Slide Slam J9 Sandbox Series
Neural Correlates of Learning Nominal Classification Rules: an fMRI study
Diego Dardon1, Hyeonjeong Jeong1, Haining Cui1, Ryo Ishibashi1, Motoaki Sugiura1; 1Tohoku University
Learning a second language requires not just learning abstract grammatical rules, but rules that rely on language-specific conceptual information. However, previous studies have explored the learning of abstract grammar rules, and there are little studies investigating the diverse set of grammar rules that rely on conceptual knowledge. Previous studies on adult second language learning reported that during the learning abstract grammar rules engaged the left inferior frontal gyrus and left ventral premotor cortex (Kepinska, de Rover, Caspers, & Schiller, 2017; Opitz & Friederici, 2003; Opitz & Friederici, 2004). In this study, we investigate the neural correlates during the learning process of nominal classification rules, grammatical systems with language particular ways of conceptualizing the natural world into categories such as liquids or plants (Kemmerer, 2014). Participants were 37 healthy, right-handed Japanese native speakers (Age 18-24, 17 females). We created a semi-artificial language based on Luganda that consisted of 72 concrete nouns with the nouns borrowed from Japanese to control for vocabulary learning and target the learning of the grammatical rules. The target grammar rule being noun class agreement between a noun and a demonstrative with each noun class taking a unique demonstrative. The noun classes divided semantically as: animate (i.e., dog), small inanimate (i.e., pen), and large inanimate (i.e., train). These semantic divisions are common for noun classes typologically (Aikhenvald, 2000). Participants learned the semi-artificial language over 3 learning phases (time 1, time 2, time 3). fMRI Scanning took place during the learning phases. During the learning phase, participants listened to 18 randomized correct noun-demonstrative combinations (learning condition) with each noun-demonstrative combination including a picture of the noun to prevent ambiguity. In addition, they heard the same 18 noun-demonstrative combinations but with the sound in reverse and a mosaic picture (control condition). After each learning phase, participants performed an offline grammatical judgment task that acted as a behavioral indicator of learning. The learning phases combined took approximately 14 minutes and the test phases combined took approximately 9 minutes for a total experiment time of 21 minutes. A one-way repeated measures ANOVA was conducted for behavioral results to compare offline test scores across the three time periods. There was a significant effect for time, Wilks’ Lambda = .32, F (2,35) = 36.72, p < .001, multivariate partial eta squared = .67. These results that scores increased over time indicating learning took place. For fMRI analysis, the contrast of interest will be the Condition (learning vs control) X Time (time 1 vs time 3) interaction to find activation involved in learning nominal classification rules. In contrast to previous studies, we predict activation in the brain areas involved in conceptual and semantic knowledge, particularly the left anterior temporal lobes and left angular gyrus in addition to the areas related to the language network such as the left inferior frontal gyrus. We hope the findings contribute to the further understanding of the learning process in real-time with regards to the interaction between conceptual knowledge and grammatical structure.