Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Multilingual Experience Has No Effect on Statistical Language Learning
Poster Session C, Thursday, October 1, 10:45 am - 12:45 pm, Wangari Maathai
Saara Kaskivuo1, Eleonore H.M. Smalle2, Riikka Möttönen1; 1University of Helsinki, 2Tilburg University
While potential cognitive advantages of bi- and multilingualism have been widely studied, comparatively little research has addressed their effects on novel language learning in adulthood. Statistical language learning (SLL)—the ability to track co-occurring regularities in continuous linguistic input—is considered a foundational mechanism of first and second language acquisition. Previous research suggests a multilingual advantage in SLL, but most studies used tasks that include additional perceptual cues (e.g., tonal or prosodic markers), rather than measuring "pure" tracking of transitional probabilities. Moreover, monolingual adults perform better in SLL when novel regularities resemble their native language (i.e., “entrenchment”). Furthermore, multilingualism has typically been treated as a binary variable despite the wide, continuous variation of linguistic experience across individuals. Here, we asked whether the diversity of multilingual experience (MLD) modulates pure SLL, and whether MLD shapes entrenchment in SLL. 127 native Finnish-speaking adults (aged 19–40) with experience in 2–11 additional languages completed an online SLL experiment. Multilingual experience was quantified using an entropy-based Multilingual Language Diversity (MLD) score derived from the Language History Questionnaire (LHQ3.0), capturing the proficiency, number, and balance of daily use across all reported languages (sample score range: 1.11–3.38). Participants completed an auditory and a visual SLL task targeting core probabilistic learning. Both tasks started with a continuous exposure stream, with hidden pseudowords that could only be segmented by tracking transitional probabilities. The tasks differed in how learning was tested: the visual task probed syllable recall for both exposed and novel structures, while the auditory task used a two-alternative forced-choice (2AFC) task (foils vs. pseudowords) with confidence ratings, testing both adjacent and nonadjacent dependencies and strength of knowledge for the newly learned items. Half of the novel words in the visual task broke Finnish vowel harmony, allowing us to probe L1 entrenchment. We calculated correlations between MLD and learning scores, and complemented our analyses with a data-driven cluster analysis combining MLD with earliest age of L2 exposure (AoE). Participants showed robust group-level learning across tasks: visual recall was significantly higher for exposed than novel trigrams (t(111)=4.67, p<.0001) , and auditory 2AFC accuracy was above chance for both adjacent (M=.62, p<.001) and nonadjacent (M=.52, p=.009) dependencies. Confidence ratings were higher for correct than incorrect responses, indicating some explicit knowledge. However, MLD did not correlate with any learning score in either modality (all |r|≤.13, all p≥.2), nor with confidence-based knowledge. Cluster analysis identified three multilingual profiles (later/higher MLD; earlier/lower MLD; later/lower MLD), but generalized linear mixed-effects models revealed no differences between profiles in any task. Disharmonious bigrams were recalled slightly better than harmonious ones, but this entrenchment-related effect was also unaffected by MLD. These findings indicate that the diversity, proficiency, and timing of multilingual experience do not influence the ability to extract statistical regularities from novel linguistic input. This suggests that core SLL is not affected by variation in adults' multilingual experience. If a multilingual advantage in language learning exists, it likely arises from mechanisms beyond statistical tracking, such as the integration of additional acoustic or prosodic cues.
Topic Areas: Multilingualism, Language Development/Acquisition