Anne Pycha

Dissertation: Morphological Sources of Phonological Length

Linguistics Department, University of California, Berkeley

Committee: Professors Sharon Inkelas, Keith Johnson, Larry Hyman, and Johanna Nichols

 

Abstract.

Diverse phonological alternations are often recruited to accomplish a single morphological effect. A familiar example comes from Finnish (Karlsson 1999), where the nominative singular can be marked by degemination (rattaa- --> ratas ‘wheel’), consonant voicing (hita- --> hidas ‘slow’), or consonant deletion (kokee- --> koe ‘experiment’). Traditional feature- and segment-based phonological theory offers no way to capture such alternations (often referred to collectively as “lenition”) in a single analysis, constituting a significant deficit in our understanding of the nature of spoken words. Alternative theories have offered explanations based on abstract hierarchies (Foley 1977, Schaefer 1982), articulatory restrictions (Ohala 1997, Kirchner 2000), or perceptual information (Harris & Urua 2001, Harris 2005). These theories, although they differ greatly, all share a focus on the lenition of individual segments. Yet the domain of diversity in phonological alternations is far wider than that. Such alternations go beyond lenition: specific morphological effects can be accomplished with mirror-image processes of fortition, as in Northern Sotho. And they go beyond individual segments: morphological effects can also be accomplished by the loss or insertion of multiple segments at a time, as in Hupa. These facts suggest the need for a very different approach to phonological diversity.

My dissertation introduces and defends Resizing Theory, whose claim is that the overall size of a morpheme can serve as a basic unit of analysis for phonological alternations. Size can be calculated along several independent phonetic dimensions, including pitch, amplitude, and duration. Decreases in morpheme size along the duration dimension, which are triggered by contact with other morphemes during word-building, can be accomplished with degemination or deletion of a single segment; this is relatively uncontroversial. Yet as I show, decreases in size can also be accomplished with the deletion of multiple segments or with voicing alternations, which are known to shorten the duration of a consonant (Lisker 1957 et. seq; Kohler 1984; Kluender, Diehl & Wright 1988). In Resizing Theory, each of these changes possesses the same status and is captured with the same formalism. In other words, it is the fact of an overall decrease in morpheme size, and not the particular feature-based or segment-based strategy used to implement it, which is linguistically significant. The same holds for increases in morpheme size.

Resizing Theory makes three major predictions, which I support with evidence drawn from my database of over 100 languages. The first prediction is the most basic one: alternations which decrease (alternatively: increase) the overall size of a morpheme should pattern together in producing specific morphological effects. The second prediction is that contact between morphemes can completely determine the surface output of a word. Of course, it is already well-established that morphology interacts with phonology in order to produce surface forms (e.g. Kiparsky 1982). The novel prediction made by Resizing Theory is that interactions between morphemes - specifically, the imperative for one morpheme to change its size - have the potential to play the only role in producing a phonological form, without reference to diacritic features or segments. The third prediction concerns null allomorphs, which have been treated as special cases of morphological “blocking” in the literature (e.g. Anderson 1986). In Resizing Theory, by contrast, null allomorphs have no special status: they are simply the logical endpoint of decreases in morpheme size, and they exist along a cline with partial and full allomorphs.

The empirical scope of Resizing Theory is at once more narrow and more broad than that of previous theories of lenition. It does not explicitly encompass consonant manner alternations, which I suggest are better accounted for by an orthogonal tendency for length to be associated with stop articulations, rather than fricative or a approximant articulations (Elmedlaoui 1993, Kirchner 2000). The theory does, however, have the potential to encompass changes in size along other dimensions, such as pitch, and I sketch several ways in the three predictions of Resizing Theory are borne out in tonal phonology.

Thus, this dissertation explicitly and formally adopts a bird’s eye view of phonological alternations that occur at morpheme boundaries, and integrates well-established facts about phonetic duration directly into the abstract unit of morpheme size. In so doing, it offers a novel solution to the problem of diversity in phonology, and makes a contribution to the literature on both the phonetics-phonology and phonology-morphology interfaces.