The architecture of grammar and the division of labour in exponence

Abstract

Languages commonly exhibit alternations governed by complex combinations of phonological, morphological, and lexical factors. An alternation of this sort will often admit a wide variety of analyses, each apportioning different roles to lexical storage and to morphological and phonological computation. Such analytic underdetermination poses a threat to falsifiability and to learnability: hypotheses can easily evade empirical disconfirmation if potential counterexamples can be redescribed in many different ways to suit the linguist's convenience, and so theories risk losing empirical content; by the same token, it becomes hard to explain how, among a plethora of choices, learners converge upon the target grammar (§2). To avert these dangers, the theory of grammar must set limits to the space of possible interactions between phonology, morphology, and the lexicon: in particular, it must ascertain the proper division of labour between storage and computation (§3), and it must constrain the ways in which morphological operations can manipulate phonological material and in which phonological processes can refer to morphosyntactic information (§4). Concerning the question of storage vs computation, this paper pursues the hypothesis that different types of alternation reflect different modes of interaction between the lexicon and the grammar. This idea is fleshed out by means of a refined dual-route approach to exponence (§3.1, §3.5), in which the well-established distinction between explicit symbolic generalization and implicit pattern association (§3.1, §3.4) is supplemented with a novel distinction between two types of lexical listing, analytic and nonanalytic, akin to Clahsen and Neubauer's (2010: 2634) contrast between ‘combinatorial entries’ and ‘unanalysed entries’ (§3.1, §3.3.1). Assuming a stratal version of Optimality Theory (OT), I show that the peculiar syndrome of properties characteristic of stem-level morphophonology arises from the fact that stem-level forms are stored nonanalytically but stem-level processes are nonetheless explicitly represented in the grammar by means of symbolic generalizations, whose status resembles that of Jackendoff's (1975) lexical redundancy rules (§3.3.1). The model provides a highly explanatory account of internal cyclic effects in stem-level domains, which I illustrate with classic examples such as English órigin ~ oríginal ~ orìginálity (§3.3.2) and còmp[ə]nsátion vs cònd[ɛ]nsátion (§3.3.3). On the issue of the interaction between morphology and phonology, this paper argues for the adoption of a restrictive stance based on general cognitive principles of modularity and locality (§4.1). A programme is proposed consisting of four hypotheses (37): that morphology selects and concatenates morphs without ever altering their phonological content (§4.2); that phonological constraints other than those on prosodic alignment may not refer to morphosyntactic information (§4.3); that output phonological representations do not contain diacritics of morphosyntactic affiliation (§4.4); and that morphosyntactic conditioning in phonology is subject to cyclic locality (§4.4). These hypotheses provide the guiding thread for an evaluation of several mechanisms currently used to describe morphologically conditioned phonological processes, including construction-specific cophonologies (§4.2.3), indexed constraints (§4.3), and readjustment rules (§4.3). The balance of argument supports a stratal-cyclic architecture for phonology—one, however, in which neither cyclicity nor stratification are innately stipulated, but both emerge from fundamental storage and processing mechanisms (§3.3.2, §3.3.3) and from timing effects in the child's linguistic development (§4.2.3).

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 88 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.