Dataset

Brody Lab Poisson Clicks Task Dataset Rats 2009 to 2024 Parsed

Parsed behavioral data from 515 male Long-Evans rats trained in the Brody Lab at Princeton University on the Poisson Clicks auditory evidence-accumulation task. The archive contains one MATLAB file per rat with parsed non-violation, non-frozen trials and associated stimulus metadata.

Access notes

The Zenodo archive is 8.1 GB. It is acceptable to download locally for MVP development under ignored `data/raw/`, while normal validation and tests should still run without the archive. Real ingestion operates on a local `.mat` file extracted from the archive.

Data formats

  • MATLAB .mat files
  • derived canonical trial tables to be generated

Expected trial-table mapping

Canonical fields the harmonizer expects this dataset to populate.

Canonical field Source mapping
choice gr mapped as 0 left, 1 right.
correct hh mapped as 0 incorrect, 1 correct.
evidence_strength Absolute value of nR minus nL.
feedback hh mapped to reward/error for the parsed behavioral outcome.
prior_context ga represented as a gamma context where available.
response_time Not available in the parsed schema described on Zenodo.
session_id Parsed rat-level file identifier until day/session breaks are available.
stimulus_value nR minus nL, the signed right-minus-left click count difference.
subject_id Rat identifier inferred from the source `.mat` filename.
task_variables left/right click counts, click-count difference, stimulus duration, gamma, reward gamma, task type, and left/right click-time vectors from bt.
trial_index Trial index within the parsed rat-level vector.

Caveats

  • The parsed archive does not include all trials, violation trials, or trials longer than 1 s.
  • Day/session breaks are not identified in the parsed release.
  • The full archive is large and should remain under ignored local raw-data storage or external storage.
  • The archive uses ZIP64/Deflate64 compression; macOS `unzip` and Python `zipfile` may not extract it.

Papers

Findings

Curves attached to this dataset on shared signed-evidence axes.

See all findings overlaid on shared axes.