Green 1976 shows that restrictions on main clause phenomena occurring in subordinate clauses are not capable of strict formalization, but respond to a variety of ‘syntactic, semantic, and pragmatic’ factors. This assessment is correct—more so than she gives it credit for: if we take more such factors into account and add elements like thematic structure and prosody, a number of the restrictions she has retained give way.

Historically, the analysis of the English auxiliary given by Chomsky 1957 has been associated mainly with the autonomous-syntax (‘interpretive semantics’) position, while the analysis of auxiliaries as main verbs first proposed by Ross 1969 has been aligned with ‘generative semantics'. This correlation is an accidental one. If less commonly discussed members of the class of auxiliary verbs are taken into account, it is possible to adduce overwhelming evidence that no line can or should be drawn between them and the items which are categorized uncontroversially as verbs. Considerations internal to several recent autonomous-syntax analyses argue for the main-verb analysis. Under the most recent revision of the AUX analysis, by Akmajian & Wasow 1975, it can be shown that at least some surface modal auxiliaries MUST have main-verb sources. We argue that this removes all principled basis for a separate category M (Modal), and suggests that all modals as well as other auxiliaries should be generated in the same way as verbs. We formulate explicitly the relevant base rules, transformations, morphological rules, and lexical categorizations to permit this. Our analysis treats all auxiliaries as verbs, is entirely consistent with current ideas in autonomous syntax, and uses no descriptive devices which would not be used in autonomous-syntax grammars of English for independent reasons. We conclude that, far from being supportive of the ‘generative semantics’ hypothesis, the main-verb analysis raises problems for it.

Examining Gordon & Lakoff's influential article (1971) on conversational postulates, this paper shows that the term postulate is misleading; that G&L's concept of conversational implicature as a case of entailment is mistaken; and that their view of the interaction of conversational implicature and syntactic rules is based on an incorrect analysis. It is suggested that some of the problems in their paper stem from a pernicious ambiguity of the phrase ‘can convey’.

We here propose and defend a system in which lexical representations of words are expressed in their surface-phonetic isolation forms, and in which phonological rules permit lexical pronunciations to be ‘undone’ in various ways that relate segments to other segments. To the extent that the phonological component is descriptively adequate, it permits morphological rules (which apply after each ‘undoing’ of a phonological rule) to parse lexical representations correctly; this accounts for speakers' knowledge that, e.g., baron and baronial are related, though they have not been given this information explicitly. We argue that this represents a significant advance over other phonological theories: it automatically captures certain alternations that require exception features or exceptional rules in other accounts, it provides a principled basis for constraining phonology in ways that are commonly thought desirable, and it explains why so much linguistic innovation takes place at the ‘bottom’ of the grammar.

This paper supports a hypothesis that poetic ‘options’ (at least on the phonological plane) are of the same types as language changes—i.e., that ‘poetic license’ consists of the relaxation, re-ordering, generalization, inversion etc. of rules of the natural language to yield metrically relevant representations that may differ considerably from outputs in the spoken language. The issue of when to measure metrical, rhythmic, or linguistic complexity is discussed, and a case is made for measuring linguistic complexity in the examples presented here.

It is argued here that English contains a distinct class of relative clauses called amount relatives. On the surface, these are very much like restrictive relative clauses, but they have a syntax and semantics which align them more with comparatives than with restrictive relatives. By positing this new class, we can account in a general way for some apparently exceptional behavior of certain relative clauses. Issues discussed include relativization of the logical subject of a sentence which has undergone There-Insertion (such as every man there is), as well as the interaction of idiomatic expressions and relative clause formation.

The investigation of data from many languages has the following results: (a) the characteristics of classifier languages are distinguished, and four types are identified; (b) defining criteria are postulated for classifiers, and it is discovered that every classifier is composed of one or more out of seven categories of classification. It is argued that classifiers typically index some perceived characteristic of the phenomenon to which the classification refers, and so the recurrence of similar noun classes in unrelated and geographically separated classifier languages shows that diverse language communities categorize perceived phenomena in similar ways.

While relational grammar regards grammatical relations as theoretical primitives that play a central role in the formulation of syntactic rules, there has been no critical assessment of the fact that grammatical relations and surface cases are often confused by both traditional and transformational grammarians. This paper shows that grammatical relations and cases must be clearly distinguished, as there are distinct rules that are sensitive to each. Investigation of the data from Japanese and Korean specifically shows that, in these languages, the rule of Quantifier Floating—which the proponents of relational grammar have alleged to be universally sensitive to grammatical relations—must, in fact, be stated in terms of cases, while the rules of Reflexivization and Subject Honorification must be formulated in terms of grammatical relations. In addition to these syntactic processes, this paper critically examines whether there are, in fact, two subjects in the so-called double-subject construction in Japanese.

The derivation of Lat. sum, es(s), est from IE *esmi, *esi, *esti involves methodological problems. The data, as they appear, e.g., in early drama, are well known, but they have been incorrectly described. At face value, the paradigm of Plautus contains allomorphic variation—sum, ess ~ ss, est ~ st—which has traditionally been explained away by ‘aphaeresis’. This has been used as a synchronic morphophonemic rule, generating the surface variants from the underlying forms /ess/ and /est/ respectively, and with the implication that the IE model is related directly to these underlying forms. But there is evidence that ‘aphaeresis’ has no linguistic basis in Latin, and the IE paradigm described must be in terms of the allomorphic variation observable in Plautus. It is claimed here that the development of sum from *esmi is related to the origin of the variation est~st (< *esti). The study is primarily concerned with the mechanism of this remodeling process, but some chronological suggestions are also made.

A morphological peculiarity of Hua, a language of Papua New Guinea, is that object and possessive pronouns with a certain class of verbal and nominal roots are infixed rather than prefixed. This is shown to result from a combination of two reinterpretations, Watkins' Law and the analytic leap, both identifiable as instances of abductive reasoning. The Hua example is particularly instructive in showing how a change from prefixation to infixation, essentially a morpheme metathesis, could have occurred gradually, and in suggesting a systematic source of counter-examples to the putative universal that the order of morphemes in a word is fixed.

The mechanism of reinterpretation can, then, be described ; but functional explanations that have been proposed are only descriptions of tendencies—exceptions to which, though frequent, are unpredictable. The necessary motives of reinterpretation remain unknown.

A number of experimental tasks were conducted in which subjects were asked to interpret and create novel noun + noun compounds. The results indicate that the semantic relationships that hold between the members of these compounds cannot be characterized in terms of a finite list of ‘appropriate compounding relationships’. Rather, the appropriateness of a given relationship depends on the use to which the compound will be put, the interpretability of the compound, and the extent to which it allows full exploitation of the informational resources of the compound form. These constraints derive in part from the fact that compounds, as lexical items, serve different linguistic functions than sentences do. This crucial distinction, and the existence of the constraints revealed by this study, are obscured by models which generate compounds from underlying clausal structures.

Most analyses of cleft sentences have assumed that the subject it of these constructions is semantically empty. It is argued here, however, that cleft sentences are reduced forms of right-dislocated pseudo-clefts, where it is a pronominal reference to the topic which appears at the end of the sentence. The fact that this NP is always pronominalized with it is shown to follow from a more general principle which governs pronominalization in ‘identifying’ sentences.

This paper questions the basic claim of M.-L. Rivero's ‘Referential properties of Spanish noun phrases’ (Lg. 51.32–48)—i.e., that a correlation exists between the ‘specificity’ of the head NP, as she calls it, and the mood of the dependent relative clause. It is demonstrated here that the indicative mood accepts both ‘specific’ and ‘non-specific’ NP's, and that the sentences in the subjunctive mood are either ungrammatical or are actually in the indicative mood. Moreoever, it is shown that the traditional distinction between referential and attributive NP's should not be interpreted as a distinction between specific and non-specific NP's. The final sections of the paper point out that Rivero's data do not show conclusively that existential import affects definite and indefinite NP's alike, nor that these two types of NP's receive their existential interpretation through the same linguistic means.

This paper argues that features should allow the expression of common phonological processes in such a way that the nature of the process will be formally explicit. Many such phonological processes will require multivalued rather than binary features. The following conventions are proposed: (1) Features are not wholly independent, but interact with each other and with syllable structure; (2) the neutral value of each multivalued feature is somewhere near the center of the scale; (3) more than one value of a feature can occur in the same segment, simultaneously (e.g. double articulations) or sequentially (e.g. affricates). Finally, a set of features for consonants is offered.

Two fundamentally different strategies may be employed by very young children learning their first language. The basic assumptions underlying the study of children's language development, however, have provided means for dealing with only one of these strategies: that which proceeds from the parts to the whole (Analytic). This paper reports on a child who evidently proceeded from the whole to the parts (Gestalt) in producing much of his early language. Since further evidence for a Gestalt strategy exists in the literature, albeit implicitly, such a strategy is probably quite widespread, and any theory of language or language acquisition needs to be able to account for it. It is also speculated that there may be neurological bases for the different language learning strategies.

A replication of D. Bickerton's analysis of fi-tu variation in Caribbean English (1971) indicates that a number of overly-strong hypotheses about variation in decreolization must be modified. Variation in decreolization is not always bidirectional, not completely directed by the pressure to acquire the standard language, and not entirely constrained by deep-structure categories. But this is not to deny that variation is systematic. Rather, it is directed by both social and linguistic pressures, and is constrained not only by purely linguistic factors, but also by lexical categories established by psychological processes.

The organization of repair in a corpus of conversations in the Lue, Yuan (or Myang), and Siamese dialects of Tai is examined with regard to the preference for self-correction described by Schegloff, Jefferson & Sacks 1977 for an English corpus. In both corpora, repair is found to be an identically organized sequential phenomenon involving repair segments in the course of ongoing talk. The initiation and outcome of repair, as well as the reticulated details of the relationship between self- and other-correction, are the same in Tai and English.

This paper addresses the problem of defining the notion ‘linguistically significant generalization’; it separates the notion of ‘generalization’ from that of ‘significance’, and proposes a rigorous objective method for determining the significance of a generalization, based on probability theory. The proposal is illustrated by applying it to examples central to theorizing in generative grammar. Precisely formulated versions of claims made by Chomsky on the use of certain abbreviatory notations (e.g. parentheses) and by Halle on certain ‘natural classes’ are investigated and shown to be not significant; i.e., they tell us nothing interesting about the nature of language. Other generalizations—on active/passive selectional restrictions, word-order universal, and noun phrase accessibility—are shown to be highly significant.

The presentation of Washabaugh 1977 represents a retrograde step in the analysis of creole systems. Though he may be correct in his belief that the constraints on complementizer variation were inadequately stated in Bickerton 1971, his decision to replace that analysis by one based on lexical diffusion and the perception of surface structures is vitiated by a number of errors, both analytic and statistical, and ignores what is most interesting about linguistic change, i.e. the capacity of the human mind to make quite abstract generalizations about broad linguistic categories.

This paper, a reply to Rojas 1977, re-affirms my conclusion (Rivero 1975) that specificity is marked in Spanish NP's by the mood of restrictive modifiers, and that it is independent of the degree of definiteness and the existential claims associated with those NP's. Specific and non-specific NP's alike may be referring expressions depending on their position in sentences. However, contrary to what I previously assumed, the notion of specificity is not connected with the pragmatic distinction between the referential and attributive uses of definite descriptions.