RT Journal Article
SR Electronic
T1 How Grammar Conveys Meaning: Language-specific spatial encoding patterns and cross-language commonality in higher-order neural space
JF The Journal of Neuroscience
JO J. Neurosci.
FD Society for Neuroscience
SP JN-RM-0599-23
DO 10.1523/JNEUROSCI.0599-23.2023
A1 Jing Wang
A1 Hui Lin
A1 Qing Cai
YR 2023
UL http://www.jneurosci.org/content/early/2023/09/14/JNEUROSCI.0599-23.2023.abstract
AB Languages come in different forms but have shared meanings to convey. Some meanings are expressed by sentence structure and morphological inflections rather than content words, such as indicating time frame using tense. This fMRI study investigates whether there is cross-language common representation of grammatical meanings that can be identified from neural signatures in the bilingual human brain. Based on the representations in inter-sentence neural similarity space, identifying grammatical construction of a sentence in one language by models trained on the other language resulted in reliable accuracy. By contrast, cross-language identification of grammatical construction by spatially matched activation patterns was only marginally accurate. Brain locations representing grammatical meaning in the two languages were interleaved in common regions bilaterally. The locations of voxels representing grammatical features in the second language were more varied across individuals than voxels representing the first language. These findings suggest grammatical meaning is represented by language-specific activation patterns, which is different from lexical semantics. Commonality of grammatical meaning is neurally reflected only in the inter-stimulus similarity space.Significance StatementWhether human brain encodes sentence-level meanings beyond content words in different languages similarly has been a long-standing question. We characterize the neural representations of similar grammatical meanings in different languages. Using complementary analytic approaches on fMRI data, we show that the same grammatical meaning is neurally represented as the common pattern of neural distances between sentences. The results suggest the possibility of identifying specific grammatical meaning expressed by different morphological and syntactic implementations of different languages. The neural realization of grammatical meanings is constrained by the specific language being used, but the relationships between the neural representations of sentences are preserved across languages. These findings have some theoretical implications on a distinction between grammar and lexical meanings.