| dc.description.abstract | Both in infant and adult speech segmentation, stress cues seem to be the most reliable markers for word boundaries. Stress cues are highly language specific in the sense that metrical systems differ from language to language. English-speaking adults are believed to take each stressed syllable to mark the onset of a new word. This Metrical Segmentation Strategy (MSS) is proposed to be a universal segmentation strategy and therefore predicted to be applied by Dutch speakers as well (Cutler and Norris, 1988). However, recent research has indicated that Dutch listeners depend heavily on their native (penultimate) stress pattern when segmenting words from a speech stream (van Ommen, in prep.). This language-specific knowledge must be learned by infants in order to be able to segment words for word acquisition. Research has shown that at the age of 3, Dutch children have mastered generalizations for regular stress assignment, and by the age of 4 they have become more sensitive to less regular stress assignment (Nouveau, 1994). The development of such knowledge towards adult- like word segmentation is yet to be documented. Therefore, the research questions of this thesis are: How and to what extent do Dutch learning children apply their knowledge about the Dutch stress system during word segmentation? And how do those segmentation strategies develop during growth of knowledge about the stress system?
To answer these questions a word-spotting task was conducted with 131 Dutch children (66 male and 65 female), age ranging between 6- to 14- years. Participants were auditorily presented with two-syllable nonsense-words with opposite stress patterns (Sw or wS), associated with two pictures. These two words were affixed to different preceding nonsense syllable-strings that were manipulated for stress position (the prefixes), creating 6 different stress conditions. Participants were instructed to press a button as soon as they recognized either of the names, while their accuracy scores and response latencies were measured.
Results showed that children segment targets faster from a speech stream when their age increases. Certain stress patterns are of greater support to the child during segmentation than others and this function changes with age. Children between 6 and 8 segment native penultimate Sw targets significantly faster from a speech stream than the less native wS targets, while children older than 8 years no longer segment Sw targets faster than wS targets. The data showed that younger children are not facilitated by the stress pattern of the prefix during segmentation, but from the age of 10 onwards, a prefix with a native penultimate stress pattern (wSw) combined with a target with a native stress pattern (wSw-Sw) has a highly facilitating effect on segmentation of the target. All findings suggest a development towards application of language specific segmentation cues in Dutch learning children, where native stress patterns are highly facilitating for word segmentation in both younger and older children and the segmentation strategies for targets with non-native stress patterns develops with age. | |
