Faithfulness in Prosodic Morphology and Phonology: Rotuman Revisited
Optimality Theory (Prince & Smolensky 1993) deals with constraints on surface forms. Yet it also depends crucially on constraints that regulate the faithfulness of the surface form to the lexical structure. The interaction of surface constraints on phonological markedness and faithfulness constraints, through ranking, is essential to characterizing particular grammars within OT. Thus, without faithfulness constraints, the claim that individual grammars differ only in how they rank a set of universal constraints would be untenable. Indeed, without faithfulness constraints, there would be no explanation for why every word in every language isn’t driven inexorably toward some maximally unmarked form, like ba, ti, or .
Faithfulness, then, is indispensable to the construction of Optimality-Theoretic grammars. It follows that the precise way in which faithfulness is understood — the implementation, rather than the core concept — is an area that should be examined closely. In this article, I will argue for a particular conception of faithfulness based on a relation of correspondence between strings of phonological elements. Correspondence was introduced as a theory of reduplicative copying (McCarthy & Prince 1993a, 1994a). Parallels between reduplicative copying and faithfulness lead to a generalized theory of correspondence (McCarthy & Prince 1994b, 1995), enlarging in several directions the conception of the nature and role of faithfulness in earlier OT work, which begins with Prince & Smolensky (1991, 1993).
I will explore several ways in which correspondence permits an expanded view of faithfulness, with a variety of desirable empirical consequences. After an introduction to Correspondence Theory in section 2, the article continues with the examination of these main themes:
•The extension of faithfulness constraints to alternations involving phonological metathesis (section 3), which has not been considered in most previous OT work.
•The extension of faithfulness constraints to the preservation of prosodic structure (section 4), subsuming both familiar faithfulness effects and phenomena previously attributed to prosodic circumscription (McCarthy & Prince 1990a). This work further develops some ideas about prosody in reduplicative correspondence (the constraint StRole) in McCarthy & Prince (1993a, 1994a), and parallels the independent developments realized in Itô, Kitagawa, & Mester (1995).
•The extension of faithfulness constraints to relations other than the one between lexical and surface forms (section 5), pursuing proposals about reduplicative correspondence in McCarthy & Prince (1994b, 1995) and about output-output correspondence in Benua (1995, forthcoming). Following Benua, this work connects with traditional notions of paradigm uniformity (see, e.g., Bybee 1985) and more recent developments along these lines, such as Burzio (1994ab), Flemming & Kenstowicz (1995), Kenstowicz (1995), Kraska-Szlenk (1995), and Orgun (1994).
These topics are strongly interconnected, since all depend crucially on Correspondence Theory and since all come to the fore in the analysis of Rotuman, which serves as the empirical focus for this article.
Rotuman is a central Oceanic language spoken on an island about 300 miles north of Fiji. This language was comprehensively described and analyzed by Churchward (1940). A very extensive secondary literature starts from Churchward's results, including contributions by Haudricourt (1958ab), Biggs (1959, 1965), Milner (1971), Anttila (1989), Cairns (1976), Saito (1981), van der Hulst (1983), Janda (1984), McCarthy (1986, 1989), Mester (1986), Besnier (1987), Hoeksema & Janda (1988), Odden (1988), and Blevins (1994).
The phenomenon that has captured most of this attention is the morphological distinction of “phase”, first recognized by Horatio Hale, philologist to the United States exploring expedition of 1838–1842. Rotuman has a contrast in major-category words between two phases, the complete and the incomplete, distributed according to syntactico-semantic principles (see Appendix B). The phase distinction has diverse morpho-phonological effects, evidenced in examples like the following (Churchward 1940, Besnier 1987):2
() Phase Differences in Outline
tokiri tokir ‘to roll’
ti u ti ‘big’
sulu sul ‘coconut-spathe’
r ko r k ‘to imitate’
i i ‘fish’
sesev sese v ‘erroneous’
hos ho s ‘flower’
pure puer ‘to rule’
parofit parofi t ‘prophet’
t fi taf ‘to sweep’
hti ht ‘to embark’
mose mös ‘to sleep’
futi füt ‘to pull’
pupui pupu i ‘floor’
lelei lele i ‘good’
keu ke u ‘to push’
joseu joseu ‘Joshua’
e. No Formal Distinction of Phase
r r ‘house’
r r ‘to do’
sik sik ‘cigar’
The principal descriptive goal of this study is to comprehend these various formal properties of the phase distinction within a single coherent view of Rotuman phonology, keeping a sharp eye on the relevance of the analysis to the theory of faithfulness.
In addition to these theoretical and empirical objectives, this article also offers discussion of several other issues of potential interest. Throughout the article, but particularly in section 3.4, evidence is presented in support of Optimality Theory over operational approaches to phonological description. The same section also addresses the question of substantive versus formal restrictions on linguistic expressions, when the matter of “C/V tier segregation” in Rotuman is addressed. The theory of templates in prosodic morphology is touched on in connection with the proper characterization of the incomplete phase (section 3.2). Also within prosodic morphology, the derivational theory of positive prosodic circumscription is called into question, and an Optimality-Theoretic alternative is presented (section 4.2).
The results in this article are set within Optimality Theory (Prince & Smolensky 1993), and they rest upon correspondence (McCarthy & Prince 1993a, 1994b, 1995), which extends and develops the original OT conception of faithfulness. Correspondence is a general way of relating representations to one another. The representations related by correspondence may be lexical and surface, base and reduplicant, or other pairs, such as a base and a derived form in root-and-pattern or truncatory morphology. Rankable constraints apply to correspondent elements, demanding completeness of correspondence, preservation of linear order under correspondence, and the like. Correspondent segments are often identical to one another, but identity of correspondents is also enforced by rankable, and therefore violable, constraints.
Correspondence was proposed as a theory of reduplicative copying in McCarthy & Prince (1993a); it is generalized to the lexical-surface relation in McCarthy & Prince (1994b, 1995); further extension of correspondence to intra-paradigmatic relations is pursued in Benua (1995, forthcoming) and below, in section 5.
Correspondence is defined as follows: