Visual Statistical Learning with Linguistic and Non-linguistic Inputs

Wang, T. & Chen, J.

Department of Chinese as a Second Language, National Taiwan Normal University

Past research has demonstrated statistical learning with auditory linguistic inputs and visual nonlinguistic inputs. The present study examined statistical learning with visual linguistic (Hebrew letters, Korean letters) and nonlinguistic (unfamiliar geometric shapes) inputs. For each type of input, 12 symbols were selected and grouped into 4 triplets. Each triplet was repeated 24 times and randomly connected to one another, resulting in a continuous sequence of 288 symbols. The co-occurrence of symbol pairs within triplet was higher than between triplets. This was the only clue for determining triplet boundaries. Presentation of the symbols and the subsequent test followed Fiser and Aslin's (2002) procedure. Chinese participants' task was tracking the symbol movement when one appeared. After the task, participants took a 2 AFC test to determine if a triplet was shown before. The accuracy of the test was significantly higher than random guessing for each type of input, indicating successful visual statistical learning. Interestingly, statistical learning with the non-linguistic symbols was always better than statistical learning with the linguistic symbols. Explanations of the difference include the nature of linguistic and non-linguistic symbols, how they are processed (holistic vs. analytic), and the amount of attentional resources being consumed by the processing.