate aspects of one theory.)

This is a new edition of a book first published in 2010; it has been expanded to include seven additional chapters. The new chapters are about topics such as language acquisition, neurolinguistics, phonetics, phonology, and semantics, whereas the chapters carried over from the first edition mainly focus on the central linguistic topic of grammar. (There is no indication that chapters which appeared in the first edition have been revised for this edition.) Rather more than half of the contributors are based in North America, most others in Europe, with a sprinkling from Australia, New Zealand, and Japan. I shall not take the space to list every chapter, but references to individual contributions as the review proceeds will give an impression of the overall range of topics and theories covered.

As one might infer from the topics of the new chapters just mentioned, not all contributions are concerned with rival theories of the same subject-matter. Patrice Speeter Beddor’s “Experimental phonetics”, for instance, appears to be entirely compatible with any particular theory about syntax. (Indeed, although this chapter is a very informative and interesting one, it is not obvious how it and some of the other new chapters belong in a book about “linguistic analysis”. It may be that the editors decided that their new edition should cover the language sciences more comprehensively than the first edition but were reluctant to modify the book title to reflect that.) The bulk of contributions, though, offer competing theories about the same or at least largely overlapping topics. Some authors make this rivalry quite explicit, for instance Vilmos Ágel and Klaus Fischer (“Dependency grammar and valency theory”) discuss the question “Is D[ependency] G[rammar] the best of the theories presented in this handbook?”

Contributions differ, too, in the extent to which they are partisan. Some are not; for instance Eric Pederson (“Linguistic relativity”) is admirably even-handed in discussing both arguments for and arguments against what is often called the Sapir–Whorf hypothesis (a term which Pederson regards as a misnomer). A larger number of contributors, though, see their chapters as opportunities to put their pet theories in the shop window and win converts. Some are more assertive about this than others. Many 21st-century linguists may not be surprised to hear that the highest levels of self-confidence are displayed by the generative grammarians. Cedric Boeckx (“Linguistic minimalism”) makes no bones about telling us that “There is no question that the minimalist program is the right strategy to account for properties of FL [i.e. the language faculty]. Its conceptual/methodological legitimacy can hardly be questioned …” Readers will no doubt make their own minds up about that.

Another respect in which contributions vary is in terms of how much prior knowledge they assume. Probably no-one would consult this kind of book unless they had at least embarked on undergraduate study of linguistics, but many contributions could be read by people with no deeper acquaintance with the subject than that. Several contributions, though, seem to expect readers already to know a fair amount of linguistic theory of the type relevant to the respective chapter. Thus, Yan Huang (“Neo-Gricean pragmatic theory”) remarks that “the constraints on Horn-scales … proposed by Levinson, successfully rules [sic, for ‘rule’] out [a given symbol-sequence] as forming a genius Horn-scale”. The term “Horn-scale” is eventually explained on a later page, for the benefit of those like myself who have never read Levinson, but “genius Horn-scale” appears nowhere else in the chapter, so far as I have seen, and I am sure I am not the only linguist who will be baffled. Or again, Guglielmo Cinque and Luigi Rizzi’s “Cartography of syntactic structures” chapter is dense with references to journal literature, many of which are far from self-explanatory. They write that

The very simplified structural representations often assumed in the
minimalist literature, expressed by the C-T-v-V system, are sometimes taken
literally … but the structure of the arguments rarely implies a literal
interpretation, and often is compatible with an interpretation of C-T-v-V as
a shorthand … with C, T, and v taken as ‘abbreviations’ standing for complex
zones of the functional structure.

C, T, etc. seem to be explained nowhere in Cinque and Rizzi’s chapter; to me their statement is totally opaque.

One “grammar” chapter, Martin Haspelmath’s “Framework-free grammatical theory”, differs strikingly from others, in that it argues that linguists should avoid allowing their descriptive metalanguage to embody assumptions about universal features of human language. If there are such universals, they should emerge from empirical descriptions of individual languages rather than being imposed on them. Haspelmath notes that “outside the field of linguistics, metalanguages do not seem to have the role of excluding impossible phenomena”. (Balthasar Bickel’s “Distributional typology: statistical inquiries into the dynamics of linguistic diversity” makes a related point when Bickel writes “There is no need to formulate one’s explanatory theory in a metalanguage that is full of notions that are unique to linguistics … and totally insulated from the rest of the cognitive and social sciences”.) Many other chapters concerned with grammar do appear to take for granted that an aim of general linguistic theory should be to devise a notation for grammatical description which permits only those languages deemed to be learnable by humans to be defined.

One consequence is that these chapters bristle with diagrams and formalisms of many different kinds: apart from familiar tree structures, various contributors also use boxes linked by arrows of diverse sorts, algebraic symbols, data-structure diagrams of the kind used in software engineering, and other things. A prize for most complex diagrammatic notation should probably go to Alice Caffarel-Cayron (“Systemic functional grammar and the study of meaning”), for a diagram labelled “Register variation and instantiation” which contains pairs of tangent circles, spreading rays, double-headed arrows in different orientations, and a curly bracket. (Whether all these elements have well-defined meanings in this author’s theory, or are intended just to suggest relationships in a more intuitive, vaguer fashion, is not entirely clear.)

The editors write that they found it impossible to organize the contributions into any logical thematic sequence, so they simply arranged them in alphabetical order of the main word in the title. This has produced an odd jumble of topics, for instance after the editors’ introduction the first two chapters are Eve Clark’s “Linguistic units in language acquisition” and Talmy Givón’s “Adaptive approach to grammar”, because “acquisition” and “adaptive” both begin with A. (In fact this scheme is not carried out entirely consistently. Francisco Yus’s “Relevance theory” is placed among neither the Rs nor the Ts, but next to the thematically-related chapter by Yan Huang mentioned above.)


For anyone wanting to look into some current linguistic theory which he has heard of but knows little about, this volume would be a good place to start. It will show what issues seem important to advocates of the theory in question, and in most cases will point the reader towards other publications that allow him to go deeper than is possible in a single chapter.

Nevertheless, the book as a whole left me rather depressed. The overall picture it presents of the current state of linguistics is probably an accurate one, but it is not a pretty picture.

Linguistics is supposed to be a scientific discipline – a standard one-line definition of the subject is “the scientific study of language”. Cedric Boeckx justifies the linguistic theory he presents here in terms of how natural sciences such as physics and chemistry work. But the evidence of this book makes it hard to take the scientific pretensions of linguistics seriously. It is normal and healthy, of course, for a science at any given moment to contain plenty of disagreement and rival views. But alongside differences about some issues, one expects to see progressive convergence towards agreement in other areas. At least, surely, one expects a substantial degree of agreement about the nature of the data to be accounted for, and about what would count as acceptable explanations, if they should be empirically supported. With this book, I have problems both with the contributors’ concept of linguistic data, and with their ideas about what count as good explanations of data. If I give specific examples, readers should please believe that I have no wish to pillory the particular contributors I quote; they happen to offer particularly clear examples of tendencies which pervade the book, and pervade much of present-day linguistics generally.

As an example of the data problem, Vilmos Ágel and Klaus Fischer claim that one difference between Hungarian and English is that the English verb ‘lie’ (tell an untruth) cannot take a complement clause expressing the content of the lie. Can it not? I googled ‘lied that’ and was offered “about 328,000 results”, beginning with ‘Have you ever lied that you had a boyfriend …’ and ‘Kelly Baker lied that a young member of her family had cancer …’ In what sense should these examples not count as English? They look perfectly normal to me. (A few of the Google examples were irrelevant because they used the noun ‘lied’ meaning a type of song, but those cases were a small minority.) At one time questionable claims about “starred sentences” could be refuted only by reference to purpose-built electronic corpora to which many linguists had no access, but nowadays Google supplies anyone with information like the above in seconds. I realize that some linguists will want to say something like “the ‘lied that’ construction may be frequent in performance but is not part of native speakers’ true linguistic competence”. However, I just do not see what sense to make of such a statement. It is like a statement “No true Englishman fears death in battle”; it is mystical rather than scientific.

Turning to the issue of satisfactory explanation: Yan Huang remarks that “we can say ‘They summered in Scotland’ [but] cannot say *‘They falled in Canada’ ”, and he explains this via a linguistic principle of “pre-emption”: the fact that ‘fall’ has the verb sense “drop down” blocks it from being used as a verb similar to ‘summer’ meaning “spend the relevant season”. It is characteristic of modern linguistics to posit abstract linguistic principles in order to explain facts which can readily be explained without reference to linguistic theory. In England we do not use ‘fall’ as the name of a season, we call that season ‘autumn’ (a word which has no alternative sense), but we too would be unlikely to say ‘They autumned in Canada’. (NB “unlikely”, not “cannot say” – I have just said it, or rather written it.) The reason is that there are (or at least have been) recognized, established social institutions, among those whose circumstances allow(ed) it, of spending whole summers, or whole winters, away from home in places with pleasanter weather; hence ‘to summer’ and ‘to winter’. In spring and autumn the weather is not extreme, so there has been no established custom of spending those seasons away. Consequently we do not usually say ‘to spring’ or ‘to autumn’, and Americans do not usually say ‘to fall’ (in that sense). But if a new custom should arise (presumably for some non-climate-related reason) of spending whole autumns away from home, very likely Englishmen would begin saying things like ‘They autumned in Wigan’, and Americans would start using ‘falled’ the same way.

Noam Chomsky’s ‘Aspects of the Theory of Syntax’ (1965) posited a complex system of linguistic “selection rules” and “subcategorization rules” in order to enable his transformational grammars to disallow sentences like:

the boy may frighten sincerity
John amazed the injustice of that decision
the book dispersed

I am sure that the particular formalisms advocated by Chomsky and by other linguistic theorists nowadays will be diverse, and different from those of ‘Aspects’. But few theorists ever seem to have realized that there was never any need for such formalisms. The examples quoted are odd not because they are “poor English”, but because they are good English and say, quite explicitly and clearly, things that no-one is likely to want to say since they are obviously untrue. (Sincerity and injustice are abstractions, and abstractions do not feel emotions. To disperse means for a closely-packed set of separate units to spread apart, but a book is not a set of separate units and hence cannot disperse.) I would be surprised to encounter an English sentence ‘The train climbed from Johannesburg to Pretoria’, but to me it would be absurd to postulate grammatical machinery to disallow it. It is an odd sentence because Pretoria is about 1700 feet lower in elevation than Johannesburg, which is not a fact about the English language or about Universal Grammar.

To my mind the whole idea of “starred sentences” is a highly questionable one, though much modern linguistic theorizing takes it for granted. My wife and I recently addressed the problem of one of our cats stealing the other’s food by buying a new type of feeding-stations with lids that open and close automatically under the control of the individual pet’s microchip. The makers, Sureflap, are an English firm founded recently by a Cambridge physicist, and the manual provided is well written. So I was initially surprised to encounter a section headed “Learning your pet into the feeder” and beginning “When learning your pet into the feeder, make sure all other pets are kept away”. (It explains how to get the mechanism to respond to a particular pet.) Surely these word-sequences are not English? – ‘learn’ does not take an animate object, or an ‘into’ phrase. But the activity described is novel, and the writer has used English in a novel way to refer to it. I might have preferred to write “Teaching the feeder to recognize your pet” – but that wouldn’t be quite right, because the change to the feeding-station is instantaneous, brought about by a single press of a button, it is the cat which has to be gradually taught to exploit its resulting behaviour. Perhaps there would be some other form of words which would have been faithful to that reality and yet deviated less from established usage; but the manual writer chose the words I quoted, and he or she is doubtless as much an English native speaker as I am, so who am I to say the wording is not English? It did not seem so previously, because no English-speaker had found occasion to use ‘learn’ that way. But now someone has had a reason to use ‘learn’ with that grammar, and I and other native speakers can certainly understand what is intended. If Sureflap prospers, in years to come probably no-one will bat an eyelid at this way of using ‘learn’.

Many theoretical linguists have a concept of “grammaticality” according to which, at a given time, some fixed (though infinitely numerous) class of word-sequences are “grammatical” in a given idiolect, though from time to time the rules of the language or idiolect change so that new word-sequences become grammatical. They would describe the ‘learn your pet’ usage as one that is currently ungrammatical (for most speakers) but which may be destined to become grammatical, under the influence of things such as the Sureflap manual. To my mind this concept of “grammaticality” is a myth. Putting words together in novel ways in order to express novel ideas is part of competent language behaviour. The ‘learn your pet’ example may be a rather extreme case which, in 2015, would make many English-speakers boggle, but less extreme cases are normal. To quote John Taylor (2012: 285):

speakers are by no means restricted by the generalizations that they
(may) have made over the data. A robust finding from our investigation
is that speakers are happy to go beyond the generalizations and the
instances that they sanction. Speakers, in other words, are prone to
_innovate_ with respect to previous usage …

Anna Babarczy and I have argued, at length and by reference to concrete statistical evidence (Sampson and Babarczy 2014), that the distinction between “grammatical” and “ungrammatical” in natural languages is an unreal one. So far as we are aware, no linguists before Chomsky’s ‘Syntactic Structures’ of 1957, not even formal linguists, ever used such a concept. But giving up that conceptual distinction undermines a great deal of what theoretical linguists believe they are doing. There are far fewer facts standing in need of explanation by linguistics than linguists commonly suppose.

Another recurring feature of the volume reviewed which seems questionable for a subject that regards itself as scientific is an undue deference to intellectual authority, evinced by many (though certainly not all) contributors. Even some who disagree with Noam Chomsky’s ideas about language nevertheless quote his writings as somehow licensing their own enquiries, rather like a mediaeval proto-scientist who felt bound to cite Aristotle before launching into an investigation which might in fact have owed little to Aristotle’s ideas. Defending a controversial idea about the cognitive abilities of human babies and apes, for instance, Ray Jackendoff (“The parallel architecture and its place in cognitive science”) writes “It is possible to read certain passages of Chomsky as endorsing such a claim”. I am not sure that this kind of deference to an influential individual is healthy for any modern scientific discipline (and I certainly do not believe that linguistics is exceptional in that respect – cf. Sampson 2015).

Conversely, some contributors ignore prior work which should not be ignored – not because it is entitled to deference, but because it is so well established that arguments for contrary points of view are unpersuasive if they do not give explicit reasons for rejecting the established view. This problem is particularly noticeable in the chapters on word meaning. Meaning in natural languages was being discussed intensively by philosophers before linguists had much to say about it; and if there was one thing that English-speaking philosophers of the 1950s and 1960s agreed on, it was that words do not have fixed, definable meanings. (As philosophers commonly put it, there is no distinction between analytic and synthetic statements.) The idea was argued at length by writers such as Ludwig Wittgenstein, Morton White, and Willard Quine, and when I first encountered professional philosophy in the 1960s it was a leitmotiv of that discipline. Many linguists today, including several contributors to this book, believe not only that word-meanings can be defined but that they have promising ideas about how to do it. Yet they rarely even mention that numerous leading members of a neighbouring discipline have given reasons to believe that it is impossible. (That could be reasonable, if linguists had already produced convincing refutations of the philosophers’ view – there is no obligation on a book like this to rehearse past errors of other disciplines. But the truth is that they never have.)

According to Ray Jackendoff, “nearly universal[ly]” the word ‘ghost’ can be defined as “a mind (or soul) lacking a physical body”. Jackendoff seems to know more about ghosts even of an English-speaking variety than I do. (If the Christian doctrine of the Second Coming is correct, are all the departed appropriately referred to in the interim as ghosts, as Jackendoff implies?) As for “nearly universal”, does Jackendoff really have evidence that most current and past human cultures have had a concept neatly equivalent to English ‘ghost’? That would surprise me. Cliff Goddard (“The natural semantic metalanguage approach”) apparently believes that all word-meanings in all languages can be defined in terms of 65 semantic atoms together with associated grammatical properties. Thus, his definition of ‘something long’ runs:

when someone sees this thing, this someone can think about it like this:
“two parts of this thing are not like any other parts,
because one of these two parts is very far from the other”
if someone’s hands touch this thing everywhere on all sides,
this someone can think about it in the same way

(‘Hands’ would be further decomposable into atoms of meaning.) Does Goddard believe that ‘long’ said of a stick or a cucumber is a different word from ‘long’ as in ‘a long way from the Earth to the Moon’ (which cannot be “touched on all sides”)?

I found some of the “new” chapters in this volume more interesting than the contributions carried over from the first edition. They discuss concrete facts that are not well-known and which seem clearly relevant to understanding how language works, whereas most of the “old” contributions discuss material which in itself is familiar and trivial, and their only concern is how best to organize that material into formal models of languages. Even when the material does require to be accounted for within linguistics, which (as we have seen) is often not so, it is easy to feel “well, we could describe the facts this way, or we could use that model”, and hard to see what might hang on the choice of theory.

If an educated but sceptical non-linguist tried to take the measure of linguistics by reading this book, it strikes me that when he put the questionable concept of grammaticality together with the lack of convergence among different approaches, he could well come to suspect that “theoretical linguistics” is little more than a self-perpetuating non-subject. He might ask himself whether practitioners are repressing awareness of the flimsiness of its foundations, simply because their livelihoods depend on the survival of the discipline. That suggestion may be entirely mistaken. But I am sorry to say that, on the evidence of the book reviewed, it is not obvious to me how the sceptic should be answered.

It may seem that this review amounts to a seriously negative evaluation of the book, but it is not. It is a negative evaluation of the current state of the discipline of linguistics. The book reflects this accurately, I believe.


Chomsky, Noam. 1957. Syntactic Structures. ’s-Gravenhage: Mouton.

Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: M.I.T. Press.

Sampson, Geoffrey. 2015. Rigid strings and flaky snowflakes. To be in Language and Cognition. Online at < >.

Sampson, Geoffrey and Anna Babarczy. 2014. Grammar Without Grammaticality. Berlin: de Gruyter.

Taylor, John. 2012. The Mental Corpus. Oxford: Oxford University Press.
Geoffrey Sampson studied Chinese at Cambridge University and linguistics and computer science at Yale. He taught linguistics, and later computer science, at the LSE, Lancaster, Leeds, and Sussex Universities, with sabbaticals at Swiss and South African universities and British research institutions. He has published on most areas of linguistics. Since becoming professor emeritus at Sussex he has been a research fellow at the University of South Africa.

Format: Hardback
ISBN-13: 9780199677078
Pages: 1216
Prices: U.K. £ 110