LINGUIST List 10.930

Wed Jun 16 1999

Review: Waals: Dutch Syllables

Editor for this issue: Andrew Carnie <carnielinguistlist.org>




What follows is another discussion note contributed to our Book Discussion Forum. We expect these discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for discussion." (This means that the publisher has sent us a review copy.) Then contact Andrew Carnie at carnielinguistlist.org

Directory

  • Niels Schiller, review: Waals

    Message 1: review: Waals

    Date: Tue, 8 Jun 1999 23:55:45 -0400 (EDT)
    From: Niels Schiller <nschillewjh.harvard.edu>
    Subject: review: Waals


    Waals, Juliette (1999), "An Experimental View of the Dutch Syllable", Holland Academic Graphics, The Hague (UiL OTS/Utrecht University Dissertation). [LOT International Series, 18.] Januari 1999. ISBN 90-5569-063-5. Paperback. xii+158 pp. Price: NLG 48.00 (excl. P&P, VAT). Discounts are available for individuals ordering directly from HAG. <www.hagpub.com>

    Reviewed by Niels Schiller, Harvard University

    Synopsis

    This thesis is a phonetic investigation of syllable structure in Dutch. It provides empirical data about the durational characteristics of single consonants as well as consonant clusters and vowels. The thesis is divided into three empirical parts, dedicated to the word onset (Chapter 2), the word coda (Chapter 3), and the intervocalic position (Chapter 4). Chapter 5 provides additional data about the role of lexical stress in syllabification. The results are summarized in Chapter 6. The thesis comprises a list of references, an index, and a Dutch summary.

    Critical evaluation

    The aim of this thesis is the empirical investigation of the relation between segment durations and syllable structure in Dutch. It generally well written although the largely descriptive form of the dissertation -- including a lot of data -- makes it sometimes a bit hard to read. The general method used by the author to obtain empirical data was to ask native speakers of Dutch to produce specific words and pseudowords including the target segments. The production of the material was recorded under laboratory conditions and the acoustic durations of the consonants and consonant clusters were measured. The motivation of this enterprise resides in the question whether it is necessary to refer to syllabic structure when interpreting durational data. The central hypothesis put forward in this thesis is the Metrical Segment Duration Hypothesis (MSDH) and states that "the relative durations of segments are a reflection of [phonological] syllable constituency". Analyses of the data support this hypothesis and show that syllables and syllable structure have to be taken into account to be able to make sense of the segment durations in the different syllabic positions.

    Chapter 1 is an introduction including a brief description of the theoretical background and a short review of the literature about the relation of segment length and syllable structure. Although articulatory data are also mentioned, the literature review focuses on acoustic studies of segment durations with emphasis on what has been found in Dutch. Also, the central hypothesis of the thesis is introduced and discussed before the background of the relevant findings reported in the literature.

    Chapter 2 reports empirical data of segmental durations in word- onset position. A brief description of the Dutch phonology is provided including principles of phonotactics and syllabification. The materials consist of monosyllabic Dutch words including singleton onsets as well as bi- and triconsonantal clusters. One native speaker produced the set of 151 words embedded in a carrier phrase two times.The results showed that the individual consonants differed significantly in length, but there was no effect of repetition or (phonological) length of the following vowel (vowel length is phonologically distinctive in Dutch). Waals found that obstruents were longer than nasals, which in turn were longer than liquids. Within the class of obstruents, voiceless members were longer than voiced ones, and fricatives were longer than plosives.

    For the clusters, so-called "compression rates" were computed, which are nothing else but the duration of a segment in an onset cluster relative to its duration as a singleton onset (expressed as a proportion). The reader of Waal's thesis will be surprised, however, that the "compression rates" reported for biconsonantal clusters in the thesis slightly deviate from the values that are obtained when the reader makes the same computation based on the values provided for the individual consonants. For example, /p/ had a duration of 154 ms as a singleton onset and 147 ms in the biconsonantal cluster /pl/. The reported compression value for /p/ in the cluster (94%) deviates slightly from the correct value computed on the basis of the data provided by Waals (9 5%). This is also true for other cases. Although this does probably not affect any of the conclusions drawn in this thesis, it is nevertheless confusing. Due to the possibility of word-initial /s/ being extrasyllabic, biconsonantal clusters were divided into two sets, i.e., obstruent+liquid clusters and /s/+consonant clusters. In the first type of cluster, the length of the obstruent was virtually the same as in singleton onsets, but the duration of the liquid was only half the duration as in single onset position. In the second type of cluster, both consonants were shorter than singleton onsets. The observations that were made for biconsonantal clusters were generally also true for triconsonantal clusters. The addition of a liquid to a biconsonantal /s/+obtruent cluster did not lead to additional compression and the duration of the liquid was the same in bi- and triconsonantal clusters. Therefore, the duration of triconsonantal clusters was completely compositional. The most important observation is that the duration of a cluster was not the sum of the durations of the individual consonants. The durations of individual segments in a cluster were shortened relative to their durations as singleton onsets. However, the shortening did not occur arbitrarily. The general cluster shortening effect was only applicable to immediately neighboring consonants, i.e., the shortening effect was local. Locality ensures that a consonant is shortened only by its direct neighbor (see triconsonantal clusters). Based on the empirical findings, the author constructed a set of duration rules. These rules were used to compute duration data. It is not clear to me, however, why the "rule data" was compared to the "human data" afterwards, when rules were constructed on the basis of the data from that same speaker. This seems circular to me (but maybe I am missing something here), and therefore it was no surprise to me that the "human data" and the "rule data" correlated very well (r = .94).

    Chapter 3 investigates consonants and consonant clusters in the word coda. Dutch phonology requires syllable rhymes to consist of at least two positions or slots. This is called the "VX condition" in this dissertation, and it states that syllable rhymes either have to consist of a long vowel (VV) or of a short vowel and a consonant (VC). Furthermore, the rhyme of the final syllable of a word may have an extrasyllabic appendix, i.e., additional consonants may be added to the final coda of a word.

    In addition to the durational questions addressed in the word onset position, relations between the nucleus and the coda were investigated. Altogether, 9 singleton codas, 19 biconsonantal clusters, and 7 triconsonantal clusters were selected. Three native speakers of Dutch produced these coda consonants five times in 1275 words and 810 nonwords. Again, digital recordings were made and acoustic measurements of the vowel and the coda consonant(s) were made. However, in the result section only the data for the nonwords were reported because Waals argued that the results for the words and their nonword counterparts were virtually equivalent but that the nonwords were "better balanced", "which should lead to cleaner results". While it is not clear to me how the nonwords should lead to cleaner results if they were virtually equivalent to the results for the words, a huge set of data was discarded without a convincing reason. It would have been nice if the author had shown (with some examples at least) that the results of the words and nonwords were really comparable. My personal opinion is to use words over nonwords whenever possible, and since the data for the words were recorded, it is a pity that they are not reported.

    The durational analysis of the singleton codas revealed three significant main effects: The effect of type of consonant indicates that there were differences between the consonant durations. Liquids were again shorter than the rest of the consonants, labial obstruents had the longest durations. Consonants that follow a short vowel were longer than consonants following a long vowel. However, as can be seen in a figure plotted by the author (Figure 2 on p. 48), this was only true for a small subset of the consonants under investigation. Unfortunately, Waals did not take the next logical step, which would be to state whether the interaction between the two factors "type of consonant" and "vowel length" was significant. Unfortunately, only the main effects were reported. The third main effect was the factor "subject", implying variability between the speakers. It is confusing that Waals states in the next sentence that "we have found that the factors subject and repetition do not have a significant influence on the duration of mono-consonantal codas" (p. 50), when the significant effect of the factor "subject" was reported before.

    For the duration of biconsonantal clusters in word coda position, Waals observed that there was a lower limit: No matter whether the coda cluster was preceded by a long or by a short vowel, it was never shorter than 115 ms. In general, the first consonant in the cluster is shortened relatively more than the second, and this effect was especially strong after short vowels. That is, coda clusters were the mirror image of onset clusters with respect to shortening of individual segments. Furthermore, the relative duration of the second consonant depended on the sonority of the preceding consonant, e.g., the duration of a second consonant in a coda cluster was relatively longer after a liquid than after a nasal. Consonants in triconsonantal clusters were not realized any shorter than in biconsonantal clusters, i.e., the locality effect observed in the onset clusters also applied to codas. Triconsonantal clusters were again shown to be compositional with respect to duration.

    Vowel length differed according to the number of consonants in the coda cluster, i.e., vowels were longest when followed by singleton codas and shortest when followed by a triconsonantal coda cluster. That means, unlike consonants, vowels were sensitive to the number of following consonants.

    Chapter 4 focused on the temporal properties of intervocalic consonants and consonant clusters. Intervocalic consonant clusters in Dutch generally consist of maximally three consonants, in some cases even four consonants are possible. The materials for this study consisted of 122 pseudoword targets. Altogether, 12 singleton consonants and 77 bi-, tri-, and quadriconsonantal clusters were investigated in intervocalic position. One hundred and twenty-two stress-initial pseudowords were created and produced five times embedded in a carrier phrase by the same three native speakers who had already served as subjects in the word-coda study. For the singleton intervocalic consonants, the factor "type of consonant" was significant. Fricatives were longer than stops, and in general the duration increased as the sonority of a consonant increased. Although labials were longer than dentals, differences due to the place of articulation were generally very small. Furthermore, the factor "vowel length" was found to be significant. Consonants were shorter after long vowels than after short ones. For some consonants, involving the sonorants and the voiced obstruents that were investigated, the differences were virtually non-existing, however. But as in the previous chapter, the author did not report whether or not the interaction between the factors "type of consonant" and "vowel length" was significant. For biconsonantal clusters, Waals observed that these clusters were shorter after a long vowel than after a short vowel, but again there seemed to be a lower limit (101 ms). Generally, differences in place of articulation did not lead to major durational differences in the individual subcategories of the biconsonantal clusters. For triconsonantal clusters, preceding vowel length did not have an effect on the duration of the cluster (although no statistics were provided except for mean values). With respect to the internal characteristics of the clusters, in some triconsonantal clusters the individual consonants were shorter than in their biconsonantal counterparts, in some they were not (see p. 88). The consonants in the quadriconsonantal clusters were not any more shortened than in the triconsonantal clusters. Vowels were found to be shortest when preceding a plosive, and longest when preceding a liquid. Vowels were longer when preceding voiced consonants than when preceding voiceless consonants. Also, vowels were shorter when they preceded a single consonant as compared to a consonant cluster, and their duration depended partly on the identity of the first consonant of a cluster.The main result of Chapter 4 is that the temporal properties of intervocalic clusters are determined by the same principle as the word-onset and word-coda clusters, namely: Group consonants observing sonority properties, elsewhere follow metrical structure. By metrical structure Waals means syllable structure based on phonological considerations. The MSDH was shown to predict also the durational behavior of triconsonantal intervocalic clusters in that duration in these clusters depended on sonority and phonological metrical structure. The MSDH was also reflected in the syllabification of intervocalic consonant clusters which follows two principles: (a) language-specific requirements (e.g., the VX condition in Dutch) and (b) the universal Maximal Onset Principle (MOP). However, in Dutch the language-specific principles seemed to be overruled by the universal principles.

    In Chapter 5 the assumption was investigated that the syllabification of consonants in intervocalic clusters is realised in the duration data. Single intervocalic consonants are syllabified with the following syllable according to the universal MOP. This is also implied by psycholinguistic data. However, Waals only cited data for English and ignores that similar syllabification experiments were carried out for Dutch by Schiller, Meyer, & Levelt (1997), which may be more relevant to the issue under investigation (i.e., the syllable in Dutch). Therefore, one would expect single intervocalic consonants to behave similarly to word-onset consonants in terms of duration. But, comparing the durations of intervocalic single consonants with word-onset consonants showed that the former were much shorter than the latter. However, this may be the result of polysyllabic shortening. The consonants in biconsonantal clusters were much longer in intervocalic position than in word-coda position. This was especially true for the second consonant. Compared to word-onsets, the first consonant in biconsonantal intervocalic clusters is much shorter which was in agreement with polysyllabic shortening. Taken together, this confirmed the expectation that clusters in intervocalic position were generally slit up but obstruent-sonorant clusters function as word-internal onsets. The MOP overrules the language-specific VX condition with the effect that any O+S cluster is dealt with durationally as an onset.

    Chapter 6 summarizes the results that were found earlier. The dissertation aimed at an investigation into the relation between segment durations and syllable structure. A relationship between metrical syllable structure and the duration of consonants can be found.

    Overall, this thesis constitutes a careful investigation of the phonetics and phonology of the syllable in Dutch. It provides a lot of empirical data, and some of the results are very interesting. The data is accounted for in terms of the Metrical Segment Duration Hypothesis, which seems to capture most of the data. However, the statistical analyses could be improved (by reporting interactions of the main effects and by conducting item-analyses). Except for some minor details that were not completely clear to me, the thesis is well written and the experimental work was carefully conducted. Therefore, this thesis will be of use for everybody interested in the phonetics and phonology of the Dutch syllable.

    bibliography

    Schiller, N. O., Meyer, A. S., & Levelt, W. J. M. (1997). The syllabic structure of spoken words: Evidence from the syllabification of intervocalic consonants. Language and Speech, 40, 103-140.

    Reviewer's biography

    >From 1994 to 1998. I worked at the Max Planck Institute for Psycholinguistics in Nijmegen, The Netherlands. In 1997, I received his Ph.D. in Psychology from Nijmegen University. In my PhD dissertation, I investigated the role of the syllable in lexical access during speech production. As of March 1998 I am affiliated with the Cognitive Neuropsychology Laboratory of the Psychology Department at Harvard University in Cambridge, USA. My research interests comprise phonological and morphological encoding in speech production, language processing in neurologically impaired patients, articulatory-motor processes during speech production, and forensic phonetics.