You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster B16, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Phrase Structure Building Evidenced by Differential Network Modulations

Chiao-Yi Wu1,2, Emiliano Zaccarella2, Angela D. Friederici2;1Nanyang Technological University, 2Max Planck Institute for Human Cognitive and Brain Sciences

The basic steps in building up language involve arranging two words of different lexical categories into a hierarchical structure. While the cognitive state of this process is widely discussed, a clear picture of its neurobiological basis has yet to be discovered. In the current study, we examined the neural mechanisms underlying basic phrase structure building in Chinese. Eighteen native speakers (8 male, mean age 23.9 years) were asked to judge whether a sequence of two visual stimuli could form a phrase to describe a noun in a 3T MRI scanner. We used syntax-driving determiners (Det), semantics-embedded classifiers (CL), and a symbol (#) to generate two-component sequences and manipulated two factors: sequence structure (phrase or list) and number of words (2-word or 1-word), resulting in four conditions: 2-word phrase (2-PH: Det + CL), 2-word list (2-LS: CL + CL), 1-word phrase (1-PH: Det + #), and 1-word list (1-LS: CL + #). Imaging data preprocessing and analyses were performed in SPM12. Differential responses to the conditions were identified in the left pars opercularis (BA44) and pars triangularis (BA45) of Broca’s area, and the left posterior middle temporal gyrus (pMTG) (cluster-level FWE-corrected p < 0.05). These regions served as regions-of-interest for effective connectivity analysis using dynamic causal modelling. Intrinsic connections were defined as bidirectional connections between every two regions for all models. Modulation effects by each task condition on any of the intrinsic connections were considered. Driving inputs from all conditions were specified on only BA44, only BA45, or both BA44 and BA45, producing a total of 192 models for comparison. All models were grouped into three families based on the driving inputs, and Bayesian model selection revealed that the models with driving inputs to both BA44 and BA45 outweighed the others (exceedance probability = 0.80). Bayesian model averaging and posterior probability (Pp) tests were conducted to evaluate the parameter estimates in the winning family, showing that the parameter estimates of all intrinsic connections and driving inputs were significantly deviant from zero (Pp > 0.95 corrected for multiple comparisons). The phrase conditions, where the first word (Det) opened a phrase, had stronger modulations on the connection from BA44 to pMTG compared to the list conditions, which highlights the pivotal role of this connection in phrase building and its association with syntactic integration. Within Broca’s area, the connection from BA44 to BA45 was negatively modulated by 2-PH whereas that from BA45 to BA44 by 2-LS. As BA44 is proposed to process syntax and BA45 semantics, the modulations by 2-PH and 2-LS may be respectively related to syntactic and semantic control mechanisms. Differential modulation effects were also found on the connections from pMTG to BA44 and to BA45, reflecting the particular demands on syntactic and semantic evaluation for the four conditions. Overall, our findings demonstrate differentiated network modulations emerging from sub-parts of Broca’s area and their interactions with pMTG in the process of phrase structure building, and provide important implications for understanding the foundation of language learning in humans.

Topic Area: Grammar: Syntax

Back to Poster Schedule