In addition to the familiar outline organization of discourse in terms of coördination and subordination, some languages of Brazil and New Guinea (including Pidgin) have an overlay organization. It involves the near repetition of substantial stretches of speech in such a way that the repeated elements are placed in prominence. These languages also use repetition in ways that are distinguishable from overlay organization. The nature of overlays puts into a new context the question of reference in linguistic theory.

The attraction of the negative to subject any is an invariant rule for all dialects of English, and appears to respond to the distributive and hypothetical features of this indeterminate. Negative concord, on the other hand, is a variable rule which distributes the negative rightward in response to affective factors. It is progressively extended from one dialect to another in an implicational series which initiates negation in new positions as the rule approaches obligatory status in the original environments.

The recently discovered manuscript page of the Codex Argenteus, containing Mark 16.12–18, is here published in facsimile and transcription. The text is examined for the new linguistic data contained in it, and a detailed discussion is offered for the most notable, if not the only, new lexical item: farwa- ‘shape, appearance’.

The heuristic principle of empirical discovery in semantics, as in all other understanding, is here derived from the particular case of words apprehended as mysterious in their environment; it is the rule that the best guess is that one which maximizes the redundancy of word and environment together. But the environment cannot be exclusively delimited; it is partly wordy and partly wordless, partly factual and partly fictionally invented, partly past and partly future, and so on without limit. Thus correspondence meaning (Hill 1970) can be separated out from ‘other’ meaning(s) only conceptually and then only for the moment; as soon as one tries to employ the distinction, it vanishes.

Process models of language, and generative grammars, receive support from typological theory. Typological study has found that some syntactic constructions can be correlated with characteristic sentence patterns. Thus consistent OV languages have relative clauses, possessives, and adjectives preceding nouns; consistent VO languages show the converse order. The parallelism between the three constructions can be accounted for in process models, and related to the position of the object with regard to the verb in both OV and VO languages. The central importance of the verb in language as indicated in these and other constructions is pointed up by recent neurological investigations supporting descriptive and generative theory which proposes that verbs have the primary role in sentences. These observations illuminate syntactic patterns in New High German, in Latin and the Romance languages; they also indicate that a framework is now available for historical and descriptive syntactic study.

This paper demonstrates that the rules of the transformational stress cycle can be rewritten and simplified by substituting binary suprasegmental features for degrees of stress. The new prosodic features proposed for English are: (1) heavy, (2) accent, (3) intonation, (4) cadence, (5) endglide, and (6) emphasis. This binary suprasegmental feature system accounts for the grammar-expounding prosodic distinctions of English in an intuitively satisfying way; it factors out accentual and intonational phenomena conflated in ‘stress’; and it incorporates linguistically significant generalizations about the syntactic conditioning of prosodic patterns. In addition, it explains rhythmic and stylistic variations disregarded by previous stress rules; and it allows revision of the stress cycle in ways which enhance its economy and explanatory power while retaining its insights and concinnities.

This article is an attempt to apply distinctive feature theory to the study of diphthongization. It shows that several types of phonological change—some of them traditionally called diphthongization, others not—involve identical processes of change, and differ only in terms of the distinctive features involved. It identifies the single principle of intra-segmental variation which underlies all such changes, and thus defines the place of diphthongization in a general typology of phonological change.

The syllable is an important phonological unit that must be formally defined within generative phonology. Evidence from Spanish, German, and other languages shows that a syllable boundary is necessary to designate the domain of certain phonological rules. The syllable can be formally and universally defined by a rule which inserts such syllable boundaries between certain sequences of segments. This rule must apply persistently throughout the phonological component, changing the position of the syllable boundaries as the sequences of segments change.

Lakoff 1970 offers seven sets of data which he believes furnish support for revising current models of syntactic description in such a way as to allow global rules, i.e. rules that make reference to more than a single point in a derivation. A close examination of these data reveals that only two sets cannot be given an adequate treatment within the framework sketched in Chomsky 1965; and even these two sets do not warrant the wide-ranging extension of current syntactic theory which L proposes. Furthermore, L's seven arguments, even if they were accepted, would furnish no support for his assertions concerning the nature of semantic representation.*

In line with recent claims, it is argued here that no rules in a transformational grammar should be extrinsically ordered. This claim is examined with respect to four well-known syntactic rules. Arguments which purport to establish that these rules must be extrinsically ordered are shown to be false, and solutions consistent with unordered rules are proposed as alternative explanations to the ordering hypothesis.

Otto Jespersen 1917 pointed out that the normal meaning of the negation of a quantitative term Q is ‘less than Q’. This fact is shown to stem from the possibility of interpreting Q as ‘at least Q’ in some non-negative environments. The analysis is extended beyond quantifiers per se to other terms involved in inclusion relations, and it is shown that such relations play a role in determining the acceptability of various grammatical constructions.

A numerical model is developed in order to establish the extent to which the principle of maximal perceptual contrast can be used in phonological theory to explain the phonetic structure of vowel systems. Preliminary results obtained with the model indicate that perceptual contrast appears to play an important role as a determinant of such systems. Therefore, it is likely that this principle (along with other factors) should be included among the variables in an explanatory phonological theory. However, the incorporation of numerically stated conditions on phonological structure appears to presuppose a formalism different from that which has developed within current descriptions of phonology. Some refinements and extensions of the present framework are suggested. It is proposed that predictions of phonological facts be derived as consequences of the structure of the mechanisms available for human speech communication and the optimization of their use. Such an extension would constitute a theory that would be different from traditional ‘Saussurean’ linguistics in several respects; e.g., it would be quantitative, and deliberately substance-based. The research reported represents a preliminary attempt to apply such a program.

In a recent paper, Bresnan claims to have shown that the Nuclear Stress Rule applies within the syntactic cycle, rather than to surface structure. Given this result, she claims to have found evidence for the existence of deep structure, for the lexicalist hypothesis, and against McCawley's 1970 analysis of English as a VSO language. B's claims are analysed here in detail. It is shown that her claims are without a basis in fact: the Nuclear Stress Rule is a global rule which applies to surface structure, but it has a global environment referring to other levels.*

Baker & Brame, in this issue of Language, correctly observe that, in three of the cases cited in my paper on global rules (Lakoff 1970), the effect of global rules can be obtained by the use of ad hoc coding mechanisms using arbitrary grammatical elements, in one case an infinite number of such elements. This raises the question of whether the elements used in grammatical descriptions should be arbitrary or whether they should have a natural basis. In addition, Baker & Brame incorrectly claim in three other cases to be able to handle global phenomena within the Aspects theory. In each case they use apparatus that goes beyond that theory. Moreover, they claim that a theory of global grammar is necessarily ‘more powerful’ than the Aspects theory and their proposed extensions. This is shown to be false.*

This paper discusses and explains some irregular Vedic imperatives in -dhi built with full-grade stem formants. Some archaic features of the aorist system are described in §§3–4.

A certain class of vowel changes caused by consonants in Germanic languages, foremost among them the OHG monophthongization (ai→ē before h, r, w, #; au→ō before h, r, dentals, #), have never been given a systematic explanatory formulation. I show in this paper that all these changes are height assimilation rules. This explanation requires the introduction of a new convention into the theory of phonology: assimilatory features in a rule environment must be interpreted as relative to the corresponding features in each of the assimilable segments. It follows that an assimilation rule may define different natural environment classes for each assimilable segment. E.g., h, r, w and h, r, dentals form natural lowering classes for front and back vowels, respectively, because h, r are low relative to all high vowels (cf. Go. i→ε, u→ɔ before h, , r), while w is low relative only to front high vowels, and dentals are low relative only to back high vowels. The assimilatory nature of these changes and the relativity principle are further supported with evidence from other Germanic languages and from Spanish, Quechua, and Diegueño. The following theoretical conclusions are drawn from this study: (a) At least two height features have to be distinguished in consonants, one for the front vowel region, the other for the back vowel region. (b) Certain detail information must be specified before binary choices are made. (c) Certain redundancy rules must be formulated in such a way that they apply in a definite order (e.g., height in dentals depends on their degree of anteriority). (d) Generally, more articulatory detail has to be accommodated by phonological theory.

In the northern Greek dialect of Samothraki, historical /r/ is reflected by zero intervocalically, by loss with compensatory lengthening between word boundary or consonant and a following vowel, by [ī] initially before consonant, by [i] between vowel and consonant, and by [r] finally. There is evidence that epenthesis of [i] occurred between /r/ and a following consonant, that /r/ was then lost with compensatory lengthening of the following vowel by a common rule, and that this long vowel was then reduced adjacent to another vowel by an equally natural rule. These simple assumptions account for the known facts.

The question considered here is the correctness of the assumption that transformational rules are extrinsically ordered. One type of argument that has frequently been given in support of this assumption is shown to be invalid: it shows that if the rules in question are to be extrinsically ordered, they must be ordered in some particular way, but it fails to show that the facts cannot be accounted for without imposing extrinsic order on these rules. Thus the assumption that rules are not extrinsically ordered is found to be viable.1

The hypothesis that the Nuclear Stress Rule (NSR) of English is the last rule within the syntactic cycle, as recently advanced in Bresnan 1971, is examined and rejected. It is shown to be observationally inadequate: it makes incorrect predictions and is unable, in principle, to capture generalizations about English prosodic stress. Evidence is presented showing that, to the extent that sentential stress depends on structure, it is surface structure that is relevant. The feasibility of assigning stress by means of the NSR, at any level, is seriously questioned, and various facts are presented which indicate that prosodic stress assignment cannot be a function of structure alone.*

Various proposals for treating the word even in a transformational grammar of English are reviewed. It is shown that the word cannot be derived from a structure in which it is directly associated with that portion of the sentence which constitutes its semantic scope. An alternative account based on principles of interpretation of derived structure is sketched and motivated. The semantic content of the word is suggested as the source of limitations on its appearance within the sentence.