Multiple Sonority Thresholds
Draga Zec
Among the constraints on syllable structure proposed by Prince and Smolensky 1993, a special role is played by those that impose a sonority threshold on the class of segments which, in a given language, may serve as syllable nuclei. In this paper, I will propose, first, that the sonority constraint on the nucleus is formally associated with the syllable in its role of a prosodic constituent, and second, that the prosodic constituents other than the syllable, the mora and the foot, exhibit a parallel behavior by imposing their own sonority thresholds. Given the prosodic hierarchy in (1) (cf. Selkirk 1979, Nespor and Vogel 1986, McCarthy and Prince 1986 among others), each constituent within the hierarchy establishes a direct relation with the level of segments by imposing a minimal sonority threshold. This is implemented by positing the SON(ority) family of constraints, which govern the sonority relations within the prosodic hierarchy, as shown in (2) (taking Optimality Theory as a frame of reference):
| (1) | w
f s m |
prosodic word
foot syllable mora |
(2) |
f
|
SON-f
SON-s SON-m s4 |
Assuming that each prosodic constituent possesses a head, corresponding to one constituent at the next lower level (marked by the h subscript), the SON constraint at each prosodic level imposes a sonority threshold on the segment at the bottom of its "head" path: SON-m, SON-s and SON-f are responsible for the minimal sonority threshold imposed, respectively, by the mora, the syllable, and the foot.
I will present two cases which provide evidence for positing three levels of sonority threshold constraints. In English, SON-m admits all segments as weight-bearing; while SON-s[+son] restricts syllabic segments to vowels, liquids (butter, murky; bottle), and nasals (sudden, sedentary). Moreover, operative at the foot level is SON-f[-cons]. Evidence for this comes from the syllables with l or a nasal in the nucleus (CL(C)), which exhibit a highly restricted distribution: CL and CLC syllables are never stressed; and there are no monosyllabic CLC words, or disyllabic CLCL words. By contrast, syllables with r in the nucleus (CR(C)), share the distributional properties of CV(C) syllables. CR and CRC may be stressed, as in murky, covert, furnace; and may occur in monosyllables, as in fur; turf, curl, terse, or in disyllables, as in curtain, curdle. The SON-f[-cons] picks out the right subset of foot-bearing syllable nuclei (following KahnÕs 1978 proposal that r is a [-cons] segment).
The second case is both more complex and more intriguing. Old Church Slavonic (OCS), as characterized in Bethin 1998, Lunt 1959 and Shevelov 1965, allows only vowels and liquids as moraic segments, yielding SONm[+son, -nasal]; and only vowels as syllable-bearing segments, yielding SONs[-con]. That the sonority threshold constraint imposed by the foot is also operative is demonstrated by the behavior of the so-called jer vowels. The jers, I and U, are high, [-tense] vowels, which exhibit a characteristic pattern of lowering. As shown in (3), word-final jers, and jers in syllables followed by vowels other than jers retain their quality, while jers followed by another jer become mid (I and U merge with e and o respectively).
(3)
| underlying | surface | gloss | underlying | surface | gloss |
| a. stolU | stolU | ÔthroneÕ (Nom) | d. otIca | otIca | ÔfatherÕ (Gen) |
| b. sUnU | sonU | ÔdreamÕ (Nom) | e. sUnIma | sonIma | ÔgatheringÕ (Gen) |
| c. dInI | denI | ÔdayÕ (Nom) | f.. sUnImU | sUnemU | ÔgatheringÕ(Nom) |
An obvious analysis is that consecutive syllables headed by jers form a foot. The crucial cases are (3)e and (3)f, footed as (sUnI)ma and sU(nImU) respectively. Due to the agency of SON-f[-cons, +tense], a jer may not head a foot. Since this is a trochaic system, the leftmost jer in a foot lowers to the closest [+tense] vowel, yielding in this case (sonI)ma and sU(nemU).
It is important to note that the three SON constraints exhibit a stable relation: SON-s is either equally or more restrictive than SON-m, and SON-f is either equally or more restrictive than SON-s. In other words, hypothetical cases such * SON-m[-cons], SON-s[+son], SON-f[-cons] or * SON-m [-cons], SON-s[-cons], SON-f[+ son] need to be excluded on principled grounds, yet constraint ranking alone cannot achieve this. Forthis purpose, I invoke the mechanism of local conjunction (Smolensky 1995). At the higher levels of the prosodic hierarchy, segments in the head position are evaluated, also, by constraints resulting from conjunction: by SON-m & SON-s at the level of the syllable, and by SON-m & SON-s & SON-f at the level of the foot, with the proviso that, for a conjoined constraint to be satisfied, each of the conjuncts has to be satisfied (following Crowhurst and Hewitt 1996). Curcially, the conjoined constraints ensure only minimal violations in cases of threshold lowering: a segment violating SON-f will still comply with SON-s, and a segment violating SON-s will still comply with SON-m.