Cortical encoding of melodic expectations in human temporal cortex

  1. Giovanni M Di Liberto  Is a corresponding author
  2. Claire Pelofi
  3. Roberta Bianco
  4. Prachi Patel
  5. Ashesh D Mehta
  6. Jose L Herrero
  7. Alain de Cheveigné
  8. Shihab Shamma  Is a corresponding author
  9. Nima Mesgarani  Is a corresponding author
  1. Laboratoire des systèmes perceptifs, Département d’études cognitives, École normale supérieure, PSL University, CNRS, France
  2. Department of Psychology, New York University, United States
  3. Institut de Neurosciences des Système, UMR S 1106, INSERM, Aix Marseille Université, France
  4. UCL Ear Institute, United Kingdom
  5. Department of Electrical Engineering, Columbia University, United States
  6. Mortimer B Zuckerman Mind Brain Behavior Institute, Columbia University, United States
  7. Department of Neurosurgery, Zucker School of Medicine at Hofstra/Northwell, United States
  8. Feinstein Institute of Medical Research, Northwell Health, United States
  9. Institute for Systems Research, Electrical and Computer Engineering, University of Maryland, United States
7 figures, 3 videos and 5 additional files

Figures

System identification framework for isolating neural correlates of melodic expectations.

(A) Music score of a segment of auditory stimulus, with its corresponding features (from bottom to top): acoustic envelope (Env), half-way rectified first derivative of the envelope (Env’), and the …

Figure 2 with 1 supplement
Low-rate (1–8 Hz) cortical signals reflect melodic expectations.

Scalp EEG data were recorded while participants listened to monophonic music. Forward ridge regression models were fit to assess what features of the stimulus were encoded in the low-rate EEG …

Figure 2—figure supplement 1
The latency of the effect of melodic expectations on the low-rate EEG data (1–8 Hz) was assessed by measuring the loss in EEG prediction correlation when a given time-latency window is removed from the TRF fit.

This loss was calculated for a sliding 50ms-long latency window and was plotted as a heat-map for each electrode. White colored pixels indicate non-significant results (p>0.05, cluster-mass …

Figure 3 with 1 supplement
Low-rate (1–8 Hz) cortical signals in bilateral temporal cortex reflect melodic expectations.

Electrocorticography (ECoG) data were recorded from three epilepsy patients undergoing brain surgery. Magnetic resonance imaging was used to localise the ECoG electrodes. Electrodes with stronger …

Figure 3—figure supplement 1
Bilateral electrocorticography (ECoG) results for Patient 3.

Electrodes with stronger low-rate (1–8 Hz) or high-γ (70–150 Hz) time-locked responses to musical notes than to silence were selected (Cohen’s d > 0.5). (A) Low-rate ECoG prediction correlations for …

Figure 4 with 1 supplement
High-γ neural signals in bilateral temporal cortex reflect melodic expectations.

Electrodes with stronger low-rate (1–8 Hz) or high-γ (70–150 Hz) responses to monophonic music than to silence were selected (Cohen’s d > 0.5). (A) ECoG prediction correlations for individual …

Figure 4—figure supplement 1
Bilateral electrocorticography (ECoG) results for Patient 3.

Electrodes with stronger low-rate (1–8 Hz) or high-γ (70–150 Hz) time-locked responses to musical notes than to silence were selected (Cohen’s d > 0.5). (A) High-γ ECoG prediction correlations for …

Figure 5 with 1 supplement
Event-related potentials (ERP) analysis.

(A) Notes with equal peak envelope were selected (median envelope amplitude across all notes with a tolerance of ±5%). Together, the selected elements were 25% of all notes. Notes were grouped …

Figure 5—figure supplement 1
Event-related potentials (ERP) analysis.

(A) Musical notes with equal envelope amplitude (within 5% of the median amplitude for all notes) were selected. Together, this corresponds to the 25% of all data. Musical notes were grouped in five …

Figure 6 with 1 supplement
Distinct cortical encoding of pitch and note onset-time during naturalistic music listening.

(A) Contrasts at each EEG channel of the TRF weights for surprise vs. entropy (top) and pitch vs. onset-time (bottom) in TRFAM. Colors indicate significant differences (p<0.05, permutation test, …

Figure 6—figure supplement 1
Scatter plots indicating the correlation between EEG prediction correlation using the acoustic regressors A for each musical piece and the average expectation score (Sp, Hp, So, or Ho) for all notes of the corresponding piece.

Significant correlations were highlighted in red. Results are shown for non-musicians and musicians separately.

Effect of musical expertise on the low-rate encoding of melodic expectations.

(A) EEG prediction correlations (average across all scalp channels) for musicians and non-musicians (*p<10−5). The error bars indicate the SEM across participants. (B) EEG prediction correlations …

Videos

Video 1
Video showing the ECoG electrode placement in 3D for each of the three participants.

Dots indicate ECoG channels. Red dots indicate channels that were responsive to the music input. The corresponding interactive Matlab 3D plots were also uploaded.

Video 2
Video showing the ECoG electrode placement in 3D for each of the three participants.

Dots indicate ECoG channels. Red dots indicate channels that were responsive to the music input. The corresponding interactive Matlab 3D plots were also uploaded.

Video 3
Video showing the ECoG electrode placement in 3D for each of the three participants.

Dots indicate ECoG channels. Red dots indicate channels that were responsive to the music input. The corresponding interactive Matlab 3D plots were also uploaded.

Additional files

Supplementary file 1

Tables indicating the coordinates (MNI) of the intracranial electrodes for each patient.

https://cdn.elifesciences.org/articles/51784/elife-51784-supp1-v1.docx
Supplementary file 2

Matlab interactive 3D plots showing the ECoG electrode placement for the first ECoG patient.

Dots indicate ECoG channels. Red dots indicate channels that were responsive to the music input.

https://cdn.elifesciences.org/articles/51784/elife-51784-supp2-v1.zip
Supplementary file 3

Matlab interactive 3D plots showing the ECoG electrode placement for the second ECoG patient.

Dots indicate ECoG channels. Red dots indicate channels that were responsive to the music input.

https://cdn.elifesciences.org/articles/51784/elife-51784-supp3-v1.zip
Supplementary file 4

Matlab interactive 3D plots showing the ECoG electrode placement for the third ECoG patient.

Dots indicate ECoG channels. Red dots indicate channels that were responsive to the music input.

https://cdn.elifesciences.org/articles/51784/elife-51784-supp4-v1.zip
Transparent reporting form
https://cdn.elifesciences.org/articles/51784/elife-51784-transrepform-v1.docx

Download links