Statistical Learning and Language Abilities in Typically Developing Children

• Main Authors: Yuval-Weiss, Neta, יובל-ויס, נטע
• Format: Dissertation
• Language: English
• Published: ProQuest Dissertations & Theses 01-01-2015
• Subjects: Adolescents; Age groups; Bilingual education; Biology; Confidence intervals; Control Groups; Design; Early childhood education; Education; Feedback; Grade 1; Grade 4; Grammar; Hebrew language; Intelligence; Language; Language acquisition; Language proficiency; Language Skills; Language Tests; Learning Processes; Learning Strategies; Linguistics; Memory; Morphology; Phonology; Preschool children; Reading comprehension; Reading instruction; Reading Skills; Reading Tests; Second language learning; Skills; Speaking; Standard scores; Statistics; Thinking Skills; Young Children
• ISBN: 9798480600599

Statistical learning, a process involving the extraction of regularities from sequences of stimuli, is a powerful mechanism in the early stages of language acquisition. However, it remains unclear whether statistical learning (SL) continues to contribute to language development beyond infancy. SL tasks can be grouped according to the statistical structure targeted for learning during the experiment (Thiessen & Erickson, 2013)− conditional relationships and distributional information while it is not sure yet whether there is one underlying process that can explain both of these learning mechanism. In the context of language development, the SL literature focuses mainly on the statistical learning of conditional relationships, whereas distributional information generally receives less attention.This study investigated the performance of typically developing children in two statistical learning tasks which differ in the statistical information to be acquired: (1) a visual task that contained conditional relationships, in which children were exposed to sequences of visual stimuli that were (unknown to the children) organized in triplets. Subsequently, they were asked to identify the triplets presented during the exposure phase and (2) a phonological memory task that contained two conditions with different amounts of distributional information. Differences in memory spans between the two conditions were used as an estimate of distributional information learning. Our aim was to explore different types of SL tests and their associations to language proficiency. Correlations between SL, vocabulary, grammar and reading were found in previous studies in both infants and adults mainly in tasks containing conditional relationships. We investigated the associations between the two SL tasks in children 6-10 years old, as well as the associations between statistical learning and language abilities in this age range. We expected to find:An increasing amount of SL in both tasks with age and an association between performance in the SL tasks and language measurements.Two experiments were conducted to test these hypotheses. Twenty preschool and 20 first grade children participated in the first experiment; 30 first grade and 30 fourth grade children participated in the second. As we predicted, SL was observed in both tasks, but its magnitude did not change significantly with age, a weak correlation was found between the SL tasks. As for the language measurements we also found correlations between SL, vocabulary and grammar.The results suggest that the magnitude of SL changes only mildly during childhood, or alternatively that the change is very gradual and thus not observed in the age range studied here. Moreover, the correlations to language measurements suggest that SL accounts (statistically) for unique variance in language skills even beyond the stages of language acquisition.