This is a project initiated by Erik Jan van der Torre and Jeroen van de Weijer (ULCL, Leiden University), which aims at bringing together a number of phonological and other studies which focus on the [voice] grammar of Dutch.
Contributors | Abstracts | Indexes
Petra van Alphen (Max Planck Institute for Psycholinguistics): Prevoicing variation in Dutch initial voiced plosives (abstract)
Mirjam Ernestus and Harald Baayen (Max Planck Institute for Psycholinguistics): Intraparadigmatic analogical effects on Final Devoicing (abstract) (this paper is also available as a PDF file; Figure 1 as PDF)
Suzanne van der Feest (University of Nijmegen), Paula Fikkert (University of Nijmegen), René Kager (University of Utrecht), Annemarie Kerkhoff (University of Utrecht) and Tania Zamuner (University of Nijmegen): Consequences of voicing acquisition for the representation of laryngeal features (abstract (PDF))
Wouter Jansen (University of Groningen): Dutch regressive voicing assimilation as a symmetric coarticulation process: acoustic evidence (abstract (PDF))
Jan Kooij (Leiden University): Intervocalic voicing in Dutch (abstract)
Marc van Oostendorp (Meertens Institute): An exception to Final Devoicing (abstract) (the draft of this paper is also available as a PDF file)
W. Leo Wetzels (Free University of Amsterdam): Universal aspects of sandhi voicing in Limburg Dutch (abstract)
Wim Zonneveld (University of Utrecht): Voice(less) in Dutch: regressive, progressive and so on (abstract)
Petra van Alphen (Max Planck Institute for Psycholinguistics): Prevoicing variation in Dutch initial voiced plosives
It is generally assumed that voiced plosives in Dutch are produced with vocal cord vibration during the closure of the plosives (referred to as prevoicing). A production study on the voicing distinction in Dutch initial plosives indicated, however, that 25% of the voiced tokens in the sample were produced without prevoicing. Several factors which affected the vocal tract volume influenced the proportion of prevoiced tokens, suggesting that prevoicing was absent due to aerodynamic limitations. Detailed acoustical analyses identified several acoustic correlates of voicing, of which prevoicing appeared to be by far the best predictor. Perceptual classification of these tokens showed that prevoicing was indeed the strongest cue to the voicing distinction, even though prevoicing was frequently absent in voiced plosives. Some of these tokens were still classified as voiced on the basis of secondary cues. These secondary cues were different for the two places of articulation (labial and alveolar).
The question then arises how the word recognition system deals with natural variation in prevoicing. A series of perception experiments (rating, discrimination, identification and priming experiments) were conducted to investigate the influence of prevoicing variation on lexical access. The results indicated that quantitative prevoicing variation in words (12 vs. 6 periods of prevoicing on their initial voiced plosive) did not affect the degree of activation of lexical candidates with voiced plosives. Qualitative variation in prevoicing, however, did have an effect: Words without prevoicing activated both voiced and voiceless word candidates, while items with prevoicing only activated voiced candidates. The presence versus absence of prevoicing in Dutch voiced plosives therefore appears to affect lexical access, but does not hinder the recognition of words starting with voiced plosives.
Mirjam Ernestus and Harald Baayen (Max Planck Institute for Psycholinguistics): Intraparadigmatic analogical effects on Final Devoicing
We will discuss two experiments documenting intraparadigmatic analogy for the interpretation of word-final obstruents in spoken Dutch. The first experiment investigated listeners' rating of voicing for word-final obstruents on a 5-point scale. Listeners heard either a complete word (condition C) or the final part (condition P) of that word. In condition C, a final underlyingly voiced obstruent was rated as more voiced than in condition P. For underlyingly voiceless obstruents, the reverse pattern was observed. The second experiment revealed longer response latencies for voiced final obstruents in words with underlyingly voiceless obstruents compared to words with underlyingly voiced obstruents. The perception and interpretation of obstruents is codetermined lexically. Given the results of Ernestus & Baayen (to appear), we interpret these data in terms of intraparadigmatical analogy arising from stored inflectional exemplars.
Ernestus & Baayen (to appear), Predicting the unpredictable: interpreting neutralized segments in Dutch. Language.
Wouter Jansen (University of Groningen): Dutch regressive voicing assimilation as a symmetric coarticulation process: acoustic evidence (abstract (PDF))
See also: Wouter Jansen's home page.
Jan Kooij (Leiden University): Intervocalic voicing in Dutch
Inflected words in Dutch show a regularity with respect to
the voicing of intervocalic fricatives that can be summarized as follows:
(i)
After long vowels and diphthongs, intervocalic fricatives are voiced,
after short vowels, intervocalic fricatives are voiceless.
Examples: graven ‘grave Pl’, vieze, ‘dirty Infl’, and dragen, ‘carry Infin’ vs stoffen, ‘substances Pl’, losse, ‘loose Infl’, and lachen, ‘laugh Infin’. In a large part of the language area, the distinction is tenuous with velars, but verbal stems in [x], written ch, still take the past tense suffix -te while stems in [γ], written g, take -de. Exceptions are a number of words with a voiceless fricative, in particular [s], after a long vowel, e.g. hese, ‘husky Infl’, and eisen ‘demand Inf’. Inflected forms with a voiced fricative after a short vowel hardly exist, however. Simplex native words pattern in the same way, but there are exceptions like tafel, ‘table’ and Pasen, ‘Easter’. In the pronunciation of French loanwords, the same pattern is followed: faseren with [z] – though written with s - vs passeren with [s]. That (i) is a true generalization is also apparent from the fact that some inflected forms which have intervocalic [s] after a long vowel can also be pronounced with [z], e.g. kruisen / kruizen, ‘cross Pl’, whereas words with intervocalic [z] do not seem to have alternative pronunciations with [s]: huizen / *huisen, ‘house Pl’.
On closer look, however, the picture becomes more complicated since there exist a number of loans, names, and recently coined words that have voiceless labial fricatives after long vowels: sofa,’ id’, Sofia, sofi-nummer, ‘social security number’, trafo, ‘transformer’. Similar formations with coronal fricatives are commonly pronounced with [z]: NASA, Lisa, Stasi. Thus, it appears that (a) the generalization in (i) is only true for inflected forms and to that extent is morphological rather than phonological, and that (b) it particularly applies to coronal fricatives, which corroborates the claim that coronals are a special category and more sensitive to the phonetic context. I will discuss these conclusions in more detail in the paper, and include some recent findings in the experimental literature on the voiced-voiceless distinction.
Marc van Oostendorp (Meertens Institute): An exception to Final Devoicing
At first sight, Final Devoicing seems a fairly stable process across the West Germanic dialect continuum: virtually all dialects devoice obstruents at the end of the syllable. On closer inspection, however, we notice that there is some variation as to the exact positions in which Final Devoicing applies. Most notably, we find dialects in which some forms are exceptions to final devoicing: words which end in voiced obstruents. In most cases, these are either plural forms of nouns, or first person singulars of verbs.
Historically, it could be argued that in these cases, an inflectional schwa has been dropped and that final devoicing has 'not yet' applied to the resulting syllable- and word-final obstruents. The problem for constraint-based phonology is that the synchronic process is opaque: it is not clear why the final obstruent does not devoice in the 1SG, while it does in the 2SG form (in most dialects of Dutch, 1SG and 2SG in inversion have exactly the same phonological shape).
In this paper, I first systematically present the interesting data from some of the (North-Hollandic) dialects which show the phenomenon mentioned here, which to my knowledge has been ignored in the international phonological literature on Final Devoicing. I then continue to give an analysis of this phenomenon, using (a) an analysis of phonotactics which posits that syllable rhymes cannot contain more than two segmental positions, (b) a theory of segmental structure which explains why this phenomenon mainly involves fricatives, (c) Optimality Theory in order to explain the variation involved. It is argued that the relevant cases, such as the 1SG form, have a different syllable structure from the 2SG; the reason for this is that the syllable structure of the 1SG still wants to mirror the complex morphological structure of this form, in spite of the fact that the 1SG-suffix is no longer visible as a schwa, but has rather turned into a null morpheme.
W. Leo Wetzels (Free University of Amsterdam): Universal aspects of sandhi voicing in Limburg Dutch
The rules of voicing assimilation in Limburg Dutch are in many respects similar to those of standard Dutch. For example, non-sonorant clusters of which the rightmost consonant is a fricative are entirely voiceless and, when the rightmost consonant is a stop, it imposes its voice value upon the entire cluster. One difference concerns past tense verb forms which, unlike in standard Dutch, obey the general rule of regressive assimilation, in which the [+voice] feature of the suffix-initial coronal stop, and spreads to a preceding non-sonorant consonant. The main theme of our discussion will be devoted to another difference between the two variants, which is the existence of an utterance-level process of sandhi voicing in Limburg Dutch by which resyllabified word-final codas are voiced. The process of sandhi voicing will be studied from a cross-linguistic perspective. More in particular, it will be argued that there is a universal implication that restricts the occurrence of sandhi voicing to languages that have either no underlying voice contrast or that have a rule of word-final voice neutralisation.
Wim Zonneveld (University of Utrecht): Voice(less) in Dutch: regressive, progressive and so on
Recently, research into Dutch Voicing Assimilation has received a new boost by attempts to capture its central properties in an Optimality Theory framework. The history of the key literature appears to be this:
pre-OT:
- Cho (1990/99) fits Dutch regressive assimilation into a comprehensive
parametric framework.
- Lombardi (1991), partly referring back to Cho, fits Dutch re- and progressive
assimilation into a revised parametric framework, using a `privative' feature
[voice], after Mester & Ito (1989).
OT:
- Lombardi (1996) typologically analyses regressive assimilation languages,
leaving progressive assimilation (in a variety of languages incl. Dutch) as a
`marked' residue.
- Lombardi (1999) similar to, although more elaborate than, 1996 (a serious
attempt to deal with progressive assimilation in Dutch has disappeared from the
survey).
- Lombardi (2001), a slightly revised version of a 1995 paper, presents the core
typological analysis (of basically regressive assimilation only) of the 1996 and
1999 papers.
- Grijzenhout & Krämer (1999) is an attempt at a comprehensive OT account
of Dutch re- and progressive assimilation.
post-OT:
- Wetzels & Mascaró (2001) provide a catalogue of difficulties for the
Lombardi framework, incl. remarks based on Dutch (with respect to: privative
feature voice, lexical vs. post-lexical status of the privative feature, slots
in the factorial typology of the OT constraint set, metaconstraints on
constraints, ways of dealing with Dutch, and so on).
In spite of all this (well-deserved) attention, none of these papers, not even those (co-)written by native speakers, appears to get a complete grip on the Dutch assimilation facts, especially the facts of progressive assimilation. This paper tries to sort out again the basic issues, and the basic and not-so-basic data (incl. data that, if they survive current scrutiny, are new to the phenomenon's empirical range.)
References
Cho, Y.Y. (1990/99) Parameters of Consonantal Assimilation, Ph.D. thesis,
Stanford University, revised version published as Lincom Studies in
Theoretical
Linguistics 15, Munich.
Grijzenhout, J., and M. Krämer (1999) Final Devoicing and Voicing Assimilation
in Dutch derivation and cliticization. Unpubl. paper,
Heinrich-Heine-Universität, Düsseldorf.
Lombardi, L. (1991) Laryngeal Features and Laryngeal Neutralization. Ph.D.
thesis, University of Massachusetts, Amherst.
Lombardi, L. (1996) Restrictions on the direction of voicing assimilation: an
OT
account. University of Maryland Working Papers in Linguistics 4, 88-102.
Lombardi, L. (1999) Positional Faithfulness and the Phonology of Voicing in
Optimality Theory. Natural Language and Linguistic Theory 17, 267-302.
Lombardi, L. (2001) Why Place and Voice are different: constraint-specific
alternations in Optimality Theory. In L. Lombardi (ed.), Segmental
Phonology
in Optimality Theory. Constraints and Representations, 13-45.
Cambridge University Press.
Mester, R.A., and J. Ito (1989) Feature predictability and underspecification:
Palatal prosody in Japanese mimetics. Language 65, 258-294.
Wetzels, W.L., and J. Mascaró (2001) The typology of voicing and devoicing.
Language
77, 207-244.
Here are the preliminary author, language and subject indexes for the book that will come out of this project. Comments welcome!
Last update: 29-12-03