Discrimination-Estimation Task (ds005381)

This OpenNeuro release (ds005381) holds the MRI data behind a Seoul National University study of how categorical decisions interact with visual working memory. In the discrimination-estimation task, participants first judged a remembered grating orientation against a reference and then reported the orientation itself. The accompanying Neuron paper uses behavior, fMRI and recurrent network models to argue that memory drifts toward a few attractor states and that the earlier decision steers this drift. The data are shared under a CC0 waiver.

Composition

The release has 50 young adults: 30 women and 20 men aged 19 to 32 years (mean 24.2). The participants table gives age, sex and ocular dominance. Most subjects have four sessions: a structural session with a 3D T1-weighted and a T2-weighted scan, followed by fMRI sessions that hold 3 to 10 task runs each, retinotopic mapping runs and resting-state runs, each with a 2D inplane T1-weighted reference. Task runs come with event files that list the stimuli and the responses. In total the raw tree has 1,220 BOLD runs.

Acquisition

All scans come from one Siemens TrioTim 3T scanner at the Seoul National University imaging center, with a 32-channel head coil according to the sidecars. The structural T1-weighted image is an MPRAGE with 0.85 mm slices. Task BOLD runs use gradient-echo EPI with a repetition time of 2 s, an echo time of 30 ms and 2.3 mm slices. DICOM files were converted with dcm2niix.

Annotations

There are no manual labels. A derivatives folder holds fMRIPrep 20.2.0 outputs, run with field-map-free distortion correction, and the FreeSurfer reconstructions from that pipeline.

Known limitations

• Single site and scanner, and a small sample of young adults.
• The release does not describe health screening, so no condition is recorded here.
• No field maps are included, which is why the preprocessing used field-map-free distortion correction.
• One subject lacks retinotopy runs and one lacks resting runs.