It was the larynx, guv

This book is a whodunnit about the origin of human language, and the culprit is not our brain but our larynx. The lowering of the larynx in Homo sapiens enabled us to utter a greater range of consonants and vowels than our lesser relatives - apes and Homo neanderthalensis . From this richer range came syllables, and the structure of syllables is similar to that of sentences. Sentences are a consequence of syllables, themselves a consequence of a lower larynx, and sentences are the hallmark of human language ... so there you have it.

Andrew Carstairs-McCarthy's argument is convoluted and requires considerable attention and linguistic knowledge to follow. Linguistic terms are not explained and there is no glossary; at the beginning, a footnote leaves one uncertain as to what the author means by NP (noun phrase): "What I call 'NPs' are strictly 'determiner phrases' in the sense of Abney (1987); but, in common with many linguists, I will continue to use 'noun phrase' in its traditional sense." A bit about NPs in the traditional sense (which tradition?) would have been welcome; so would a hint of what Abney "strictly" meant by "determiner phrase". As the reference is to a Massachusetts Institute of Technology PhD thesis, and Abney does not appear from the bibliography to have written anything since 1987, readers are clearly not encouraged to find out. This is a bit worrying, because the author sees the hallmark of human language in the dichotomy between noun phrase and sentence.

It so happens that I spent some of my life studying human languages that do not distinguish between noun phrase and sentence, between verb and noun. In Tolomako, for instance, an Austronesian language of Espiritu Santo, Vanuatu, na gapuku is at once a complete sentence meaning "I have fire" and a noun phrase meaning "my fire" as in na gapuku mo nate , "my fire has gone out" ( na is the article, gapu "fire," ku "my", mo nate "has died"). Put that in the subjunctive and you have te gapuku which alone, as a sentence, means "Give us a light, will you?" But as a noun phrase, as part of a sentence, it means "my (hypothetical) fire". Yes, in those languages, not only can noun phrases be used as sentences, but nouns are inflected for mode, indicative ( na ) or subjunctive ( te ), just like verbs (except that Austronesianists call those modes realis and irrealis ). To top it all off, nouns can take the same inflections as verbs, in which case they mean "to turn into" ( mo vatu "it turned into stone" versus na vatu "stone"). Verbs may take the same inflections as nouns, too. The fact is so well known to Polynesianists (Polynesian is a sub-family of Austronesian) that they coined a term to cover both verb and noun: "base". Carstairs-McCarthy, who lectures at the University of Canterbury in New Zealand and briefly mentions Maori and Polynesian languages, should know the above.

The syllable structures he invokes are also illusory. Many languages allow for only the simplest syllable structure: a vowel, preceded or not by one consonant (eg Tolomako). In others, only a wild stretch of imagination allows one to distinguish syllables; a word typically consists of a string of consonants with one vowel at or near the end (eg Sakao, Vanuatu: tmlkkleprn ), optionally broken by epenthetic vowels in loose regressive vowel harmony to facilitate pronunciation, varying in their positions and in their quality from speaker to speaker. Yet these languages are perfectly adequate for communication. My informants often asked me about rockets and radio waves and their subsequent questions showed they understood. As Austronesian languages stretch from Madagascar to Easter Island and the Philippines, via Indonesia, Malaysia and Taiwan, ignoring them is quite a feat. And as the author also manages to refer to syllables in Mandarin Chinese without mentioning tones, one may be forgiven for thinking that the whole tale is not told.

He is helped here by his espousal of "mainstream" linguistic theory (read: Chomsky's). Mainstream linguistic theory is a closed system. Whenever a sentence predicted by the theory does not occur in a corpus of texts or elicits a "you must be mad" look from native informants, the accessory theory of competence versus performance is brought in: the native was competent enough to produce or understand the sentence, but his performance was wanting. This leads to a fine disregard for data. Thus, this French example is totally unacceptable to native speakers: " Que se passe-t-il? - Le chat qui est tombé par la fenêtre ." (Myself a native French speaker, I happened to have two informants at hand: my wife, French, bilingual and her mother, French, monolingual.) Many readers may be too exhausted after wading through Carstairs-McCarthy's complex, highly technical and jargon-laden arguments to retain enough critical sense to counterargue that if syllable structure mimics sentence structure it is because sentences came first, sentence structure reflecting thought structure; or that the lowering of the larynx, supposed to have brought the phonetic range necessary for language, is humbug. Our much pooh-poohed cousins Homo neanderthalensis could have made do with just two consonants (a cough and a retch for instance), two vowels (an aarrrgh and a hey), and eight tones, which, assuming the simplest syllable structure, left them way ahead, with 48 different syllables, when Rotokas manages only 35 with its six consonants, five vowels and no tones (Rotokas, a Papuan language of New Guinea, is famous for having the fewest phonemes of all).

This book demonstrates what ails the linguistics that calls itself mainstream. Parochialism: not one mention of Igor Mel'cuk or Yorick Wilks, but references to dozens of obscure authors. The pre-eminence of theory over data: none of the English examples is drawn from a corpus; the lone French example is unacceptable to native speakers. Ad hoc exclusion: ignoring tone even in Chinese; ignoring the non-distinction in many languages between verb and noun, sentence and phrase.

Until linguistics departments make fluency in at least one unfamiliar language a prerequisite for exposure to various theories, linguistics will remain in the sorry state of astronomy when it was astrology, of meteorology when it was rain dancing. Those departments that also insist on the learning of at least one signed language may perhaps nurture, at long last, the Isaac Newton of linguistic science .

Jacques M. Guy, a computer scientist, spent 15 years at the Australian National University working on Austronesian languages and their taxonomy.