[PS-3.20]The sensitive period for associative learning of nonadjacent dependencies: an ERP-study

Paul, M. 1, 2 , Männel, C. 1, 3 , Mueller, J. L. 4 , van der Kant, A. 5 , Wartenburger, I. 5 , Höhle, B. 5 & Friederici, A. 1

1 Department of Neuropsychology, MPI for Human Cognitive and Brain Sciences, Stephanstr.1, Leipzig, Germany
2 Berlin School of Mind and Brain, Luisenstr. 56, Berlin, Germany
3 Department of Neurology, Stephanstr. 1A, Clinic for Cognitive Neurology, Medical Faculty of the University of Leipzig
4 Institute of Cognitive Science, Osnabrück University, Wachsbleiche 27, Osnabrück, Germany
5 Linguistics Department, University of Potsdam, Karl-Liebknecht-Straße 24-25, Potsdam, Germany

Grammatical dependencies between non-adjacent elements are important building blocks of language. For example, the sentence ?The sister is singing? requires the listener to track grammatical relations between the auxiliary ?is? and the verb suffix ?-ing?. Recent event-related potential (ERP) studies revealed that 4-month-old infants can learn these dependencies by merely listening to correct examples, while adults only learn under active task conditions (Mueller et al.,2009; Friederici et al., 2011). In the current ERP study, we aim to specify at what age grammar learning changes from infants' associative, effortless processing to adults' more controlled, effortful processing by testing German children between 12 and 36 months of age. We mirror natural grammar learning by using a non-native language and expose children to Italian sentences containing a non-adjacent dependency between an auxiliary and a verb suffix (e.g. ?La sorella sta
cantando? (The sister is singing)). Following learning blocks with correct examples, the comparison between ERP responses to correct and grammatically incorrect sentences (i.e. non-adjacent dependency violations) reveals up to what age children process non-adjacent dependencies from mere exposure. So far, ERP-responses in 36-month-old children suggest that at this age, children no longer employ associative mechanisms in learning non-adjacent dependencies in natural language.