Cyclicity and Linearity in Morphology: The View from Icelandic, Gã, and Kabyle

Abstract

Much work in the generative tradition assumes that the grammatical architecture consists of independent modules. A number of interesting questions arise concerning the ways in which these modules communicate with one another at their interfaces. This dissertation investigates the role of cyclicity and linearity in the interaction of syntax, morphology, and phonology. It is couched in the Distributed Morphology theoretical framework (DM; Halle & Marantz 1993, 1994), and draws on data from the Icelandic (Germanic, Iceland, iso:isl), Gã (Kwa, Ghana, iso:gaa), and Kabyle (Amazigh, Algeria, iso:kab) nominal and pronominal domains. There is a disparity between morphosyntactic and morphophonological bodies of literature concerning the location of cyclic boundaries, stemming from varying approaches to cyclicity within syntactic and phonological frameworks. Morphophonologically-oriented approaches propose that the cyclic domain of morphological application is located at a phonologically-defined prosodic boundary (Kiparsky 1982). Morphosyntactically-oriented approaches propose that this domain corresponds to the syntactic phase (Embick 2010). Drawing on data from Icelandic and Kabyle, I find that phonology, and consequently, morphology, is local and cyclic at phase boundaries. Apparent exceptions to phonological phase cyclicity may result from morpheme-specific cophonologies. These findings support a DM-compatible morphophonological framework like Cophonologies by Phase (Sande et al. 2020). Additionally, I argue that the Phase Impenetrability Condition is inactive in phonology, but that phase-based phonological faithfulness constraints may result in pseudo-Phase Impenetrability effects. Proposing that cyclic domains are isomorphic between syntax and morphology raises the question of whether the morphological component of grammar is maximally simple, such that the linear phonological string can be read directly from the hierarchical morphosyntactic structure (Svenonius 2012), or if morphology requires certain operations to be sensitive to linear order, suggesting that the morphosyntactic structure is linearized prior to lexical insertion (Arregi & Nevins 2012). Introducing new data from Gã, I demonstrate that certain portmanteaux must be generated based on linear adjacency. This finding supports a classic DM-type approach wherein morphological operations, some of which may be sensitive to linear adjacency, mediate between syntax and morphology. Although phonological domains of application may map to syntactic phases, phonological representations cannot be read directly from syntactic structure.