Lexical overlap in foreign language speech segmentation in primary-level students

Von Holzen, K., Schnieders, M., Wulfert, S., & Hopp, H. (2023). Lexical overlap in foreign language speech segmentation in primary-level students. Presented at EuroSLA 32, Birmingham, UK. doi: 10.6084/m9.figshare.24041883.Policy makers base decisions about the ideal starting date for foreign language (FL) teaching on FL learning outcome studies at secondary level which often support a later start (Jaekel et al., 2017). However, their results are influenced by the quality of instruction rather than the learning abilities by the students (Baumert et al., 2020).To study students’ initial learning abilities, we focus on a key skill: speech segmentation. Adults initially struggle to extract or segment word forms from continuous speech in an FL but succeed when target words partially overlap in form and meaning with their L1 equivalents (i.e. cognate: English: /kraʊn/; German: /kroːnə/; noncognate: English: /skɪn/; German: /haʊ̯t/; Shoemaker & Rast, 2013) or when target words are adjacent to previously familiarized words (Cunillera et al., 2010). To identify the optimal age for starting FL learning at primary-school level, we study the role of lexical overlap with German for the segmentation of English speech among 1st graders early in the school year and among 2nd graders towards the end of the school year before they have received instruction in English.In a word recognition study, English-German cognate and noncognate word pairs (n = 160) were embedded into an otherwise identical English sentence frame (She reduced her crown/skin mursk to poverty). Within each sentence frame cognate and noncognate words were followed by a noncognate pseudoword (i.e. mursk). Students listened to the sentence followed by an isolated probe-word and indicated via button press whether they heard the probe word in the sentence.In the talk, we will report results from 48 1st graders and 48 2nd graders to be collected by July. So far, we have analyzed results from 27 first-graders. A general-linear mixed model revealed no difference in target accuracy for cognate and non-cognate words (p = 0.83), nor was there a difference in d’prime between them (p = .97; Figure 1). There was also no difference in target accuracy for cognate and non-cognate following words (i.e. mursk, p = 0.53) or in d’prime between them, (p = .97; Figure 2). These preliminary results provide suggestive evidence that 1st grade students do not yet transfer knowledge from their L1 to segment words in an unknown FL. I