Ordin, M. 1, 2 & Soto, D. 1, 2
2 IKERBASQUE - Basque Foundation for Science
Using a classical artificial language learning paradigm, we tested the ability of bilingual (Basque-Spanish) and monolingual (Spanish) listeners to use the statistical probabilities (TP) to segment two artificial languages presented dichotically, and to probe the metacognitive abilities of bilinguals and monolinguals regarding the segmentation decisions. Each artificial language was composed of six tri-syllabic words with the syllabic TPs within words equal to 100%, and the syllabic TPs straddling the word boundaries equal to 20%. The learning phase was followed by a dual forced choice test pitting the statistical words versus random 3-syllabic sequences, and a confidence rating (how sure the listeners are in their choice). The results did not reveal the effect of the ear (during learning/recognition/whether the words were presented to the same or different ears during the learning and recognition phases). Both groups performed equally and significantly above chance, with the overall success rate around 60%. Monolinguals assigned the same confidence rating both to correct and incorrect responses. Bilinguals, on the other hand, assigned substantially and significantly higher confidence rating to correct responses than to incorrect responses. This shows that bilinguals exhibit higher metacognitive sensitivity compared to monolinguals in statistical learning tasks.