In his book, Carstairs-McCarthy (1999, 2000) attempts to ground language evolution in syllabic structure, by emphasising the similarities between syllabic and syntactic structure. Syntactic structure universally instantiates a difference between reference and truth, which has been much examined in philosophical circles. Carstairs-McCarthy demonstrates that this universal difference is not necessary for a workable language, raising the question of why it should have come to be the linguistic universal that it is, encoded into the syntax of all languages. His answer to this question is that the distinction reflects the roots of syntax in syllabic structure. The argument depends on making a strong analogy between syllabic and syntactic structures. The problem is that the simplified syllabic and syntactical structures compared may both have roots in more general cognitive information-management principles, weakening the key claim of a direct line of descent from syllables to syntax. Indeed, Carstairs-McCarthy provides arguments in favour of this cognitive interpretation. Although readers may not, therefore, be wholly convinced by the main argument of this thought-provoking book, they will certainly come away educated on a wide range of issues relevant to language structure and evolution.
2. The structure that he wishes to champion for this role is the simple tree-structure of a maximal syllable. This is a structure formed minimally of a single phonological element (the nucleus), which optionally (but often) has two other phonological elements attached to it. For example, the non-word 'pid' is composed of three phonological elements that map onto its three letters, with the vowel being the nucleus. Three elements combined in a tree structure allow for three possible tree structures: a two-level binary tree with a pair to the right or left of a single item, and a ternary tree. Carstairs-McCarthy reviews arguments for all three possible structures as underlying syllables. However, he claims that the disagreements hinge on how to weight internal asymmetries that are widely agreed upon, greatly limiting the extent of the disagreement. One agreed-upon asymmetry exists between the nucleus and the other two elements, with the nucleus being the only non-optional element. A second set of asymmetries arise between the element to the right of the nucleus (called the onset) and that to the left (the coda). In all languages, the onset appears preferentially (but not necessarily) to the coda when there is only one element attached to the nucleus. When it exists, the coda tends to have a closer association with the nucleus than the onset, attaching itself to form a single unit (the rhyme). In sum, evidence favours the binary tree structure [(onset) [nucleus ((coda))]], where the number of parentheses increases with the degree of optionality, and square brackets indicate the tree structure.
3. The main argument of The Origins of Complex Language is an attempt to motivate a strong analogy between this (more or less) universal syllabic structure, and the structure of sentences. Simplifying the matter rather drastically (as Carstairs-McCarthy acknowledges), most human languages have a basic sentence structure which may be represented in the above notation as [(NP) [VP ((NP))]]. When there is only one element, it is a verb phrase (VP), consisting either of a single verb ('Eat!') or a set of words which together act like a verb ('Eat quickly!'). The VP is the nucleus. When there is a second element (usually a noun or a noun phrase [NP] which acts as a noun), there is a strong tendency across languages for it to be attached to the left of the verb, as in '[(The hungry dogs) eat.]'. The analogy is completed because a second NP - the object or complement- usually attaches to the right of the obligatory VP. It modifies that VP in such a way as to form a coherent component with it: [(The hungry dogs) [eat ((the family cat)).]]. Just as with the syllabic structure described above, there are a multitude of exceptions and complications to all these regularities, many of which Carstairs-McCarthy argues may be ignored for one reason or another.
4. The binary tree-structure of a basic sentence instantiates a distinction, which Carstairs-McCarthy argues is key to understanding human language, between a noun-phrase and a sentence. On this distinction hinges the matter of truth value. Truth may be ascribed to a sentence, but not to a noun-phrase. This assertion may seem trivial. However, trying to make sense of the complications of the relation between truth and reference has been a central focus of analytic philosophers of language for many decades, and it has turned out to be anything but trivial. Carstairs-McCarthy reviews the philosophical issues, prior to focusing on a question that has not received much attention: why should this distinction between a noun phrase and a sentence be encoded so strongly (and so universally) in linguistic grammar?.
5. It is not sufficient, as he explains at some length in an early chapter of his book, to simply give the answer that it is encoded in grammar because it is an important distinction. One problem with such a dismissive approach to the problem is that, from a comparative communication perspective, it is not clear that the distinction really is necessarily important. It is, at any rate, not important enough to appear in any other animal's communication system. Non-human animals' (and very young humans') communications cannot be easily classified as either noun phrases or sentences (does a vervet alarm call mean 'Eagle!' or 'There is an eagle overhead!' or 'Run!'?), and yet for all that they have clear utility. How did it come about that human beings found insufficient the utility that suffices for all other animals? What lead us to separate referential communications from truth statements?
6. Such questions are difficult to answer. There is no consistent explanation of the nominal/sentential distinction on semantic, pragmatic, or cognitive principles (p. 161). Moreover, it is quite possible to imagine a structured language -and Carstairs-McCarthy works out the exercise for his readers in Chapter 2- that maintains the distinction without encoding it on a grammatical level. "It follows that, if human language was descended from something like a primate call system, there is no stage at which the sentence/NP distinction would have had to emerge in order for it to express everything that modern human language can." (p. 150). Why doesn't any 'monocategoric' language exist? Why do all human languages encode, in their very structure, a distinction that no other animal makes at all, and which need not be encoded structurally anyway?.
7. The answer developed in 'The Origin Of Complex Language' is that the nominal/sentential distinction arose from the syllabic distinctions whose structure it mirrors. In rather balder terms than those presented by the author, the claim is that the syllabic distinction served as a template for the development of syntax. Such arguments from analogy, being necessarily post-hoc, are tricky to make convincingly. They depend on the identification of plausible isomorphisms of structure as well as the outlining of plausible routes by which that isomorphism might have been developed.
8. Carstairs-McCarthy has some trouble with making the isomorphism convincing. The kinds of structures that he claims to be analogous to one another are simple and ubiquitous. It is simpler to add a single unit to an already existing, possibly complex, unit than it is to bring a greater number of units into coordination at once. As a result, many composite structures and compositional processes in the natural world admit of descriptions in terms of the binary trees. Arguments of a privileged status for syllabic and sentential structures usually depend on the time-honoured rhetorical trick of using selective examples. If one wishes to deride the many (albeit contentious) claims of an analogy between nonlinguistic functional composition (such as occurs during tool development or when producing complex motor processes) and grammatical tree structure, it is easy to find examples in which one constituent is "not of a different kind in the sense in which a verb phrase is a different kind of syntactic unit from a sentence" (p. 186). It is equally easy to find examples which support those same claims. How is one to decide which examples are better? Linguists are fond of insisting that the distinctions they draw are of a different kind than those drawn by others. It sometimes appears to the rest of us that they alone are privileged to such clear-cut determinations of natural kinds, and that they alone have the ability to derive universal classifications from consideration of a few examples of their own choosing. Those of who missed the class in which these enviable skills were imparted are apt in our ignorance to find such claims unconvincing.
9. The claims of isomorphism of structure between syllables and phrases are further compromised by the fact, already alluded to above, that the identified isomorphism in this case depends upon a rather severe over-simplification, especially on the syntactical side. Carstairs- McCarthy is careful, however, to acknowledge this, and, in many cases, to try to justify his over-simplifications on theoretical, quantitative, and/or empirical grounds.
10. The bulk of his argument attempts to develop the more richly- constrained second line, by outlining a plausible route by which an isomorphism between syllabic and syntactic structure might have evolved. He is a suitable and thoughtful guide to the theory he is putting forth, providing throughout his book a suitable number of caveats and warnings of the uncertainty of evolutionary speculation. Nothing relating to the evolution of language can be certain or complete. The best one can hope is to suggest a location for one small piece of a jigsaw puzzle that is clearly too large for anyone to view in its entirety.
11. Working with this modest goal and within the levels of uncertainty he readily acknowledges, Carstairs-McCarthy describes an engaging, if sometimes frustrating, hypothesis, which attempts to justify the syllabic/syntactic isomorphism by explaining how it ties together four simple phenomena which otherwise seem disconnected. The first has already been mentioned: the mystery of why we should have universal grammatical encoding of a distinction made by no other communication system. The second phenomenon is the universal (cross-species) avoidance of synonymy. Young unilingual human infants and other primates undergoing language-training or its analogue have been shown to have an unlearned default assumption that a new symbol must have a different meaning than an old one (multi-lingual infant language learners present an interesting counter-example which Carstairs- McCarthy does not discuss). The third phenomenon is the lowering of the human larynx, a physiological change which has been much discussed by those interested in language evolution. Finally, Carstairs-McCarthy ties in vocabulary size. Human beings have enormous memory storage capacity in their lexicon. Many of us cannot remember the name of a person we met at dinner last night, but can nevertheless draw on a vocabulary of a couple of hundred thousand words in expressing our displeasure at this fact. What is special about words, that we are able remember them so well?.
12. In caricature form the bulk of argument goes like this: The lowering of the larynx enabled our ancestors, who already had a repertoire of calls, to produce many new sounds. This ability gave our ancestors 'unused' sounds that could be taken over for new purposes, allowing previously inexpressible meanings to be expressed. In order to minimise the memory load inherent in adding many new calls to vocabulary, they re-used calls that already existed in their repertoire, put together by concatenation. Thus "A way has been found whereby a large proportion of the potential call repertoire can be put to use, in accordance with synonymy-avoidance pressures, without imposing an intolerable burden on the memory or an impracticable task on the learner" (p. 133). This opened the way for dual-patterning, the characteristic of human language whereby the components of signals can be analysed at two levels, one (word-level) meaningful, and one (subword-level) meaningless.
13. A vital core of the argument is to explain why these new 'words' should be organised along the same lines as their own structure, so that syntax should come to mirror phonology. Alas, this vital core is the weakest section of the book. The nature of the analogy between syntactical and phonological structure is elaborated in detail, and great pains are taken to show that the analogy need not necessarily have been so exact as it is. However, when it comes to explaining how the analogy came to be so exact, Carstairs-McCarthy has surprisingly little to say, and what he does say seems to undermine his own argument. The vital question is posed on p. 163: "is there any reason why 'kick' rather than 'boy' or 'ball' should gravitate towards the nucleus-like position?". By this point in the book, the reader hopes and expects to be introduced to an explanation that is crucially dependent on the way syllables are structured. However, Carstairs- McCarthy's response to the question he has posed is "The answer is yes. With the action 'kick' in a nucleus-like position, expressed syntactically like a verb, there is room to accommodate both its arguments 'boy' and 'ball' in marginal positions in the same clause." And why should this be desirable? Because "On the other hand, if one of the arguments (let's say, 'boy') is placed in a nuclear position, then fitting 'kick' and 'ball' around it involves a greater deviation from the syllabic model". One cannot justify an explanatory analogy by pointing out that any alternative explanation would weaken the analogy!.
14. There is more to the justification of syntactical structure than this. However, none of it seems to depend crucially on any continuity between syllabic and syntactic structure. Instead, it depends on biases in human cognition ("human beings represent events mentally in terms of stable objects and transient actions or states" [p. 164]) and some formal conditions for their identifiability (i.e. "material that occupies margin-like positions should be relatively homogeneous and relatively distinct from what occupies the nucleus-like position" [p. 164]) and management (i.e. single clause syntactic representations are preferable to multi-clause representations because the argument management is less messy [p. 165]).
15. These are interesting, intelligent, and quite plausible bases for the identified isomorphism. However, their exposition does little to advance the book's main thesis of a strong evolutionary continuity between syllabic and syntactic structure. Rather, it seems to pick up on some more general principles of information management that might underlie both structures. We need organising principles in order to be able to identify data structures. One good way to organise data structures is to have a core feature with the roles of that feature's modifiers marked by position, and organised in a binary fashion which makes it easiest to augment the data structure by minimising the required organisational overhead. If we accept these axioms, we should expect to see formal isomorphisms between various aspects of language, as well as in other areas of human cognition. Notwithstanding Carstairs-McCarthy's attempts to downplay the isomorphism in other spheres, I believe that such isomorphisms are indeed evident in many areas of human cognitive activity, from early infant play to the design of structured computer programs.
16. Books like 'The Origin Of Complex Language' require a boldness that is easily discredited. No human being can possibly fully master the complex multitude of issues that Carstairs-McCarthy addresses. Evolutionists will take issue with his casual use of 'pressure' as an explanatory principle, his bias towards single-dimensioned adaptivity, and his glib explanation of how novel advantages can spread through a population even if they have no initial adaptive value. Syntacticians will point out that he has greatly over-simplified the syntax of human languages. Physiologists will wish to emphasise that other explanations for larynx-lowering have been offered than those he considers. Philosophers will point to discussions of truth and reference that reject the analytical axioms. Neuropsychologists will point out that his review of grammatical deficits in aphasia is highly selective. Cognitive psychologists will probably wonder why he didn't cut out all that stuff about language, and just write the good stuff about the formal structure of cognition. Nonetheless, careful readers from any of these disciplines will come away from his book better-educated. There is a great deal in this work to think about, and many good points are raised that make one think about language and language evolution from a novel perspective. If I find the core argument unconvincing, I do not for that reason wish to dismiss this book as an unworthy work. It educated me on many issues I had not considered, and helped me to think coherently about them. If our ultimate goal is out of reach- if we will never be able to correlate all the information bearing on language evolution, and arrive at the single truth we can all comprehend and accept as authentic- then we can ask nothing more than that we be able to think coherently about the nature of this fascinating problem. Carstairs-McCarthy helps us to do that, and earns our gratitude for having done so. Books like 'The Origin Of Complex Language' require a boldness that we should not discredit, but do all we can to encourage.
Carstairs-McCarthy, A. (1999) The Origins of Complex Language: Oxford: Oxford University Press.
Carstairs-McCarthy, A. (2000) The origins of complex language. PSYCOLOQUY 11(082) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/2000.volume.11/ psyc.00.11.082.language-origins.1.carstairs-mccarthy http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?11.082