Poster E77, Monday, March 26, 2:30-4:30 pm, Exhibit Hall C
Age-related differences in the underlying mechanisms of temporal statistical learning
Noémi Elteto1, Karolina Janacsek1,2, Dezso Nemeth1,2; 1Eotvos Lorand University, Budapest, Hungary, 2Brain, Memory and Language Lab, Hungarian Academy of Sciences, Budapest, Hungary
Statistical learning underlies many day-to-day activities during the entire lifespan since it is crucial in the acquisition of perceptual, motor, cognitive, and social skills. Janacsek et al. (2012) reported that the basic ability to pick up implicitly triplets that occur with high- vs. low-probability in a sequence – measured by raw reaction time (RT) - is superior in children and it is decreased around age 12. Yet, the qualitative ontogenetic changes that give rise to the quantitative differences in performance are not understood yet. Here our aim was to characterize performance in more detail by estimating ex-Gaussian parameters of the RT distributions. We re-analyzed the dataset from the Janacsek et al. study where participants between 4–85 years of age were trained on a probabilistic sequence learning task. First, we confirmed the decreasing developmental pattern of temporal statistical learning; the difference in the mu parameter (mean value) between the high- and low-probability triplets was the highest before 12 years of age. Importantly, the sigma parameter (variance) was larger for the high- than the low-probability triplets in children, and this difference was gradually reversed through adolescence. However, triplet types were not differentiated by the tau parameter (exponential rate). This suggests that while children have acquired some high-probability triplets more than others, adults learned these approximately similarly. Therefore, we propose that the learning of high probability events per se undergoes a shift from weighting specific events to learning whole probabilistic structures around age 12.
Topic Area: LONG-TERM MEMORY: Skill learning