Vision for action

# README — Vision for Action fMRI Dataset

## Dataset Overview
This 3T fMRI dataset was collected for the **Vision for Action** project to investigate how the human visual system represents **locomotive action affordances**—the actions that environments afford for navigating through them (e.g., walking, biking, climbing).

The dataset accompanies the publication:

**Bartnik et al. (2025). _Representation of locomotive action affordances in human behavior, brains, and deep neural networks_. PNAS.**

Data were collected at the **FMG Research Lab**, University of Amsterdam (REC, Amsterdam, The Netherlands).

---

## Contents

Each participant completed **four scanning sessions**:

### **Session 1 (ses-001): Functional Localizer + Scene pRF**
- Six runs of a **category localizer** (block design) with six object/scene categories.
- One or more **population receptive field (pRF)** mapping runs.
- One **T1-weighted anatomical scan**.
- BOLD fMRI acquisitions and corresponding fieldmaps.

### **Sessions 2–4 (ses-002, ses-003, ses-004): Main Vision-for-Action Tasks**
Across these three sessions, participants performed three different tasks using the same set of 90 scene images:

1. **Locomotive Action Affordance Task**
Participants indicated which locomotive actions the environment afforded (walking, biking, driving, climbing, swimming, boating).

2. **Object Presence Task**
Participants selected which object categories were present (building/wall, plant/tree, furniture, road/street, body of water, stones).

3. **Fixation Color Task (Control)**
Participants reported the color of the fixation cross (blue, red, orange, purple, yellow, cyan), independent of scene content.

The order of tasks was **counterbalanced across subjects**.

Each of these sessions contains:
- Six functional **BOLD** runs (event-related design)
- A **T1-weighted anatomical scan**
- Optional fieldmaps
- Associated `*_events.tsv` files

---

## Stimuli

A subset of **90 natural scene photographs** was selected from a larger set of 231 images to:

- Sample indoor, outdoor-natural, and outdoor-manmade environments
- Maximize variance in **locomotive action affordance ratings**
- Cover the first two principal components of affordance space

Each scene was shown for **1 s**, followed by a jittered ISI of **2.2–5.4 s**.
Stimuli subtended approximately **20° of visual angle**.

---

## fMRI Acquisition

All data were collected on a **Philips Achieva 3T** scanner using a **32-channel SENSE head coil**.

### **Functional Localizer (ses-001)**
- EPI sequence
- TR = 2000 ms, TE = 27.63 ms
- 36 slices, 3×3×3 mm
- 140 volumes per run

### **Main Task Sessions (ses-002 to ses-004)**
- Multiband EPI
- TR = 1600 ms, TE = 30 ms
- 56 slices, 2×2×2 mm
- Multiband factor = 4
- 281 volumes per run
- Acquisition label in BIDS: `acq-Mb4Mm2Tr1600`

### **Structural Imaging**
A **T1-weighted anatomical scan** (1 mm isotropic) was collected at the start of each session.

Raw BIDS data provided here are **unprocessed**.

---

## Events Files

For each functional run in sessions 2–4:

- An `*_events.tsv` file is provided
- Contains onset times, durations, and trial-type information
- Response prompts and button presses are included where applicable

The event structure follows the task design described in the corresponding publication.

---

## Behavioral Task Structure

Each run contained:
- **90 trials** (all stimuli presented once per run)
- Randomly interspersed **response screens** (3.8 s)
- Tasks allowed **multiple responses** in the affordance and object conditions

The fixation task required a single color judgment on each response screen.

---

## Anonymity

To comply with University of Amsterdam regulations:

1. All participants signed consent permitting data sharing.
2. No mapping exists between personal information and subject IDs.
3. All structural data have been **defaced** using `pydeface`.
4. Subject identifiers (`sub-XXX`) were randomly assigned.

---

## Citation

If using this dataset, please cite:

**Bartnik, C. G., Sartzetaki, C., Puigseslloses Sanchez, A., Molenkamp, E., Bommer, S., Vukšić, N., & Groen, I. I. A. (2025). Representation of locomotive action affordances in human behavior, brains, and deep neural networks. _Proceedings of the National Academy of Sciences_.**

---

## Contact

For questions about the dataset, please contact:
**I. I. A. Groen** — [e-mail]