| Human Nature Review ISSN 1476-1084 | Table of Contents | What's New | Search | Feedback | Daily News | Submit A Manuscript |
PDF of this article
Download Adobe Acrobat Reader
Email the reviewer
Reviewer's web site
Search for papers by Trask, R. L.
Author's web site
Search for papers by Baker, M. C.
Publisher's web site
Send a response to this article
Search the web for related items
Contact the Editor

The Human Nature Review  2002 Volume 2: 77-81 ( 13 February )
URL of this document http://human-nature.com/nibbs/02/trask.html

Book Review

The Atoms of Language: The Mind’s Hidden Rules of Grammar
by Mark C. Baker
New York: Basic Books. 2001.
xi + 276 pp. ISBN 0-465-00521-7

Reviewed by R. Larry Trask*, Professor of Linguistics, School of Cognitive and Computing Sciences, University of Sussex, Brighton, BN1 9QH, UK.

Since the pioneering work of Joseph Greenberg in the 1960s, linguists have realized that the grammars of languages do not normally consist of arbitrary collections of properties, but that instead they tend strongly to fall into more-or-less well-defined constellations of properties. In spite of a great deal of investigation, the reasons for these consistent though far from exceptionless co-occurrences of grammatical properties have remained largely obscure.

In the early 1980s, the linguist Noam Chomsky proposed a speculative but spectacular explanation. As part of his more general view that human beings are genetically endowed with the principles of a “Universal Grammar” (UG), Chomsky suggested that the grammars of natural languages could differ significantly only in the “settings” for a (presumably small) number of possible “parameters”. In other words, every child is born already “knowing” what the principles of grammar are and what the possible settings are for each parameter, and acquiring a first language consists only of guessing, from the linguistic input, what the appropriate setting is for each parameter in the language being acquired. (It is usually assumed that any given parameter has only two possible settings.)

This proposal has attracted enormous attention, and in some quarters it has already achieved the standing of a piece of truth. Nevertheless, it has been fiercely criticized by linguists and others, on a variety of grounds. One of these grounds is this: the data on which Chomsky’s proposals are based are drawn almost entirely from the study of English and a handful of other European languages which are known to share a common ancestry with English. In the view of the critics, sweeping universalist generalizations based on such a tiny and unrepresentative sample of the world’s languages are at best premature and at worst absurd.

This new book by Baker (hereafter ‘B’) attempts to address this criticism, and it moreover offers a more complete outline of Chomsky’s parametric view than has otherwise appeared in the literature. The book is more popular than specialist, and B writes well, so non-linguists should find it easily readable, except perhaps for a few minor details of notation.

The amount of linguistic content is rather modest not least because B devotes so much space to pursuing a perceived analogy between the parametric view of grammar and the periodic table of chemistry. It might have been wise to spend less time on chemistry and more on linguistics.

B’s idea is to sketch out a set of parameters and their settings, and to argue that these parameter settings correctly predict the possible combinations of observed grammatical properties in languages, and even to some extent the observed frequency of the possible combinations in the world’s languages. To pursue this goal, he looks in some detail at the grammar of the American language Mohawk about as different as you can get from a typical European language and much more briefly at the grammars of other languages of Africa, Asia, the Pacific and the Americas.

Now, before I begin my discussion of the details of B’s argument, I must lodge a protest against one very general point. Throughout the book, B repeatedly asserts that there can be no alternative to his parametric account of grammatical variation other than a mere vague appeal to culture. This appeal to culture he easily and properly disposes of, but then probably no linguist would defend such a notion anyway. What B fails to do is to acknowledge the existence, or even the possibility, of purely linguistic accounts other than the parametric one. B may well consider the other proposals, such as the restricted-pathways view mentioned later in this review, to be unsatisfactory, but he should at least recognize their existence, and not try to deny the very possibility of other accounts a priori.

The centrepiece of B’s presentation is an account of Mohawk. Mohawk is a polysynthetic language, in which noun objects can easily be incorporated into the verb. So, for example, ‘John likes the baby’ can be expressed either as ‘John he-likes-her the baby’ or in a form glossed literally as ‘John he-baby-likes’, with ‘baby’ incorporated into the verb. Observing the extensive obligatory verbal agreement in Mohawk, B proposes a parameter, the “polysynthesis parameter”, which is supposedly set at “yes” in Mohawk but at “no” in European languages. In a language with the “yes” setting, we are told, verbs must include some overt expression of all subjects and objects. B goes on to conclude that this “yes” setting makes a number of predictions for a grammar possessing it:

(1) Subject-agreement and object-agreement in the verb are obligatory;

(2) The order of phrases in a sentence is free;

(3) Object-incorporation is possible;

(4) Verb-forms lacking agreement, such as infinitives, are prohibited;

(5) Independent reflexive forms like English himself are prohibited;

(6) Independent indefinite nominals like English anybody and nobody are prohibited.

All these predictions are true for Mohawk. Now, by B’s reasoning, any language for which (1) holds has the “yes” setting for polysynthesis, and therefore (2) - (6) must also hold. But this prediction is plainly false. Take Basque. In Basque, statement (1) holds: a Basque finite verb must agree in person and number with its subject, its direct object, and its indirect object, and it must do so regardless of whether the phrases agreed with are overtly present or not just as in Mohawk. Statement (2) is also true for Basque. But all of the remaining statements are false in Basque: Basque has no object-incorporation; it has several verb-forms lacking agreement; it expresses reflexives with independent forms comparable to himself; and it has independent indefinites like anybody. B’s interpretation therefore appears to be falsified.

Just to make things worse, B states on p. 50 that a language must choose between case-suffixes and verb-agreement to perform the task of exhibiting the argument relations between verbs and nominals subjects, direct objects, indirect objects. This is an implicit parameter not discussed further in the book. But his statement is false: Basque does both of these things, in spades, and there are other languages which do neither. The Amazonian language Nadëb even combines very extensive object-noun incorporation into verbs with a general absence of both case-marking and verb-agreement a catastrophic counterexample to B’s position. The great linguist Edward Sapir famously remarked that “all grammars leak”, but the parametric approach begins to look uncomfortably like a sieve.

B himself goes on, on pp. 143-155, to note that his polysynthesis parameter also predicts the wrong outcome in several other languages, which exhibit various mixtures of the predicted and unpredicted properties. He therefore proposes three possible ways of fixing the problem: (1) split a parameter into two or more smaller and more specific parameters; (2) accept that parameters can make conflicting requirements, and introduce machinery for resolving conflicts; (3) allow a parameter to have more than two settings. He prefers the third way out, but he acknowledges that no one of these routes appears to work satisfactorily at present. Indeed, his revised version of his polysynthesis parameter, on pp. 150-151, asserts only that a language may set the parameter to require overt expression in verbs of all, or any, or none, of the participants. This seems to amount to nothing more than an admission that a language can do anything it damn well pleases. In the end, B contents himself with an expression of confidence that a solution can be found.

The existence of the (apparently numerous) “mixed” or “compound” languages, with complicated combinations of properties that seemingly cannot be neatly fitted into any sets of parameters at all, represents an enormous obstacle for B’s parametric account, and this is an issue which clearly merits far more attention than it receives here. On pp. 166-167, B suggests that the grammars of some of these refractory languages must contain various ad hoc “movement” rules, whose only function is to get things out of the positions where the parameters say they should be and into the positions where they actually occur. But this kind of “solution” is no solution at all: it reminds me of Ptolemy’s epicycles upon epicycles, to force the resolutely elliptical planetary orbits into the required circles.

B goes on to survey a number of the prominent grammatical differences which have been reported among the world’s languages: differing word orders, presence or absence of adjectives, presence or absence of ergativity, presence or absence of topic-prominence, presence or absence of question-movement, differences in reflexive domains, and so on. For each difference, he at once proposes another parameter, finally obtaining a list of thirteen explicitly stated parameters, a list which is clearly not intended to be exhaustive, and which does not include one or two others noted implicitly but not present in his final list.

I have two observations on this procedure. First, it looks unprincipled. If every grammatical difference we can uncover is to be at once assigned to yet another putative “parameter”, then we are in danger of reducing the entire parametric approach to nothing more than a mere list of observed differences, with no explanatory power. This is a matter which needs to be addressed, but B has little to say about it.

My second observation is more serious. Take ergativity. An ergative system is one in which the subject of an intransitive verb is treated grammatically like the direct object of a transitive verb, while the subject of a transitive verb is treated differently. B, of course, proposes to handle this with an “ergative case parameter”, with the settings “yes” and “no”. He acknowledges that this account is not adequate, but he does not pursue the matter. In fact, though, we now know a great deal about ergative languages, and it is clear that ergativity is as far as can be from an either/or phenomenon.

There is no known language which is wholly ergative. Instead, every ergative language exhibits ergativity in some circumstances but not in others, and the range of observed ergative systems is enormous, though the differences are not arbitrary. Historical and functional linguists have enjoyed a good deal of success in explaining how ergativity arises and why it appears in the circumstances in which we find it. But B says not one word about this success, and proposes instead a “parameter” which not only explains nothing but doesn’t come remotely close to getting any facts right. In this case, the parametric approach is plainly a large step backward from the understanding already achieved outside the parametric framework. (To be fair, B does very briefly note the problem, but he seemingly concludes only that more parameters are needed.)

A similar problem arises with the “agreement principle” (not a parameter) stated on p. 154. This rather helpless approximate summary of some observed facts about agreement systems does no more than to assert that certain systems are possible, and it entirely overlooks the progress historical linguists have made in understanding the rise and extension of agreement. B’s stipulation explains nothing, while the historical approach explains a good deal.

And that brings me now to my own specialist area of language change. To begin with, it is not easy to see how the grammars of languages can change at all in the parametric world-view. If children are exposed to a language which has fixed a certain set of parameter settings, how can they possibly acquire different settings? Yet we know and B readily acknowledges that the grammars of languages can and do undergo dramatic changes, changes which must absolutely require large changes in parameter settings. Word order can change enormously; ergativity can be gained or lost; object-incorporation can be gained or lost; and so on. How can such things possibly happen if our mental grammars are constrained by parameter settings?

In his final chapter, B essays a few speculations. He suggests, somewhat speculatively, that changes in parameter settings arise by the process that historical linguists call “shift of markedness” though B does not use the term.

Every language has marked constructions alongside its ordinary unmarked ones. For example, alongside unmarked I can’t recommend this book, English has the marked version This book I can’t recommend, and alongside unmarked The Mongols sacked Kiev it has the marked Kiev was sacked by the Mongols. B proposes that a change in a parameter setting occurs when, for some reason, speakers come to use marked forms with ever greater frequency, until the marked forms become more frequent than the originally unmarked ones, at which point the children acquiring the language find it more economical to take the originally marked form as the unmarked form, at once inducing a shift in the setting of a parameter.

Is this plausible? There is no doubt that shift of markedness is a real phenomenon, and one involved in some instances of grammatical change. But it is certainly not the only pathway of change in grammar, and in fact the standard handbooks devote much more attention to other and quite different pathways, ones not even mentioned by B. It is conceivable, of course, that B’s parametric approach might have something to say about these other mechanisms, but B is silent here.

A further difficulty with a parametric view of grammatical change is that changes must in this view be abrupt: a new generation of children must change the setting of a parameter, and the grammar of the language must change at once. Is this what happens? B assures us that it is. But he cites only a single example in support from the history of English and his sole source of information is the work of his fellow Chomskyan linguist David Lightfoot. Now, Lightfoot has been arguing for years that change in grammar really is abrupt. But his work has been savaged by historical linguists, who have demonstrated time and again that changes in grammar even the cases cited by Lightfoot are almost always slow, gradual and incremental, taking place over generations and often over centuries. On the face of it, then, the predictions of the parametric account are simply dead wrong. True, Lightfoot has located one change in the grammar of English which, from the written records, seems to have been quite abrupt, but one such case is hardly enough. And, of course, there are obvious difficulties with the notion that a whole generation of children can change a parameter setting all at once.

There is only one conceivable way of salvaging the parametric account in the face of such recalcitrant data: it must be maintained that gradualness in grammatical change is simply an illusion, and that the underlying grammars are in fact changing abruptly. B in fact attempts to defend this position, but only very briefly and inexplicitly, and only in connection with his favoured pathway of shift of markedness.

There is a further point. Historical linguists have established that a grammatical system cannot, in general, change arbitrarily from one pattern to another. Instead, change in grammar must normally proceed along an identifiable pathway and the possible pathways are now rather well understood. Just to cite one example, a language which lacks case-suffixes cannot acquire these out of thin air. Rather, it can only acquire them from identifiable sources via identifiable pathways of change. Historical linguists have already enjoyed a good deal of success in identifying these pathways, and no account of grammatical change can reasonably ignore these findings. But B is silent on this topic: he fails even to show any awareness of the importance of pathways, and he sticks to his belief in changes in parameter settings.

B’s final chapter is largely devoted to an account of why parameters should exist at all, and how they might have arisen. But his wholly speculative suggestions are unconvincing, and this chapter adds very little to the value of the book. True, he need not explain the origin of parameters if he can demonstrate their reality, but the space might better have been used to shore up the evidence for that reality. And his final pages, on the value of linguistic diversity, are excellent in themselves but of no great relevance to his argument.

To sum up, B has done us a great service in presenting the parametric view of grammar in such a clear, explicit and readable book. His presentation is often seductive, and I sometimes find myself willing his account to be right. Like most people, I prefer positive results to negative ones, and the parametric view, if it can ever be substantiated, will represent an enormous step forward in our understanding of language.

But my judgement is that B has not made a compelling case. While B is able to cite some evidence in support of his case mostly in the form of grammatical correlations which stand up tolerably well it appears that almost every one of his proposed parameters at once runs into substantial difficulties. To his credit, he recognizes at least some of these difficulties though certainly not all and he tries to address them, but he himself acknowledges that his attempts are less than successful, and he constantly admits that more work must be done before any parameter can be regarded as established. The parametric view of grammar remains a hope for the future, no more, and the obstacles are more formidable than a reader of this book may suppose.

Buy The Atoms of Language from Amazon United States of America Amazon.com  Amazon United Kingdom Amazon.co.uk  Amazon France Amazon.fr  Amazon Deutschland Amazon.de  Amazon Japan Amazon.co.jp Amazon Canada Amazon.ca

Computer-generated translation of this page French français German deutsch Spanish español Portuguese português Italian italiano Russian Russian JapaneseJapanese Chinese (Traditional) Chinese (Traditional)Arabic Arabic― also try this alternative fast translation service.

© Larry Trask. 

* Larry Trask is a leading expert on the Basque language; he works on typological, universal and historical aspects of morphology and syntax. Among his recent publications are a popular introduction to linguistics, an edited volume on time-depth in linguistics, a handbook of English usage, and a dictionary of English grammar. With eight colleagues, he has just completed a large reference grammar of Basque, to be published shortly. He has won a Leverhulme Fellowship, and he is now on research leave, compiling an etymological dictionary of Basque. He is also writing a book on the origin and evolution of language.


Trask, R. L. (2002). Review of The Atoms of Language: The Mind’s Hidden Rules of Grammar by Mark C. Baker. Human Nature Review. 2: 77-81.

US -

Amazon.com logo

UK -

Amazon.co.uk logo

The Human Nature Review