Principles of syllabification for Swedish — a methodological study

One aim of the research project “Prosodic coding and decoding of words in fluent speech: a cross-language investigation between Swedish, Greek and French (ProCoF)” is to investigate the syllable from a phonological and phonetic point of view. Due to the differing opinions about the rules for syllabification in Swedish, a methodological study has been carried out in order to shed light on the consequences of various principles for syllabification and to select the optimal algorithm which, at the same time, should be compatible with the best algorithm for French. Syllables in Standard Swedish According to tradition, syllabification has been carried out relying exclusively on the segmental structure of words and utterances. However, languages may also have prosodic features like quantity and stress that, by nature, make syllabification more difficult. In contrast to French, Standard Swedish has not only quantity expressed in one segment, e.g. the vowel like German, Dutch or Danish, but shows the characteristic quantity pattern of complementary length which means that a stressed long vowel is followed by a short consonant and a stressed short vowel by a long consonant or consonant cluster within a morpheme. It is interesting to note that the tonal prosodic feature of the two Swedish word accents, acute vs. grave, is of no significance to syllabification in Swedish. In the literature on Swedish syllables, various principles for syllabification are to be found. They may be summarised like this: (1) a stressed syllable attracts as many consonants as possible (Aurén, 1869; Lyttkens and Wulff, 1885). At the same time, the consonant cluster must be admissible initially and finally in words. (2) as many medial consonants as possible are attached to the following syllable (Noreen, 1907). The sonority hierarchy proposed by Jespersen (1932) yields the criteria for the division. (3) the boundaries between syllables are determined in the first place by phonotactic considerations and in the second place by the degree of coarticulation (Malmberg, 1968). A good demonstration of the consequences of these three syllabification algorithms can be shown using the word 'aspar (aspens): 'asp ar (according to 1 above) 'as par (according to 2 above) 'a spar (according to 3 above) In recent years, a more phonologically oriented approach to syllabification has emerged. Kristoffersen (1991), in his work on Norwegian syllable structure, favours three principles: (1) the Onset Rule stated by McCarty and Prince (1986), (2) a weak version of the Sonority Sequencing Generalisation and (3) the minimally constraint on stressed syllables. His treatment of syllable division is strikingly similar to Gårding’s algorithm (Gårding, 1967). The aim of this methodological study is to analyse a given speech sample according to different algorithms for syllabification, to compare the results and assess them and, based on