|
Article Excerpt
Abstract. In man, y languages that permit coronal laterals to follow labial and velar stops in complex onsets, sequences of a coronal stop followed by a coronal lateral are prohibited. Standard accounts rule out coronal-lateral clusters as an effect of the Obligatory Contour Principle, but this approach cannot explain languages such as Mong Njua and Katu, which neutralize the coronal-velar place contrast yet still allow coronal-lateral clusters to appear. Recent work in Dispersion Theory (Flemming 1995, 2002, Padgett 2003a, b,e) has argued that Optimality Theory (Prince and Smolertsky 1993/2004) must also include systemic constraints that evaluate phonological forms in the context of the larger system of contrasting forms in a language. This paper offers a new Dispersion-theoretic analysis of restrictions on onset clusters involving laterals. Systemic markedness constraints penalize indistinct coronal-velar contrasts in different pre-lateral contexts. Directionality of neutralization is determined by faithfulness constraints on input place, whose ranking can vary across languages and dialects (Hume 2003, Hume and Tserdanetis 2002). The proposed analysis solves problems with earlier accounts and also encompasses typological patterns from over forty languages, including velarization in early Romance sound change and Mexican Spanish loanword adaptations from Nahuatl. (1) - 1. Introduction. In many languages that permit coronal laterals to follow labial and velar stops in complex onsets, sequences of a coronal stop followed by a coronal lateral are prohibited. For example, English allows the clusters in 1 a and 1c but prohibits the coronal stoplateral clusters in 1b.
(1) a. /pl/,/bl/ plead, bleed, plank, blank b. */tl/, */dl/ c. /kl/,/gl/ clue, glue, class, glass
Standard generative analyses invoke similarity avoidance and formalize the restriction as an effect of the Obligatory Contour Principle (henceforth, OCP; Leben 1973, McCarthy 1986), which prohibits adjacent identical segmental specifications. The OCP bans the coronal-coronal clusters in 1b but allows clusters in which the initial stop is labial as in 1a or velar as in 1c. However, the OCP cannot account for languages in which coronal stop-lateral clusters occur either in free variation with or to the exclusion of velar stop-lateral clusters. In Mong Njua (Northern Thailand; Lyman 1974) and Kam (South Vietnam, Laos; Wallace 1969), the OCP-violating clusters themselves are not problematic. Rather, these languages prohibit a contrast between coronal and velar stops before laterals.
[FIGURE 1 OMITTED]
Flemming (1995, 2002) considers these restrictions as evidence for Dispersion Theory. In addition to the standard faithfulness and markedness constraints of Optimality Theory (Prince and Smolensky 1993/2004), Dispersion Theory also includes constraints that govern the well-formedness of phonological contrasts. These constraints are SYSTEMIC inasmuch as they evaluate phonological forms in the context of the larger system of contrasting forms in a language. In this paper, I develop an analysis of stop-lateral clusters within the version of Dispersion Theory elaborated by Padgett (2003 a, b, c), which admits input representations and input-output faithfulness constraints. Furthermore, I adopt the proposal by Hume (2003) and Hume and Tserdanelis (2002) that place markedness is not universal but can vary cross-linguistically. I argue that different patterns of place neutralization in stop-lateral clusters are determined by the interaction between systemic markedness (SPACE) constraints, which penalize indistinct coronal-velar contrasts in different segmental contexts, and Max(place) constraints, whose language-specific ranking determines the directionality of neutralization (i.e., to coronal, velar, or variably to both). I show how the proposed analysis avoids the problems faced by an alternative Dispersion-theoretic account, suggested by Flemming (1995, 2002), that appeals to perceptual distinctiveness of affricates. In addition, the analysis put forth here predicts that stop-lateral onset cluster phonotactics should follow independently attested place assimilation asymmetries in the language, which is substantiated on the basis of available data from English, Catalan, and Mong Njua.
The place neutralization account is further extended to cover data from Romance and other languages. First, no variety of Spanish allows /dl/ clusters, while only some varieties allow /tl/. Harris (1983) proposes dialect-specific filters against adjacent coronal noncontinuants, while Martinez-Gil (2001) and Gerlach (2004) suggest a possible analysis in terms of the OCR These analyses fall short when the [+continuant] allophone [[eth]] of Spanish /d/is taken into account, but such allophony is unproblematic for the Dispersion-theoretic analysis developed here. Since (i) the release burst of a stop provides significant cues to place of articulation, (ii) voiced stops have quieter bursts than voiceless stops, and (iii) voiced continuants have no audible burst, it follows that a [dl-gl] or [[eth]1-yl] contrast is less perceptible than a [tl-kl] contrast. This difference in perceptibility is captured by a universal ranking of Space constraints penalizing coronal-velar contrasts in different contexts, with permissible clusters determined by dialect-specific rankings of Max(place) constraints along the hierarchy. Because this approach does not require reference to continuancy specifications, it does not face the same problems as an OCP-based account of Spanish /dl/. Additionally, the proposed hierarchy makes typological predictions that extend beyond Spanish, which are confirmed on the basis of data from Tobin's (2002) survey of initial clusters in forty different language varieties. The significant generalization is that in a given language, the existence of a [dl-gl] contrast entails the existence of a [tl-kl] contrast, but not vice-versa. Finally, /tl/ often neutralizes to /kl/, as evidenced in early Romance sound change and in Mexican Spanish loanword adaptations from Nahuatl, although in some cases /tl/ is maintained or varies with /kl/. These patterns are accounted for by different rankings between faithfulness and systemic markedness constraints, lending further support to an analysis in terms of coronal-velar place neutralization.
This paper is organized as follows. Section 2 examines onset cluster data that are problematic for an OCP account and develops an analysis in terms of systemic markedness and Max(place) constraints. Section 3 further explores the inadequacies of an OCP account, with additional data from Spanish dialects. Section 4 examines the typological predictions of the novel Dispersion-theoretic analysis, while Section 5 extends the analysis to early Romance sound change and to Mexican Spanish loanword phonology. Section 6 summarizes the analysis and provides some concluding remarks on the issue of universal place markedness.
2. Perceptual Distinctiveness and Coronal-Velar Neutralization Before Laterals. There are at least two languages in which an OCP analysis cannot explain the distribution of coronal and velar stops before laterals in complex onset clusters. First, consider the Mong Njua data in 2, taken from Lyman (1974:30-32, 38). Coronal and velar stops vary noncontrastively before laterals in 2a, but no such variation is found involving the labials in 2b. Coronal and velar stops are contrastive before vowels, as shown in 2c.
(2) a. [kl[??]-tl[??]] 'dog' [[k.sup.h]la ~ [t.sup.h]la] 'to run; jump' [[??]glua ~ ndlua] 'flash (as of lightning)' [[??][k.sup.h]l ~ n[t.sup.h]l] (no examples) b. [pla[??] 'stomach' [[p.sup.h]lu] 'cheek; face's [mblo[??]] 'leaf' [m[p.sup.h]lai] '(finger-)ring' c. [t[??]u] 'to dam up water' [k[??]u] 'barking-deer'
Second, the Katu data in 3 show that coronal- and velar-lateral clusters do not coexist in the same dialect (Wallace 1969:72). Velar-lateral clusters in the Andiem dialect correspond to coronal-lateral clusters in the Phuhoa dialect. The former dialect lacks /tl,dl/, while the latter does not have /kl, gl/.
(3) a. [kl[??]m] 'urinate' [gluh] 'go out' (An[??]di[??][??]m) b. [tl[??]m] [dluh] (Phuhoa)
The OCP cannot account for the data from Mong Njua in 2a because coronal stoplateral clusters actually do occur, in free variation with velar stop-lateral clusters. Like English 1b, the Andiem dialect of Katu (3a) has velar stop-lateral clusters but excludes clusters with an initial coronal stop. However, the OCP cannot explain the Phuhoa data in 3b, in which coronal stop-lateral clusters appear to the exclusion of velar stop-lateral clusters.
As Flemming (1995, 2002) argues, the data from Mong Njua and Katu suggest that indistinct coronal-velar place contrasts are problematic rather than the coronal-lateral sequences themselves. The formant transitions and release burst of an oral stop provide significant cues to place of articulation (Lieberman and Blumstein 1988, Olive et al. 1993, Stevens 1998). In oral stops that are released into a following lateral, coronal and velar constrictions are produced at or behind the lateral constriction. As a result, these stops are not well differentiated by their release bursts when coarticulated with a following lateral (Kawasaki 1982). The release bursts of labial stops are not affected by coarticulation in the same way because the labial constriction is formed in front of the lateral constriction. Flemming (2002:134) captures the similarity of coronal-lateral and velar-lateral clusters in terms of the auditory specifications shown in 4:
(4) [pl] [tl] [kl] transitions: F2 1 3 3 burst: diffuse + - - noise frequency 2 4 4
In Flemming's system, auditory dimensions such as formant transitions and noise frequency are assumed to be multi-valued features, represented on a numerical scale. Diffuseness of stop bursts, however, is treated as a binary feature since quantifying degrees of 'peakedness' versus flatness of fricative spectra appears to be unmotivated (Flemming 2002:20). According to the auditory representations in 4, coronal and velar stops that are released into a following lateral have identical specifications for F2 transitions and for diffuseness and noise frequency of the...
|
