Embedding the past and the future into speech perception

Poster Session A - Saturday, April 13, 2024, 2:30 – 4:30 pm EDT, Sheraton Hall ABC

Hua Fan¹ (202121198438@stu.blcu.edu.cn), Lei Zhang¹, Yuting Meng¹, Ling Liu¹; ¹Beijing Language and Culture University, Beijing, China

How we perceive current speech is shaped by past and future information. While numerous studies have independently demonstrated the influence of preceding contextual information and subsequent tasks on speech perception, it remains unclear how the brain integrates these two types of information to modulate the perception of current speech. Here, we used MEG to record responses from participants engaged in a speech perception task, where they were instructed to determine whether the tones (tone task, subject n = 30) or consonants (consonants task, subject n = 19) of the two preceding and following syllables were the same. Using the multi-voxel pattern analysis method, we separately analyzed the neural signals when participants listened to the first and second syllables. By comparing the tone task and consonant task, we uncovered that phonemic representations are adapted in response to the speech perception task at hand. Specifically, when perceiving the second syllable, the task-related phonemic features of the first syllable, such as the tone feature in the tone task, will be automatically reactivated. Furthermore, we used representational similarity analysis combined with multidimensional scaling analysis to explore the representation of the second syllable. Our results revealed that the brain actively aligns the second syllable to the task-relevant phonemic features of the first syllable, with the second syllable being reassembled to match the tone/consonant features of the first syllable for the tone/consonant task. Overall, our results indicate that the perception of current speech is adapted to more efficiently match subsequent tasks based on preceding phonemic information.

Topic Area: LANGUAGE: Lexicon