Intonational Speech Prosody Encoding in Human Auditory Cortex

This dataset contains data and results associated with the manuscript, "Intonational Speech Prosody Encoding in Human Auditory Cortex", as well as code used to analyze the data and generate the figures of the manuscript. 

intonatang-2017.7.17.tar.gz contains the entire project, including neural data, stimulus sound files, analysis code, and documentation. The code and documentation can also be viewed on Github at https://github.com/ChangLabUcsf/intonatang.

We additionally included each block of neural data and a zipped file containing the experimental, acoustic stimuli as separate files in this dataset. The data comprise three experiment types, "Speech", "Non-speech control", and "Non-speech missing f0 control". The neural data files are named with a subject identification number and a block number.

Speech:

1. EC113_B13
2. EC113_B20
3. EC113_B21
4. EC118_B3
5. EC118_B7
6. EC118_B13
7. EC122_B30
8. EC122_B40
9. EC122_B43
10. EC122_B53
11. EC123_B4
12. EC123_B5
13. EC123_B10
14. EC125_B13
15. EC125_B1044
16. EC129_B10
17. EC129_B16
18. EC129_B37
19. EC131_B47
20. EC131_B48
21. EC137_B7
22. EC137_B10
23. EC142_B36
24. EC142_B37
25. EC143_B9
26. EC143_B11
27. EC143_B13

Non-speech control:

1. EC122_B33
2. EC122_B45
3. EC123_B11
4. EC123_B16
5. EC125_B30
6. EC129_B40
7. EC129_B42
8. EC131_B54
9. EC131_B59

Non-speech missing f0 control:

1. EC137_B9
2. EC137_B11
3. EC142_B38
4. EC142_B40
5. EC143_B10
6. EC143_B12
7. EC143_B14

These .mat files contain the following variables: 

•  badTimeSegments - (n_badTimeSegments x 2)
  • This variable contains manually marked time segments containing epileptiform, electrical, or movement artifacts. Each row indicates one bad time segment, with the start time and end time in seconds.
• bcs - (n_bcs)
  • This array contains manually marked bad channels. These channels from the ECoG grid either had continuous epileptiform activity or signal indistinguishable from noise. The channels are indexed from 0.
• ECXXX_BXX_hg_100Hz - (n_chans x n_timepoints)
  • This variable contains the mean high-gamma analytic amplitude signal for each channel, sampled at 100Hz. The mean is taken across 8 bands between 70-150Hz. The variable name contains the subject number, ECXXX, and block number BXX. 
• ECXXX_BXX_log_hg_100Hz - (n_chans x n_timepoints)
  • This variable contains the mean of the natural logarithm of the high-gamma analytic amplitude signal for each channel. The log is taken for each of the 8 bands between 70-150Hz and then averaged.
• experiment
  • This variable holds the experiment type and is either "Speech", "Non-speech control", or "Non-speech missing f0 control".
• sentence_numbers - (n_trials)
  • The integers in this array are the sentence number condition for each trial in this block. The sentence number conditions depend on the experiment type.
  1. For the "Speech" experiment, the four sentences indicated by 1, 2, 3, and 4 are "Humans value genuine behavior", "Movies demand minimal energy", "Lawyers give a relevant opinion", and "Reindeer are a visual animal".
  2. For the "Non-speech control" experiment, the sentence number conditions indicate which sentence from the main experiment the amplitude contour for the control stimuli came from. A sentence number of 5 means that the amplitude contour was flat. 
  3. For the "Non-speech missing f0 control", the sentence number holds information about the composition of the stimulus (which harmonics were present), whether noise was added, and how much the pitch range was stretched. 
     • 0: 4h + 5h + 6h, no noise, stretch = 1
     • 1: f0 + 2h + 3h, no noise, stretch = 1
     • 2: 4h + 5h + 6h, noise, stretch = 1
     • 3: 4h + 5h + 6h, noise, stretch = 0.5
     • 4: 4h + 5h + 6h, noise, stretch = 2
• sentence_types - (n_trials)
  • The sentence type is the intonation contour condition. Across all experiment types, a sentence type of 1 is Neutral, 2 is Question, 3 is Emphasis 1, and 4 is Emphasis 3.
• speakers - (n_trial)
  • The speaker conditions depend on the experiment.
    • The speaker condition for the "Speech" experiment is an integer between 1 and 3. 1 is the low-formant, low-pitch male speaker. 2 is the high-formant, high-pitch female speaker. 3 is the low-formant, high-pitch female speaker. The absolute pitch values of speakers 2 and 3 match, while the formant values of speaker 1 and 3 match.
    • The speaker condition for both of the two non-speech experiments are either 1 or 2. 1 means low absolute pitch (male) and 2 means high absolute pitch (female).
• stims - (n_trials)
  • This array holds the stimulus name that was played for each trial. The names refer to the wav files in the tokens, tokens_nonspeech, and tokens_missing_f0 folders, for the "Speech", "Non-speech control", and "Non-speech missing f0 control" experiments, respectively.
• times - (n_trials)
  • This array contains the onset times of each trial in seconds.