Tone Circles and Contrast Preservation

This squib investigates a type of circular chain shift, known as a tone circle, found in dialects of the Min language of the Sino-Tibetan family, specifically the Xiamen dialect. Tone circles are problematic for classical Optimality Theory (OT), leading many researchers to conclude that they are instances of paradigmatic replacement. I argue that tone circles are indeed phonological processes. Thus, an enrichment of the theory beyond that of classical OT is required. I discuss how a solution along the lines of contrast preservation (Łubowicz 2003) can fill the gap. I hasten to add that Łubowicz's theory excludes the possibility of circular scenarios that do not exhibit any neutralization of contrasts. In line with this prediction, Xiamen does exhibit neutralization. The analysis sketched here exploits Łubowicz's proposals and, without making any changes to her theory, is capable of explaining the Xiamen tone circle.

The squib is organized as follows. Section 1 discusses Xiamen tone sandhi. Section 2 argues that tone circles are active phonological processes. Section 3 discusses tone circles as optimal solutions to contrast preservation. Section 4 offers a brief conclusion.

1 Xiamen Tone Sandhi

Xiamen, like many Southern Min languages, exhibits what has become known as a tone circle. Sandhi tones appear in nonfinal position within the tone sandhi domain. Citation tones appear in final position of the tone sandhi domain or in isolation.¹ (1) gives some examples of tone sandhi in Xiamen, where the underlying tone is italicized and the sandhi tone is boldfaced (Chen 1987). The tone sandhi are summarized in (2), which clearly shows their circular nature. The numbers refer to "tone letters" as described by Chao (1930): 5 represents the highest pitch level, and 1 represents the lowest pitch level.

(1)

1. a. we-24 'shoe'              we-22 tua-21 'shoe laces'

2. b. wi-22 'stomach'            wi-21 pih-22 'stomach ailment'

3. c. ts'u-21 'house'           ts'u-53 ting-53 'roof top'

4. d. hai-53 'ocean'              hai-44 kih-24 'ocean front'

5. e. p'ang-44 'fragrant'     p'ang-22 tsui-53 'fragrant water'

(2) Citation and sandhi tones in Xiamen (Chen 1987)

Early attempts to analyze this process relied on highly unnatural rules or stipulations (Wang 1967, Yip 1980), which led some researchers to suggest that the pattern of tone sandhi is paradigmatic. The fact that so many Southern Min languages exhibit similar or identical tone circles suggests that a systematic explanation is in order. In fact, two recent works (Hsieh 2005, Mortensen 2002) have proposed OT analyses relying on antifaithfulness constraints (Alderete 2001). The current proposal appeals to contrast preservation, which does not rely on ad hoc rules or on antifaithfulness constraints. The key difference between the contrast preservation approach and the rule-based and antifaithfulness approaches concerns the motivation for the chain shift. The contrast preservation approach admits only circular chain shifts that achieve some sort of neutralization, whereas the other approaches do not have this restriction. We shall see that the Xiamen tone circle does exhibit neutralization spurred on by a high-ranking markedness constraint.

Let us begin with the representation of tone. Most analyses assume a register feature that bisects the tonal space into the Upper and Lower registers, and a pitch feature, designated [hi] or [lo], that fine-tunes the pitch of the tone within the register.² The two most widely debated representations for contour tones are the contour tone unit (CTU) (Bao 1999, Yip 1980, 1989) and the tone cluster (Duanmu 1990, 1994). For simplicity, I assume a CTU approach; however, the ensuing discussion is compatible with either approach. CTU analyses typically assume a single register feature and a branching pitch feature to represent a contour tone. Example (3a) represents a high-rising tone. The branching Pitch node indicates that the tone moves from the low end of the Upper register to the high end. I use the abbreviated form in (3b) throughout.

(3)

1. a.

2. b. [U,lH]

Abstracting away from the feature-geometric representations and concentrating only on the feature content of the tones, I suggest a plausible set of features for Xiamen tones in (4).

(4) 44...