Immediate constituent analysis is based on an implicit assumption that linguistic structures—especially syntactic structures—are layered structures amenable to analysis by progressive dichotomous cutting. This paper suggests an alternative assumption, viz., that some linguistic structures are layered while others are ordered like beads on a string. Negatively this alternative assumption leads to a rejection of conventional immediate constituent analysis along with its reluctance to recognize more than two constituents on the same level of cutting. Positively, it leads to the development of a string constituent analysis in which grammatical strings are discovered and described (with mention of any observable layering characteristics), and in which constituent substitution points within the strings are likewise discovered and described. It is important to note that, while string constituent analysis is not geared to the assumption of the relevance of layering at all points in the structure of a language, it does not rule out the observation and description of layering where this is relevant.

In his article Proto-Indo-European Reality and Reconstruction, Ernst Pulgram states that ‘when we reconstruct, through the methods of comparative historical linguistics, an array of asterisked Proto-Indo-European forms, the procedure itself implies that the result of our endeavours is a uniform construct’, and asserts that ‘anything in linguistics that is timeless, non-dialectal, and non-phonetic, by definition does not represent a real language’ (422). He then bases his further discussion of the impossibility of learning about ‘Real Proto-Indo-European’ on the presumed unreal nature of the reconstructed proto-language: ‘No reputable linguist pretends that Proto-Indo-European reconstructions represent a reality’ (423). Hence, for Pulgram, the formulaic character of our reconstructions of individual items renders it impossible to make a serious attempt to reconstruct texts in a proto-language: ‘To write an Aesopian fable or the Lord's Prayer in reconstructed Proto-Indo-European is innocent enough as a pastime, but the text has no doubt little if any resemblance to any speech ever heard on earth’ (423).

It has often been regretted that, while the methods of modern linguistics have achieved notable progress in the field of phonology, corresponding results for syntax have been considerably less impressive. At least one reason for the truth underlying this regret is undoubtedly the inability of the linguist to satisfy the very requirements which he himself has thought it necessary to impose on an adequate syntactic description. His persistent failure has resulted not so much, I think, from the lack of knowledge or insight into the nature of linguistic systems but rather in large measure from the unrealistic nature of these requirements.

Chomsky's rules do not at present generate intonation patterns as elements in terminal strings, nor do they show the optional and obligatory choices of intonation patterns that are available with any given sequence of formatives. The following modifications of his rules are suggested as a way of incorporating intonation patterns.

[Vowel nasality is a distinctive feature in American English in cases such as camp, hint, and bunk, as compared, respectively, with cap, hit, and buck. This is demonstrated by the results of a number of experiments with synthetic speech and magnetic tape, by data obtained from spectrograms and kymograms, and by the descriptive analysis of pertinent utterances by American speakers. A general rule for the distribution of this phenomenon is given in the conclusion.]

A discussion of color categories as they function in contexts which are unfamiliar to the reader introduces frustrations and misunderstandings, particularly when the reader's concerns are practical as well as academic and when they involve cross-cultural communication and interaction, for example during anthropological, psychological, or medical research. The reader who questions this observation may enjoy reading the following passage a second time, after he leaves the final word of our article.

It has generally been assumed that a linguist would never encounter in the course of his linguistic practice a case of two different phonemes with mutually exclusive distributions and with allophones phonetically similar. This, however, is the case in the phonemic system of Toba-Batak, a Malayo-Polynesian language of northern Sumatra. In that language the two phonemes /o/ and /ɔ/ have phonetically similar allophones, for they are adjacent vowels on the vowel chart; and the two are so distributed that no two whole utterances are distinguished solely by the alternation /o/ ∼ /ɔ/. There are, to be sure, two minimal pairs on the word level: /jolo/ ‘front’ ∼ /jɔlɔ/ polite particle, and /do/ ‘do (in the tonic sol-fa)’ ∼ /dɔ/ syntactic particle. But these minimal pairs are enough to demonstrate that the two phonemes contrast only if we take the lenient approach to the problem of phonemic analysis exemplified by Zellig Harris's 1941 statement: ‘In identical environment (distribution) two sounds are assigned to two phonemes if their difference distinguishes one morpheme from another.’ That is to say, if we are willing to assume that phonemic analysis may comfortably be based on a previous morphemic analysis, our Toba-Batak problem is solved without difficulty.

It is odd that linguists, who pride themselves on the rigor and scientific nature of many of their concepts, should nevertheless be so tolerant of vague, unverified statements in some parts of their field. To take an example, Bloch has made many contributions to linguistic theory in a long series of excellent publications, but he does not appear to have adopted any scientific procedure to check the validity of the phonetic statements (e.g. those about tongue positions during vowel sounds) which often occur in these works. It seems as if he regards the justification of phonetic statements as outside his field as a linguist. Thus he says, ‘Qualities are identified by ear; but in linguistic works they are traditionally defined in terms of their assumed production by the vocal organs.’ This uncritical acceptance of the traditional descriptive techniques is, as we shall show, inadvisable. Disclaimers such as ‘articulatory terminology is used only because a usable auditory terminology has not yet been developed’ (loc.cit.) are not sufficient excuses for making seemingly dogmatic assertions about physiological facts.