This dataset contains multi-unit spike recordings from the auditory cortex of Squirrel monkeys, collected during passive listening tasks. Monkeys were presented with a variety of auditory stimuli, but this documentation focuses on two key stimulus classes used in our study:

- TIMIT: English sentences from the TIMIT corpus

- mVocs: Monkey vocalizations (e.g., grunts, screams)

The dataset is organized as follows:

root/

├── stimuli/ # Metadata about the auditory stimuli

├── sessions/ # One subdirectory per recording session

│ ├── 180413/

│ │ ├── *_MUsp.mat # One file per recording channel

│ │ └── ... # Additional channel files

│ ├── 180420/

│ └── ... # More sessions

├── session_metadata.yml # Annotations about sessions (location, area, hemisphere, etc.)

├── LICENSE.txt # Dataset license

└── README.md # This file

- stimuli/ contains:

  - out_sentence_details_timit_all_loudness.mat

  - SqMoPhys_MVOCStimcodes.mat

  - MonkVocs_15Blocks.wav

- Each session directory (e.g., 180413/) contains:

  - Multiple '*_MUsp.mat' files (each corresponding to a recording channel)

- 'session_metadata.yml' includes:

  - Session-level annotations (brain area, hemisphere, bad session exclusions)

  - Stimulus repetition counts

  - 2D coordinates for each session

Each *_MUsp.mat file contains spike data from one recording channel. The following variables are populated when the file is loaded:

The spike struct contains spike-level data. Key fields:

- events: s × t matrix of spike waveforms  

  - s: number of detected spikes  

  - t: number of samples per waveform  

- spktimes: vector of spike times (in seconds from recording start)

- amStimcode, fmStimcode, dmrStimcode, mVocStimcode, timitStimcode: vectors of length s, each specifying the stimulus played at the time of the spike

- trial: vector of length `s`, mapping each spike to the trial in which it occurred

The 'trial' struct contains trial-level metadata. Key fields:

- stimon: vector of stimulus onset times (in seconds)

- amStimcode, fmStimcode, dmrStimcode, mVocStimcode, timitStimcode: stimulus code vectors, one per trial

> Note: All stimulus and trial IDs follow MATLAB-style indexing (i.e., start from 1).  

> When working in Python, make sure to adjust for zero-based indexing if needed.

TIMIT stimuli consist of English sentences used during stimulus playback.

- Metadata is stored in: stimuli/out_sentence_details_timit_all_loudness.mat

- Main variable: sentdet — a list of structs (one per stimulus)

Each element in sentdet contains the following key fields:

- sound: waveform (numpy array)

- soundf: sampling rate (Hz)

- duration: total stimulus duration including silence (in seconds)

- befaft: tuple (bef, aft) specifying silence before and after the sentence

Monkey vocalizations (e.g. grunts, screams) were played as naturalistic stimuli.

- stimuli/SqMoPhys_MVOCStimcodes.mat  

  - mVocsStimCodes: list of stimulus IDs (MATLAB indexing)

- stimuli/MonkVocs_15Blocks.wav  

  - A .wav file concatenating all monkey vocalizations with silent gaps

This dataset is shared under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

This dataset has been used in the following studies:

- Ahmed, B. et al. (2025)  

  Deep Neural Networks Explain Spiking Activity in Auditory Cortex.

  PLOS Computational Biology (In press)

- Downer, J. D., Bigelow, J., Runfeldt, M., & Malone, B. J. (2021)

  Temporally Precise Population Coding of Dynamic Sounds by Auditory Cortex

If you use this dataset, please cite the dataset itself:

> Multi-channel auditory cortex electrophysiology in squirrel monkey  

> doi:10.5281/zenodo.16175377