Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Phase-amplitude coupling binds the FFR and envelope tracking in human intracranial recordings
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Hao Zhu1, Chen Yao5, Junxi Chen6, Danny T M Chan7, Xing Tian4, Xiangbin Teng1,2, Patrick C M Wong1,2,3; 1Brain and Mind Institute, The Chinese University of Hong Kong, Hong Kong SAR, China, 2Department of Psychology, The Chinese University of Hong Kong, Hong Kong SAR, China, 3Department of Linguistics and Modern Languages, The Chinese University of Hong Kong, Hong Kong SAR, China, 4NYU-ECNU Institute of Brain and Cognitive Science at NYU Shanghai, Shanghai, China, 5Department of Neurosurgery, Shenzhen Second People’s Hospital, The First Affiliated Hospital of Shenzhen University, Shenzhen, China, 6Department of Neurosurgery, Guangdong Sanjiu Brain Hospital, Guangzhou, China, 7Division of Neurosurgery, Department of Surgery, Prince of Wales Hospital, The Chinese University of Hong Kong, Hong Kong SAR, China
The frequency-following response (FFR) and envelope tracking index two different aspects of speech on two different timescales, but whether they share a neural population — and if so, how they are bound — has not been tested directly in humans. We recorded stereoelectroencephalography (sEEG) from 10 patients with drug-resistant epilepsy (1,453 bipolar channels covering bilateral cortical and subcortical regions; seizure-related channels excluded) during passive listening to 600 repetitions of a Cantonese Tone-4 syllable (a lexical low-falling tone) presented at 1.25 Hz. For each channel we measured 1.25 Hz spectral power (envelope tracking; paired t-test against neighboring-band power, FDR corrected) and inter-trial phase coherence (ITPC) at the fundamental frequency F0 (FFR; Rayleigh test, FDR corrected), partitioning channels into slow-only (n=198), fast-only (n=74), both-tracking (n=147), and neither-tracking (n=1,034) pools. On every channel we then computed phase-amplitude coupling (PAC) between 1.25 Hz phase and 70–120 Hz amplitude, benchmarked against an empirical null built from the neither-tracking pool, and a Bayes-factor (BF) comparison of an envelope (harmonic stack at 1.25/2.5 Hz) versus a fundamental-only (1.25 Hz alone) model. The FFR was distributed across cortical and non-brainstem subcortical sites: of 221 fast-tracking channels (fast-only + both-tracking), 14 sat in subcortical structures (7 hippocampus, 5 putamen, 2 medial pulvinar thalamus), extending F0-locked coding beyond canonical auditory cortex. A subpopulation in superior temporal gyrus (STG), Heschl's gyrus (HG), insula (INS), and Rolandic operculum (RO) carried both envelope tracking and the FFR on the same channel (147/1,453 channels; envelope power and F0 ITPC correlated across channels, r = 0.32, p < 10⁻⁴). Critically, 35 of the 147 both-tracking channels exceeded the 95th-percentile PAC null, with a group-mean modulation depth ~49% of F0 amplitude across the slow cycle; PAC did not exceed the null in slow-only or fast-only pools, arguing against a simple co-occurrence account. Bayesian model comparison further dissociated the two pools: envelope-like channels in auditory and peri-sylvian cortex (STG/HG/INS/RO) favored an envelope model (BF > 10³), whereas putamen channels — the only basal-ganglia nucleus densely sampled — favored a fundamental-only model (BF > 10⁵). Putamen channels nonetheless showed significant F0 ITPC, indicating a fundamental-locked but harmonic-less code. Lateralization was carried by the slow code only: on the slow-tracking pool, a per-ROI laterality index (on the 1.25 Hz power-excess metric; subject-cluster bootstrap, FDR across ROIs) showed right-lateralization in STG (LI = 0.12, p < 0.05) whereas F0 ITPC on the fast-tracking pool was not lateralized in any ROI under the same test (all p ≥ 0.5). Thus, the envelope tracking and FFR — typically studied as separate signals — are bound by phase-amplitude coupling on a shared population in auditory and peri-sylvian cortex. The basal ganglia carry a parallel syllable-rate code lacking the harmonic-stack signature, and only the slow code shows hemispheric asymmetry (right-lateralized in STG).
Topic Areas: Speech Perception,