Temporally structured motor and auditory representations in covert syllable production

• 成果类型: Article
• 署名作者: Orpella, Joan; Mantegna, Francesco; Oderbolz, Chantal; Assaneo, M. Florencia; Poeppel, David
• 署名单位: Georgetown University; New York University; University of Oxford; Universidad Nacional Autonoma de Mexico; New York University
• 刊物名称: PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
• ISSN/ISSBN: 0027-8424; 1091-6490
• DOI: 10.1073/pnas.2536563123
• 发表日期: 2026-09-15
• 页码: e2536563123
• 关键词: speech production; inner speech; magnetoencephalography; Decoding; covert speech; SPEECH; phenomenology; language; imagery; models; brain; fmri; WORD; meg

摘要

Inner speaking-the deliberate generation of speech through covert production mechanisms-is a common human experience. Current accounts propose that it engages sensorimotor processes similar to those underlying overt speech. Here, we asked whether motor and auditory representations during inner speaking maintain the granularity and temporal structure observed in overt speech. Using magnetoencephalography and time-resolved decoding, we examined neural dynamics across two experiments. Participants covertly produced consonant-vowel syllables designed to emphasize either articulatory or acoustic phonetic components through contrasts in the consonants or vowels, respectively. Across experiments, neural activity closely paralleled stages for overt speech. Consonants (articulatory contrasts) were more strongly decoded from speech motor cortex before the expected syllable onset, whereas vowels (acoustic contrasts) were more strongly decoded from auditory cortex immediately afterward, consistent with the onset of internally generated percepts. Vowel discriminability in auditory regions was further graded according to separation along a perceptually relevant acoustic dimension in formant space approximated by the F2-F1 projection, indicating specificity in auditory representations. These findings reveal temporally structured, phoneme-level content-specific sensorimotor representations during inner speaking, consistent with production-based accounts involving internal forward models.