Sonority and cross‐linguistic acquisition of initial s‐clusters

Abstract

This paper examines the findings and implications of the cross‐linguistic acquisition of #sC clusters in relation to sonority patterns. Data from individual studies on English, Dutch, Norwegian, and Hebrew are compared for accuracy of production as well as the reductions with respect to potential differences across subtypes of #sC groups. In all four languages, a great deal of variability occurred both within and across children, but a number of general patterns were noted. While all four languages showed similar behaviour in reduction patterns, clear differences between the three Germanic languages–English, Dutch, and Norwegian–and Hebrew were found in the accuracy rate of #sC clusters that have a larger sonority distance between C1 and C2, in a binary split of /s/+[+cont.] versus /s/+[−cont.], and in #sC clusters versus non‐/s/‐clusters. Patterns in cluster reductions were, in general, supportive of the factorial typology of Pater and Barlow, while pointing at some difficulties for the predictions of the headedness approach advanced by Goad and Rose, and Jongstra. This cross‐linguistic description of patterns of s‐clusters increases our knowledge of typical phonological development and helps selecting targets in clinical context.

• Sonority
• s‐clusters
• phonological acquisition

Notes

1. Among these are openness of the vocal tract (Jespersen, 1922; Donegan, 1978; Halle & Clements, 1983; Lass, 1984; Katamba 1989; Durand, 1990), propensity for spontaneous voicing (Chomsky & Halle, 1968; Kenstowicz, 1994; Brakel, 1979), resonance (Foley, 1977; Heselwood, 1998), acoustic energy (Keating, 1988; Goldsmith, 1990; Wright, 2001), loudness (Selkirk, 1984; Clements, 1990; Lavoie, 2000), continuity of the spectrum amplitude (Stevens, 1987; Stevens & Keyser, 1989), tonality (Boersma, 1998; Heselwood, 1998) to name a few.

2. There are several other proposals which differ in details. For example, Selkirk (1984) draws a finer distinction among fricatives. Also, considerable dispute exist with respect to the relative order of voiced stops and voiceless fricatives. While some rank voiced stops higher (Jespersen, 1904; Bolinger, 1962; Brakel, 1979; Boersma, 1998), others have suggested the reverse order by ranking voiceless fricatives higher (Lass, 1984; Selkirk, 1984; Katamba, 1989; Durand, 1990; Blevins, 1995; Gnanadesikan, 1997). Yet, others treat the two groups of sounds equal in sonority (Venneman, 1972; Hooper, 1976; Larson, 1990).

3. Epenthesis (e.g. blue [blu]→[blu]), and coalescence (e.g. swim [swim]→[fim]) also found, but these patterns are much less common than reductions.

4. The detailed account of these individual studies are found in a special issue of the Journal of Multilingual Communication Disorders (vol. 4, no. 3): Yavaş & Core (2006) on English, Ben‐David (2006) on Hebrew, Gerrits and Zumach (2006) on Dutch, and Kristoffersen and Simonsen (2006) on Norwegian. Also, it is worth mentioning that the choice of languages in our study initially targeted comparison among the three Germanic languages. The availability of Hebrew participation gave us an opportunity to compare these languages with one coming from another family. Currently, we are expanding our inquiry by the addition of Croatian, a Slavic language, examining data from children with phonological disorders for the same structures.

5. The reader is referred to the references cited in note 4 for further details.

6. The differences between Hebrew and the other languages in Figure 2 can also be attributed to the special nature of // in Hebrew acquisition. It is one of the last consonants to be acquired and when being misarticulated, it tends to be deleted in child phonology, even when not in clusters, so although all the children in the Hebrew data produced // correctly, perhaps it was not fully acquired by then.

7. The cluster /sk‐/ is rare in Dutch. The frequency of /sk‐/ in written newspaper texts (CELEX database) is less than .5% according to Jongstra (2003). This cluster only occurs in loan words such as “ski”, “skateboard”, “skelet”. Since we expected these words to be familiar to children, we included them in the instrument.

8. The fact that /sp‐/ initial clusters are rare in Hebrew and appear only in loan words, can be attributed to the spirantization constraint (Schwarzwald, 1976; Bolozky, 1978; Adam, 1993) remnants of which occur in modern Hebrew, such that the stops /p, b/ do not occur in clusters in C2 position.

9. In the comparison of the individual targets, the difference between /sn‐/ and /sw‐/ (48% and 76% accuracy respectively) is significant (p<.002) in English (Yavaş & Core, 2006).

10. As indicated in appendix B, the clusters in parentheses are rare.

11. This hypothesis is based on studies on phoneme segmentation by adults (Treiman, 1989), which states that rhymes containing a vowel followed by a consonant of low sonority index such as a stop or a fricative are easier to segment than rhymes containing a vowel followed by a consonant with a high sonority index such as a nasal or a liquid.