Medieval Spanish had a complex system of strident obstruents. Slightly divergent paths of evolution of these segments have resulted in the major phonological differences between modern Castilian and Latin American dialects. These widely studied sound changes are here re-interpreted as changes in the sets of phonological rules of Castilian and American dialects. The role of the theories of markedness and linking, especially with respect to the features [distributed] and [strident], is discussed in some detail, and seems to provide insight into puzzling problems. In passing, it is seen that, contrary to popular belief, there has been no merger of [θ] and [s] in Latin American dialects as [s]; rather, Castilian diverged from other dialects in replacing [s] with [θ].

The use of phonons in stratificational phonology is compared with the use of marked vs. unmarked features by transformationalists. In connection with a Bulgarian vowel alternation, the simplicity of the total description is taken as a basis for preferring one possible assignment of markedness over another within the stratificational view. The assignment of markedness preferred here runs counter to that taken as universal in recent transformational discussion. At the same time, it allows an analysis which agrees with some basic intuitions of classical phonemics. Markedness, it is concluded, is not necessarily universal; and its implications for simplicity must also be considered.

Dyirbal and Gumbaiŋgar are alike in that (1) possessive phrases and relative clauses are marked by the same or similar affixes, probably cognate between the two languages; and (2) the syntax of possessive phrases is similar. Relative clause constructions are on the surface less similar: in Dyirbal the relative clause marker is always attached to the verb, but in Gumbaiŋgar it is attached sometimes to the verb and sometimes to a noun. It is shown that in each language possessive phrases can be dealt with as a special instance of the relative clause construction; and that a common set of rules can be set up for the two languages, but with different conditions and slightly different ordering. An ‘affix transfer’ rule is needed to generate all possessive phrases (but is not required for relative clauses) in Dyirbal; in Gumbaiŋgar this rule is needed for all possessive phrases and for some relative clauses. The syntactic similarity of the two languages may be evidence of genetic relationship.

Fallacious reasoning has led transformationalists to conclude that natural language cannot be produced by a finite state device. An alternate argument is proposed, based on a distinction between two types of generative mechanisms: iteration and recursion to a depth of one. A device which employs these mechanisms is a finite state device; and this paper contends that such a device is adequate to describe the data of natural language, including English embedded relative clauses. The proposed solution also handles multiple-branching constructions such as are found in coordination, and describes the intonation pattern of sentences containing strings of relative clauses.

Three different reflexes of Proto-Slavic *g and the reflex of *di̭ changed to fricatives in early Slavic. A hitherto unnoticed sequential constraint in Common Slavic, along with the geographical limitations of these changes, permits us to date them relative to the loss of syllable-final obstruents in CS1., to the fall of the jers, and to one another. Typological analysis shows that the source of the phonetic changes was a pronunciation rule implying phonemically tense vs. lax obstruents in CS1. The relative chronology established is significant for several problems of chronology within Slavic and for its relation to Baltic; and the set of concentric isoglosses defined throws light on the pattern of expansion of early Slavic.

In Part I (§§1-2) of this paper, previously published, it was argued that a ‘mirror image’ convention must be incorporated into linguistic metatheory in order to account for various syntactic phenomena in the most insightful way. This convention may be interpreted as a notational device serving to collapse rules that are, in a special sense, mirror images of one another. Here, in Part II, it is shown that there is also lexical and phonological motivation for such a convention.

Vowel harmony in Sahaptian languages is an automatic assimilation of phonemes of a ‘recessive’ set to those of a ‘dominant’ one in the same sequence. It is describable with phonetic features of phonological segments, as opposed to class markings on whole morphemes. In fact, in the Sahaptin dialects, explaining palatalization of front velars requires setting up a preceding phonetic rule of vowel harmony, and different late rules of vowel merging. Historically, we can view the vowel harmony as a phonetic process arising in the dialects by parallel (but not identical) restructuring from a Proto-Sahaptian system of morphological ablaut.

This paper describes an investigation regarding the degree of awareness of a number of different morpheme structure conditions in Turkish displayed by native speakers of that language. It appears that not all the morpheme structure conditions that can be formulated on the basis of the regularities observable in Turkish lexical items are internalized by native speakers; the test results indicate, rather, that with respect to their awareness of certain morpheme structure conditions, native speakers fall into two main groups, neither one of which can be said to know the relevant conditions in precisely their ‘correct’ form. These results support the view that speakers may well be only very imperfectly aware of regularities that do not fall within the system of fully productive rules of their language.

This study reports an experiment concerning the spontaneous responses of young children to commands differing in structural format and semantic content. The results indicate that syntactic comprehension exceeds production in ‘telegraphic’ speakers. Based on these results, conjectures are offered about the techniques which a child might use in coping with his linguistic environment.

The thesis of this paper is that a ‘mirror image’ convention should be incorporated into linguistic metatheory. This convention may be interpreted as a notational device; it serves to collapse rules that are significantly similar in that they are mirror images of one another, in a sense that can be made fairly precise. Part I gives a preliminary formulation of the convention and discusses syntactic evidence; Part II, to be published in Lg. 45:4, will deal with lexical and phonological evidence.

A general framework for the study of the acquisition of grammar is proposed as a means toward a decision between competing theories, e.g. innate versus learned. The different functions of grammatical acquisition are distinguished as modules such that each represents a separate task for a comprehensive theory of the child's acquisition of grammar. The notion of an ‘analysis device’ is contrasted with that of a ‘behavior-contingent’ device. Some aspects of the operation of behavior-contingent devices are subjected to simulated experiment; the results suggest that response-learning theories and behavior-contingent models cannot be ruled out a priori and may, in fact, be developed into powerful tools in the study of the acquisition of grammar.

This paper examines ideas basic to the Praguean concepts of markedness and neutralization in the light of problems which arise in contemporary phonological theory. The partial reformulation of phonological theory suggested here goes beyond proposals by previous authors in that it provides a plausible explanation for a number of universais concerning stem-initial consonant clusters. The present proposal includes the claim that there is a universal, ordered set of redundancy rules—in the form of neutralization rules—which place constraints on the distribution of marked features in the lexicon. It is suggested that at least some of these universals are due to limitations of the peripheral speech production and perception systems.

This paper presents evidence that semantic notions—such as presupposition, speaker’s and hearer’s beliefs about the world, and previous discourse—must be taken into account in a complete treatment of the distribution of some and any in conditional, negative, and interrogative sentences. Syntactic conditions alone will not account for the fact that, in certain sentence types, the two forms occur with different meanings.