Stresw&schwa in Faetar
[ Home ] [ Up ] [ Computational / computer index ] [ diss ] [ DOUBLE MARQUAGE DU SUJET DANS LE FRANÇAIS PARLÉ ] [ Endangered language fieldwork, SECOL 1999 ] [ Faetar grammar ] [ Stresw&schwa in Faetar ] [ diacritics OPHS ] [ NHVTresearch  proposal ] [ CLS paper ] [ SUBJECT PRONOUN VARIATION ] [ Yankee Doodles in Dialectography: Updating New England ] [ SURF FINAL REPORT DRAFT___8/12/97 ] [ Morphology index ]

Stress and schwa in Faetar

Naomi Nagy

Abstract

This paper explores three possible accounts of the prosodic structure of Faetar, a Francoprovençal Apulian dialect: the (lexically) Marked Stress Hypothesis, Marked Vowel Quantity Hypothesis, and the Marked Vowel Quality Hypothesis. Unlike Italian, Faetar's phonology includes schwa, but the phonemic status of schwa has not previously been examined. Several arguments show that Faetar is a language in which stress is lexically marked and schwa is not a phoneme, but rather a reflex of some other vowel(s) in unstressed position. Support for the Marked Stress Hypothesis (and against the other hypotheses) includes: vowel alternations in verb paradigms, the existence of raddoppiamento sintattico, the retention of historical stress assignment, vowel/schwa allophonic alternation in pretonic position, the categorical absence of non-schwa vowels in post-tonic position, and the restriction of stress to the three final syllables.

Acknowledgements

1. Introduction.

The first, the Marked Stress Hypothesis, is that primary word stress is marked in the lexicon. In that case, vowel quality (full vs. reduced vowel) is predictable: tonic vowels are never schwa, post-tonic vowels categorically reduce to schwa, and pre-tonic vowels (variably) reduce to schwa. Under this analysis, schwa is not phonemic.

The second possibility is the Marked Quantity Hypothesis: that vowels are lexically marked as long or short (bi- or monomoraic). Then we can observe that stress appears on the rightmost long (bimoraic) vowel, and short (monomoraic) vowels may reduce to schwa. The lack of a moraic difference between vowel and schwa in Faetar will be shown by the existence of raddoppiamento sintattico in Faetar.

The Marked Quality Hypothesis, that vowel quality is lexically marked (schwa as a mid-central vowel vs. other vowels), is one other possible alternative. In that case, the location of word stress is fully predictable: it always appears on the rightmost non-schwa vowel. Schwa, then would be phonemically distinct from other vowels: the mid-central vowel would be a phoneme. The word-position distribution of vowels undermines this hypothesis: if schwa were a phoneme, why would it be the only vowel with the privilege of appearing post-tonically?

Like many nonstandard languages and dialects, Faetar has no systematic orthography, long-term historical record, or native speakers trained as linguists. This eliminates many of the tools often used by linguists to determine underlying structure. Therefore, other methods are required to choose among these three hypotheses. I present several arguments to support the first hypothesis, and contradict the latter two. Before presenting these arguments, I provide necessary background information about Faetar and the languages with which it is in contact.

2. Background phonological information.

2.1. Faetar consonants and vowels.

(1)

(adapted from Nagy 1996:119 and Valente 1975:12)

The phonemic inventory of Faetar consonants is shown in 2a. Additionally, Faetar has phonemic geminate consonants word-internally (Nagy 1994). There are contrastive geminates for all the consonants except /µ/, /Ò/, /v/ and the affricates.

(2) a.

(adapted from Nagy 1996:116 and Valente 1975:12)

2.2. Faetar stress.

(3) Possible stressed syllable positions

In Faetar, any vowel except schwa may be stressed, and any vowel may appear in an unstressed syllable, as shown in 4.

(4) Possible stressed and unstressed vowels

These facts can be accounted for in a straightforward manner. One binary foot, a generalized trochee (Kager 1993), is built at the right edge of the word. Some words have one extrametrical syllable at their right edge, but no more than one. Thus the possible structures for the right edge of Faetar words are:

(5)

5 shows only the assignment of vowels to syllable nuclei. The strong (left) branch of any foot may have a coda consonant, the weak branch may not. The limitations of one extrametrical syllable and a binary-branching foot restrict stress to one of the rightmost three syllables of the word.

Nagy & Reynolds (1997:44-5) show that the variable process of word-final deletion (of segments and syllables) is more likely to occur in Faetar words with antepenultimate stress than in words with penultimate or ultimate stress. This can now be explained as deletion of unfooted segments/syllables, a less costly deletion than for footed segments/syllables, supporting the analysis that extrametrical syllables exist in words with antepenultimate stress. Such extrametrical vowels reduce to schwa as an intermediary step in the deletion process. Similarly, pretonic vowels, all unfooted, may reduce to schwa.

An unassociated mora appears in the structure for (all?) ultimate stressed words, which satisfies the requirement that a trochee be binary-branching. The source of the unassociated mora may be from analogy with Italian, or it may exist for historical reasons. Faetar exhibits much variable word-final deletion, a synchronic symptom of the diachronic process of increasing apocope mentioned above. Final unassociated moras may be remnants of word-final segments present at an earlier stage.

2.3. Apulian phonology. While distinctions between Faetar and standard Italian have already been highlighted, the relevant comparison is to the local Apulian dialect, as this is the variety with which Faetar speakers have regular contact. The principal source used for this description of Apulian is Valente 1975, and his findings are confirmed by recordings of speakers from towns neighboring Faeto (Nagy 1996). Valente (1975:64) writes, 'Apulia represents a unified and continuous linguistic area that has its center of gravity and of dissemination in the region of Bari [my translation].' The most distinctive trait of the dialect is that Latin A in both stressed and final syllables is intact, that is, not reduced (ibid 54). All other unstressed vowels reduce to [\] (ibid 36). In Apulian, stress is always on one of the last three syllables (Valente 1975:29). From the data available, it seems to have the same stress system as standard Italian. The phonemic inventories of Apulian, are given above in 1b and 2b. Shading in these two figures indicates differences between the inventories of Apulian and Faetar.

2.4. Raddoppiamento sintattico. Like many southern Italian dialects, Faetar exhibits a process of word-initial consonant lengthening following a word-final stressed vowel, known as raddoppiamento sintattico (RS). I adopt Repetti's (1991:307) analysis of RS: 'raddoppiamento is caused by an empty mora which is present in the underlying representation (not added by a rule) ... [the mora] is now filled by spreading from the initial consonant of the following word.' If there is an empty mora that needs to be associated to an element in the following word, then the vowel in that syllable can be linked to only one mora to begin with. That is, the structure of a word-final stressed open syllable is that shown in 5c. As in Italian, this provides the necessary structure for RS to apply. In Section 3.3, acoustic evidence that RS does occur in Faetar is described. I make use of the presence of RS to illustrate that the Marked Stress Hypothesis is viable and the Marked Quantity and Marked Quality Hypotheses are not.

3. Empirical evidence of RS in Faetar.

3.1. Historical source for RS.

(6) RS in the neighboring towns and closest cities (Jaberg & Jud 1928-40)

The phrases in 7 give more recent evidence of lexically triggered RS in Apulian. They are extracted from narratives recorded in three southern locales and transcribed in Valente (1975:66-69). The list includes phrases which have contexts for RS, but includes some instances where the lengthening did not occur, or at least was not transcribed.

(7) RS patterns in near-by dialects (Valente 1975:66-69)

3.2. Design of experiment.

For this experiment, the durations of word-initial consonants in various contexts were examined in the speech of one representative speaker: a 29 year old woman recorded in 1994. 87 segments were measured: all of the clearly recorded instances of word-initial /m/ and /f/ in the storybook recording. After transcribing the speech in full, transcriptions were searched for all appearances of word-initial /m/ and /f/. A list of the phrases examined, and the relevant consonant duration for each, appears in the appendix. Recordings were digitized and the duration of the initial segments measured from on-screen spectrograms using Xwaves on a Unix Sparcstation in Mark Liberman's lab. The measurements were grouped according to the class of the preceding segment: consonant (C), full vowel (V), or schwa (\).

3.3. Results of experiment.

These data illustrate two points. The first is that RS occurs in Faetar. This is shown by the difference between length of word-initial consonants which follow RS-triggers (V) and those following non-triggers (C, [\]). The second is that V triggers RS and [\] does not, indicating a difference in prosodic structure between syllables with schwa and those with other vowels.

Historical evidence also supports lexically marked stress, as Francoprovençal, the source language of Faetar, has lexically marked stress. Although Francoprovençal requires that stress be on one of the last two syllables, there are minimal pairs like ['†ata] 'sing, 2nd sg. imperative' vs. [†a'ta] 'sing, 2nd pl. imperative' (Martin 1990:673). There is no way that stress could be predicted to appear on different syllables of the two members of this pair, as they are otherwise identical.

4.3. The Marked Quality Hypothesis. One alternative approach is that Faetar vowel quality is lexically marked. The vowel phonemes are then /i, u, e, o, a and \/. By rule or constraint, stress will be placed on the last syllable with a non-schwa vowel. While apparently descriptively adequate, such an approach cannot account for alternation in verb paradigms. Many paradigms contain pairs as listed in 10: some verb forms have a stressed full vowel in the last syllable of the root and other forms have an unstressed schwa in the same position (Giuliani 1995:147). Presumably, the underlying root forms of various elements of a paradigm are identical.

Second, if we assume that vowel quality is marked in the lexicon, there is no way to correctly assign stress by rule or constraint in Faetar, given the diversity of word types shown in 3. For example, if we take a common Faetar word, ['brok\l\] 'fork', there is no way to predict that it is pronounced ['brok\l\] rather than [br\k\'le] or [br\'kel\]. A regular right-to-left trochee-forming grammar would produce the latter form, rather than the appropriate one, if schwa were a phoneme.

————. 1994. Language Contact: Italian (?) Geminates in Faetar. Belgian Journal of Linguistics 9.111-128.

———— & Bill Reynolds. 1997. Optimality theory and variable word-final deletion in Faetar. Language Variation and Change 9.1.37-56.