Foreign-language speech segmentation in ab initio child learners: The role of sublexical and lexical L2 overlap and phonological awareness

Von Holzen, K., Schnieders, M., Wulfert, S., & Hopp, H. (2023). Foreign-language speech segmentation in ab initio child learners: The roles of sublexical and lexical L2 overlap and phonological awareness. Poster presented at the 48th Annual Boston University Conference on Language Development, Boston, MA.One of the first steps in acquiring a new language is segmenting the continuous speech stream to find individual word forms. Research on adults segmenting a foreign language (FL) has shown both the continued application of sublexical (phonotactic) cues from the L1 (Finn & Hudson Kam, 2008) and facilitation for words overlapping in form and meaning with their L1 equivalents (i.e. cognates: English: /kraʊn/; German: /kroːnə/; noncognate: English: /skɪn/; German: /haʊ̯t/; Shoemaker & Rast, 2013). We study whether these sublexical and lexical L1 effects extend to elementary school children. These children start to learn an L2 together with learning how to read and developing their phonological awareness skills (Anthony & Francis, 2005), which has been found in particular to support later L2 acquisition in school-aged children (Hu, 2014). In two studies, we examine the effects and interactions of developing L1 lexical and sublexical knowledge and phonological awareness skills on the segmentation of English speech among German 6- to 9-year-olds before they receive instruction in English. 85 German 6- to 9-year-olds listened to English sentences followed by an isolated probe-word that either did (familiar) or did not (novel) appear in the sentence. Children indicated via button press whether they had heard the probe word in the sentence (word acceptance). In Study 1, target pseudowords appeared in a context which provided a clear L1-German phonotactic cue to a word boundary (e.g., lv; dal_vouchen) or an L1-ambiguous cue (kv; dack_vouchen). In Study 2, target words were English-German cognate and noncognate words (She reduced her crown/skin mursk to poverty). To measure phonological awareness skills, we used a modified phoneme-manipulation task from Bialystok et al. (2003). General-linear mixed models revealed a main effect of word status in Study 1 (β = 1.79, SE = 0.06, p < .001) and Study 2 (β = 1.05, SE = 0.05, p < .001) and an interaction between word status and phonological awareness in Study 1 (β = 0.12, SE = 0.02, p < .001) and Study 2 (β = 0.08, SE = 0.02, p < .001). Word acceptance was higher for familiar compared with novel probe words (Figures 1a and 2a