This paper proposes principles to revise the strict ‘double articulation’ of standard linguistic theory, and define a class of optional rules of postlexical phonology whose output bears meaning. The class includes prosody-modifying processes of Central Alaskan Yupik Eskimo (CAY)—which, when applied to a word or prosodic phrase, add linguistically significant expressive or other pragmatic meaning. Their postlexical rule status is suggested by their phonological form, their gradience and dependence (in some cases) on above-word prosodic phrasing, and above all by their intricate inter-ordering among the obligatory prosody rules of CAY postlexical phonology. The phonological treatment prescribed by the proposed principles is shown to handle such features naturally and easily, whereas treatments in keeping with strict double articulation do not. The proposed revision of the double articulation principle promises to preserve modularity in a way which simplifies individual grammars, and which makes testable empirical predictions about universal grammar.

This study offers a theory of agreement within the framework of lexical-functional grammar. In this theory, verbal affixes may mark either grammatical or anaphoric agreement. In grammatical agreement, a NP bears an argument relation to the verb, while the verbal affix expresses redundantly the person, number, and gender class of the NP. In anaphoric agreement, the verbal affix is an incorporated pronominal argument of the verb, and the coreferential NP has a non-argument function—either as an adjunct of the pronominal argument, or as a topic or focus of the clause or discourse structure. Grammatical topics have syntactic properties deriving from a theory of discourse functions. The minimal difference between an incorporated pronoun and a grammatical agreement marker is the presence or absence of a semantic attribute in the lexical content of the affix. Likewise, the minimal difference between a subject NP and a topic NP is the function, rather than the phrase structure attributes of dominance, precedence, and category. This theory offers an explanation for the close relation between grammatical and anaphoric agreement, as well as clear criteria for distinguishing anaphoric and grammatical agreement by their discourse, syntactic, and even phonological effects. The Chicheŵa language (Bantu) has both grammatical agreement with the subject and anaphoric agreement with the object, and related languages show various distributions of these two agreement types.

In previous accounts, the word order of Papago (Uto-Aztecan, Arizona) has been described—in terms of syntactic roles—as SOV, SVO, and VSO. However, discourse data show that (surface) order is most insightfully accounted for by strong pragmatic principles. Information preceding the verb is either pragmatically marked, or is information for which the hearer is instructed to ‘open a new active discourse file’. Other information follows the verb. Syntactic role correlates highly with the discourse-pragmatic and semantic status of the information encoded, but has almost no correlation with order. Therefore, in terms of the typological tradition represented by J. Greenberg, J. Hawkins, and others, no particular order of syntactic roles should be taken as basic. Rather, languages in which order is not based on syntactic role should simply not be forced into an order typology based on syntactic role.

This paper examines the phenomenon of ergativity and its relation to patterns of surface grammar and information flow in discourse. Corresponding to the grammatical pattern of ergativity (exemplified in the A vs. S/O distribution of verbal cross-referencing morphology in Sacapultec Maya) there is an isomorphic pattern of information flow: information distribution among argument positions in clauses of spoken discourse is not random, but grammatically skewed toward an ergative pattern. Arguments comprising new information appear preferentially in the S or O roles, but not in the A role—which leads to formulation of a Given A Constraint. Evidence from other languages suggests that the ergative patterning of discourse extends beyond the ergative type to encompass accusative languages as well. Given the linguistic consequences of a type-independent Preferred Argument Structure, it is argued that language-internal phenomena as fundamental as the structuring of grammatical relations can be shaped by forces arising out of discourse, viewed as the aggregate of instances of language use.

The aim of this paper is to re-examine constraints on relative clause formation in languages of the world. The main tasks of the study are thus two-fold: to present a new formulation of one of the main constraints on relativization, and to challenge an assumption of the most robust typological interpretation of relative clause constraints to date—the Noun Phrase Accessibility Hierarchy.