Archangeli (1987) has pointed out that the hierarchical model of feature representation combined with the statement of phonological rules in terms of conditions and parameters offers the advantage that it allows the expression as a single rule of unitary processes that must be stated as multiple operations within other frameworks. In this paper I will offer an example of this (cf. Hualde, 1987 for another example). I will show that a seemingly complex process of palatalization that must be stated as two related but different operations within a linear model, can be straightforwardly captured in the hierarchical /parametrical approach by taking into account the geometrical structures on which the palatalization rule applies; in particular, the branching structures created by a rule of place assimilation. I will assume that assimilatory processes have the effect of creating complex structures where features or nodes are shared by several segments. From this assumption we can make predictions about how other rules may apply to the output of a process of assimilation. These predictions are very different in some cases from what one would expect from a formulation of the rules in a linear, feature -changing framework. In the case to be examined here, the predictions made by taking into account derived geometrical structures receive very strong confirmation. I will consider a rule of palatalization in two Basque dialects. In one of them, the process of palatalization can be captured quite simply by a linear rule. In the other dialect, the facts appear as more complex and requiring several operations within a linear framework, but are actually simpler to state within a geometrical /parametrical framework. Only within such a theory can we capture the fact that the more pervasive palatalization observed in this second dialect arose from a simplification in the rule that other dialects possess.

No description available.

No description available.

A major claim of this paper is that the distinctive features of lexical accent are formally identical to those of tone, or at least to a subset of tonal features. The terms accent and tone have been used in many different ways in the literature, but throughout this paper I will use both terms to refer only to lexical features that surface as contrastive pitch, length, volume and/or other features of prominence. By lexical I mean features those phonetic realization cannot be predicted by any regular metrical structure or phonological rule. I am assuming that the placement of stress is always determined by a set of language- particular (but parameter - based) rules which build metrical structure, with the location of exceptional stress indicated by a lexical diacritic called accent. Examples of such systems of rules are described in Hayes (1982), Hammond (1986) and Halle and Vergnaud (1987 a and b). Although metrical structure has generally been associated with non-tonal languages, there are also some tonal languages which exhibit the presence of metrical structure. Examples of such languages include Creek (Haas (1977)), Malayalam (Mohanan (1982)) and Capella Trique (Hollenbach (1988)). Thus the presence of metrical structure is not sufficient in itself to distinguish a non-tonal language from a tonal language. What, then, distinguishes these two categories from one another? There are two general distinctions which have traditionally been made in classifying languages as tonal versus non-tonal. One distinction is that many tonal languages exhibit a variety of lexically contrastive tones, while mast, if not all, of the degrees of stress in a non-tonal language can usually be explained using only one kind of lexical accent. Thus, tonal languages can have more than one kind of lexical tone, whereas non -tonal languages can have lexical accent but not tone, and there is apparently only one kind of lexical accent. I will discuss this apparent asymmetry in section three. The other distinction between tonal and non-tonal, for which I present counterevidence in this paper, is that autosegmental status has been attributed to tone, but not to accent, in a number of languages; see, for example, Goldsmith (1976), Williams (1976) and Pulleyblank (1983). For all such languages, the Universal Association Convention (UAC) (Goldsmith (1976)) predicts the location of most tones, with the remaining tones accounted for by lexical pre- linking. From an examination of the literature it appears, then, that the main distinction between the terms tonal and non-tonal is that tonal languages have lexical tone while non-tonal languages have lexical accent. Formally, both of these devices are lexical diacritics, but they appear to differ in that tone can be an autosegment, while no such status has ever been claimed for accent. Therefore, the question to be addressed in this paper is this: Can an accentual diacritic have autosegmental status? Using data from Mayo, a Uto-Aztecan language of northwestern Mexico, I will show that the answer is yes. The implication, then, is that accent is formally the same as tone, or at least the same as one variety of tone. A significant claim follows from this. If accent is formally the same as a tone, then no language can exist in which lexical accent occurs independently of all tonal features. As far as I know, no such language has been shown to exist. The paper is organized as follows. Section one presents the data and provides two possible analyses of Mayo stress using the theory of Halle and Vergnaud (1987 a and b) (henceforth H&V). I show that Mayo has lexical accent which floats in underlying representation (UR), just like an autosegmental tone. Section two demonstrates that stress assignment crucially has to precede and follow reduplication, thus indicating that the rules of stress assignment are cyclic and that lexical accent refloats at the end of each cycle. In section three I explore the theoretical implications of this analysis and propose that accent is formally the same as tone.

American Sign Language (ASL) signs are claimed to be composed of four parameters: handshape, location, movement (Sto]çoe 1960) and palm orientation (Battison 1974). This paper focuses solely on handshape, that is, the configuration of the thumb and the fingers in a given sign. Handshape is significant in ASL and Chinese Sign Language (CSL); that is, minimal pairs exist for handshape in each. Thus, the two ASL signs in (1) differ in one parameter: the handshapes are different, but the location, palm orientation and movement are the same. Similarly, the two CSL signs in (2) differ in one parameter: handshape. A logical next question asks if handshapes are further divisible into parts; more specifically, are handshapes composed of distinctive features? This question is not new; in fact, researchers have made many proposals for ASL handshape features (Lane, Boyes -Braem and Bellugi, 1979; Mandel, 1981; Liddell and Johnson, 1985; Sandler, 1989; Corina and Sagey, 1988 and others). This paper focuses on the proposal of Corina and Sagey (1988). In Section 2, I outline the proposed system for the distinctive handshapes of ASL, of which [lateral] is a part. Then using data from ASL and CSL, I give three arguments in support of the claim that there is not sufficient justification in ASL or CSL for the feature [lateral]. First, I show in Section 3 that the prediction which follows from the claim that [lateral] applies only to the thumb, namely that the thumb behaves differently from the other fingers, is not borne out by CSL data. Second, I argue in Section 4 that since other features (proposed by Corina and Sagey, 1988) can derive the same phonetic effects as [lateral], [lateral] is unnecessary to describe thumb features in either ASL or CSL. Third, in Section 5, I use ASL and CSL data to argue that the notion of fingers as "specified" or "unspecified ", although intuitively pleasing, should be discarded. If this notion cannot be used, the feature [lateral] does not uniquely identify a particular set of handshapes. I show that CSL data suggests that two other features, [contact to palm] and [contact to thumb] are independently needed. With these two features, and the exclusion of [lateral], the handshapes of both ASL and CSL can be explained. In Section 6, the arguments against [lateral] are summarized.

The goal of this paper is to develop a fully formalized, (i.e. generative) privative vowel-feature theory which is a viable alternative to conventional feature theory for vowels. To do this, certain revisions to the theory of Particle Phonology will be proposed.

Madurese, a Malayo-Polynesian language, is of particular interest to theories of vowel harmony and feature geometry because of the interaction of consonants with vowels, and the problem of representing both transparent and opaque segments within the same language. Vowels divide into two sets, occuring exclusively after each of two sets of consonants. Isolation of this process is somewhat complicated by loan words showing no alternation or containing non -native vowels or consonants. In this paper I will examine vowel- consonant interaction in native Madurese words. All data are from H.N. Kiliaan (1904), Madoereesch- Nederlandsch Woordenboek. and Stevens (1968), Madurese Phonology and Morphology, along with additional data from Stevens (1980), "Formative Boundary in Phonological Rules."

In this paper, I provide an analysis of vowel harmony in Izi, an Igbo language spoken in the East - Central State of Nigeria. Using data from Meier, Meier, and Samuel (1975; hereafter MMB), I argue that harmony in complex verbal structures in Izi is inadequately accounted for within a level ordered model of lexical phonology (Kiparsky 1982, Mohanan 1982, etc...), claiming instead that harmony facts are best accommodated within a non-level ordered approach (cf. Halle and Vergnaud 1987, Halle and Kenstowicz 1991; Halle, Harris, and Vergnaud 1991). In sections 1 and 2, I provide a description of the general pattern of the [ATR]-based vowel harmony system in Izi and motivate [+ATR] as the only value of the feature [ATR] present at the level of underlying representation. In section 3, data are presented demonstrating the inadequacy of a level -ordered treatment of vowel harmony in verbal structures. Finally, in section 4, I propose an alternative, non-level ordered analysis that derives the attested harmony facts via cyclic rule application at a single level. Crucially, particular morphemes in verbal structures are claimed to undergo a pass of the cyclic rules prior to concatenation, a phenomenon which I call selective cyclicity.

No description available.

One of the pillars of phonological research has been the desirability of representing phonological processes as being local in application. Locality, as a principle of the grammar, constrains the relation between the trigger and target elements of a phonological process to one of adjacency. Adjacency, within the framework of Autosegmental Phonology and Underspecification theory, consists of two varieties: tier adjacency and structural adjacency (Myers (1987)). Tier adjacency examines linear relations among elements within an isolated tier of the representation (e.g. the tonal tier), while structural adjacency examines these relations mediated through the skeletal core, which organizes and maintains the linear relations between phonemes and their constituent elements. Locality and Adjacency are not, simply the preserve of featural relations and their skeletal core. The core itself, whether viewed as C/V slots, X/X' timing slots, or Root nodes, is organized into the grander structures of the Prosodic Hierarchy (e.g. syllable, Foot, etc.) . The formation of these units is a phonological process and as such subject to the same principles. A portion of the on -going debates in metrical theory has focused on whether metrical structure, in particular Foot structure, is limited to binary constituents. Kager (1989) proposes an extreme Binarism, with all metrical structure initially being limited to binarity. Hayes (1987) and Prince (1990) only commit to a strong preference for binary Feet. Halle & Vergnaud (1987) propose a system allowing binary, ternary, and unbounded Feet. The principle of Locality with its requirement of adjacency argues for a binary -view of metrical structure where the trigger and target of the structure building process are un- metrified elements. The most serious challenge to this view is the existence of languages which employ ternary constituents, e.g. Cayuvava, Chugach Alutiiq. These languages have been cited as evidence in arguing for a theory capable of generating ternary Feet. In a framework designed to maintain strict locality surface ternary constituents must be derived from underlying binary structures. This paper proposes a solution to this problem which relies on the ternary constituent being a complex constituent composed of a binary Foot grouped with an adjacent syllable. This constituent is not a Foot, but rather a Prosodic Word. Generating an iterative ternary Prosodic Word requires a new algorithm for building metrical structure. This algorithm builds metrical constituents in an opportunistic manner. Opportunistic building creates metrical constituents as soon as possible, instead of applying one particular structure building rule across the whole string before the next rule applies. This paper examines these issues through the metrical structures of the Alutiiq dialects described by Leer (1985a). The rich and detailed work of Leer serves admirably as a base for elucidating the issues of ternarity. Unfortunately, the ramifications of these proposals beyond the issue of ternarity can only be briefly alluded to in this paper. Length constraints do not permit me to present all aspects of these proposals in the full detail they require for their justification.