Distinctive [voice] Does Not Imply Regressive Assimilation (with Pétur Helgason)

In a recent paper, van Rooy and and Wissing (2001) distinguish between what they call the "broad interpretation" and the "narrow interpretation" of the feature [voice]. According to the broad interpretation (Lisker & Abramson 1964, Kingston & Diehl 1994), languages with a two way [voice] contrast may implement this contrast phonetically with any two of the following: voice onset precedes plosive release, voice onset immediately follows plosive release, voice onset substantially lags behind plosive release. According to the narrow interpretation (Jakobson 1949:389, Keating 1990; Iverson & Salmons 1995; Jessen 1989, 1998; Jessen & Ringen 2002) [voice] is employed only when actual vocal fold vibration is present. They suggest that languages with the distinctive [voice], on the narrow interpretation, always have regressive voice assimilation.

The purpose of this paper is to show that Swedish employs the feature [voice] on the narrow interpretation: voice onset precedes plosive release in utterance initial position, (voiced) plosives are produced with vocal fold vibration intervocalically and word finally, but there is no regressive assimilation of [voice].

Six native speakers of Central Standard Swedish were recorded in a sound-treated room at Stockholm University. The speakers read a list of words containing stops from both stop series of Swedish, referred to here as fortis vs. lenis stops. The stops occurred in word-initial position, in intervocalic position and in word-final position, as well as in word-medial and -final clusters. The vast majority of instances of word-initial lenis stops exhibited prevoicing. All intervocalic lenes were voiced and practically all word-final lenes were voiced as well.

Among the test words were two suffixes, one /-t/ and one /-d/e.

köpa 'buy' kö/p+t/ > kö[pt] 'bought (ppart)' kö/p+d/e > kö[pt]e 'bought (past)'

väga 'weigh' vä/g+t/ > vä[kt] 'weighed (ppart)' vä/g+d/e > vä[gd]e 'weighed (past)'

Our data indicate clearly that in underlying mixed voice/voiceless clusters, [-voice] is dominant and the cluster is voiceless on the surface. Hence the claim that languages with narrowly defined voiced plosives exhibit regressive assimilation of voice is incorrect. In Swedish both progressive and regressive assimilation to [-voice] is found.

We present an OT account of the Swedish data which involves both features [±voice] and [±spread glottis].

The Analysis of Geminates: The Evidence from Sinhala

(with Robert Vago)

True geminates have been analyzed by Hayes (1989) as inherently heavy, containing a mora bearing single root node, and by Selkirk (1990) as inherently long, containing double root nodes. Most subsequent researchers have adopted the syllabic weight treatment. The controversy continues, however, with Hume et al. (1997), Davis (1999, 2002), Muller (2001), Ringen and Vago (2002a, b) and Arvanti and Rose (2003). The weight analysis requires that geminates be heavy with respect to weight sensitive processes, and short with respect to quantity sensitive processes. Both claims are problematic as shown by Ringen and Vago (2002a, b). In contrast, analyzing geminates in terms of double units on the length tier leaves open their role in syllabification and weight assignment. This leads to the prediction that geminates should exhibit all possibilities with respect to syllable weight. This is the case: a geminate consonant may be moraic or nonmoraic, in coda or onset.

In this paper we show that, contrary to Davis' (2002) claim that Sinhala provides strong evidence in favor of the moraic treatment of geminates, there is a straightforward two root analysis of geminates in Sinhala. Consider the Sinhala forms: [mal-ə] sg., [mal] pl. 'flower'; [mænik-ə] sg., [mænik] pl., 'gem'; [mull-ə], sg., [mulu] pl., 'corner'; [pætt-ə] sg., [pæti], pl.'area'. Davis bases one of his arguments for the moraic treatment of geminates on the fact that an epenthetic vowel shows up in the plurals of nouns that have a geminate in the singular, i.e. that have roots ending in geminates ([mull] and [pætt]). He takes this to be the reflex of the underlying mora of the consonant. However, there is a straightforward two root account. In the two root theory, the crucial difference between a geminate and non geminate is the number of root nodes in the input, rather than the presence or absence of a mora. In our analysis, the well-known constraint against deletion of root nodes (MAX) is high-ranked to account for the fact that degemination does not result in deletion of a root node. As in Davis' account, we assume Dep (no epenthesis), *Gem# (no word-final geminates), *V: (no long vowels), Dep [±bk] (don't insert [±bk]). We assume a low-ranked faithfulness to the feature [consonant], the constraint *Cmplx (no complex codas). The apparent epenthetic vowel results because the second root node of the geminate is not deleted to satisfy *Gem#. Superscript R's indicate the number of root nodes a segment is associated with. (We assume with Davis that the backness of the epenthetic vowel results from spreading of the backness of the root vowel.)

/mul^RR/	Dep	MAX	*Gem#	Dep [±bk]	*Cmplx	*V:	ID-IO Cons
a. mul^RRu^R [mullu]	*!
b. mul^R [mul]		*!
c. mul^RR [mull]			*!		*
d. mul^Ri				*!			*
e. mul^R w^R					*!		*
f. mu^RRl^R [muul]						*!	*
g. Fmul^Ru							*

We conclude that the Sinhala facts add to the body of evidence that lend support to the claim of Ringen and Vago (2002a, b) that all descriptions of single root, moraic geminates are reanalyzable as double root representations.

Geminates: Heavy or Long?

(with Robert Vago)

True geminates have been analyzed by Hayes 1989 as inherently heavy, containing a mora bearing single root node, and by Selkirk 1990 as inherently long, containing double root nodes. In this paper we present a typological investigation of geminates, bring together arguments against the prevailing inherent weight analysis, and motivate the double root (alternatively, skeletal) representation as the only universal property of geminates. The weight analysis as a universal position requires that geminates be heavy with respect to weight sensitive processes and short with respect to quantity sensitive processes. Both claims are problematical, as we will show. In contrast, analyzing geminates in terms of double units on the length tier leaves open their role in syllabification and weight. Indeed, this is the case: Geminates are found in both moraic and nonmoraic flavors, in both coda and onset positions, as we will show. All descriptions of single root, moraic geminates known to us are reanalyzable into double root representations, augmented by language specific weight assignment.

Voice and Aspiration in Austrian German Plosives

(with Sylvia Moosmüller)

It is well-known that so-called "voiced" plosives in German, including Austrian German, are voiceless except between vowels where they are (sometimes) voiced (i.e. have vocal fold vibration during closure). Nonetheless, in the phonological literature, the contrast is often treated as one of [±voice]. This leaves a rather substantial mismatch between the phonological description and the phonetic facts. Jessen and Ringen (2002) have recently presented experimental evidence in support of the position that the contrast in northern Standard German (NG) is one of [±spread glottis]. It is often suggested that in Austrian German there is a two-way contrast of plosives, but no aspiration. This raises a question about whether the contrast in Austrian Standard German (AG) can possibly be one of [spread glottis] vs. non-[spread glottis]. This paper investigates this question. We present experimental results and argue that AG, like NG the appropriate feature of contrast is [spread glottis].

Voice and Aspiration: Evidence from Russian, Hungarian, German, Swedish, and Turkish

(with O. Petrova, R. Plapp, S. Szentgyörgyi)

The purpose of this paper is to investigate a variety of languages with laryngeal contrasts that have usually been characterized in the literature of generative phonology as having a two-way [voice] contrast and to show that by adopting a narrower interpretation of [voice] to cover only those languages which exhibit prevoicing in word-initial stops, a better understanding of the laryngeal contrasts and assimilation of laryngeal features in these languages is possible. We consider Hungarian, Russian, German, Swedish, and Turkish, which have all been analyzed as having a two-way [voice] contrast for stops. We suggest that the feature [voice] is indeed appropriate for Hungarian and Russian, that the feature of contrast in German is [spread] and that, in Swedish and Turkish, both [voice] and [spread] occur in underlying forms.  Analyses are provided for these stop systems in the framework of Optimality Theory.