How Autosegmental is Phonology?

Abstract

UC Berkeley Phonology Lab Annual Report (2013) How Autosegmental is Phonology? Larry M. Hyman University of California, Berkeley Paper presented at the Colloque du Reseau Francais de Phonologie, Paris, June 25-27, 2012 Introduction In this paper I address the following questions: How appropriate/effective are traditional autosegmental representations? What insights do autosegmental representations help us express? Where do autosegmental representations fall short? And finally, what does this all mean for phonology? Such questions are particularly timely, as numerous scholars have abandoned both earlier assumptions of classical autosegmental phonology, particularly as concerns representations. I believe there still is general acceptance of the “basic autosegmental insight”, i.e. that certain features are semi-autonomous from each other and from their anchors, e.g. tones vs. tone-bearing units (TBUs), which are represented on separate tiers (Goldsmith 1976a,b; Pulleyblank 1986, 1989). However, as recent developments in phonology have shown decreasing interest in representational issues, there has been less and less reliance on the key insights of autosegmental phonology. In what follows I will focus on the interaction of the following properties assumed in traditional autosegmental phonology: (i) Assimilation = spreading (Hayes 1986a). Whether concerning tone, vowel harmony or other processes, assimilation is captured by spreading a feature from one anchor to another, represented by a dashed association link, as in (1a). This contrasts with the earlier view represented by SPE (Chomsky & Halle 1968) where assimilation was expressed via feature copying, as in (1b). a. X X b. [+F] X X [+F] X X [+F] [+F] Such a view has been either sidestepped or abandoned in frameworks such as optimal domains theory (Cassimjee & Kisseberth 1998) and “search and copy” (Nevins 2010, Samuels 2011). In certain cases of apparent long-distance assimilation, e.g. consonant harmony, spreading has been replaced by agreement by correspondence (Rose & Walker 2004, Hansson 2010). While Rose & Walker suggest that both spreading and agreement by correspondence are needed for different processes, others have been extending ABC more generally, e.g. to vowel harmony (Rhodes 2010) and local assimilations (Shih 2011). The direction definitely seems to be away from assimilation as spreading. (ii) No line crossing (Goldsmith 1976a, b). A second property concerned the prohibition against association lines crossing each other. While critically evaluated by Coleman & Local (1991), most practicing autosegmentalists continued to assume that assimilation by spreading could not produce representations such as in (2). X 1 X 2 [+F] [-F] X 3