Experiments

Experimental design and task structure for the LAION-fMRI dataset.

Experimental Design

Overview

The LAION-fMRI dataset includes data from multiple experimental paradigms designed to investigate neural responses to visual stimuli.

Design Types

The experiments utilize established fMRI design paradigms:

Block Design

  • Sustained presentation of stimulus category

  • Block duration: 12-20 seconds

  • Inter-block rest periods

  • Multiple blocks per run

Event-Related Design

  • Individual trial presentations

  • Variable inter-trial intervals

  • Jittered timing for optimal estimation

  • Randomized trial order

Mixed Design

  • Combination of block and event-related elements

  • Allows investigation of both sustained and transient responses

Task Structure

Experimental Sessions

Each participant completed multiple scanning sessions:

Session Structure:
├── Session 01: Anatomical + Localizer
├── Session 02: Main Experiment Runs 1-4
├── Session 03: Main Experiment Runs 5-8
└── Session 04: Validation Runs
Session structure overview

Session structure overview in the LAION-fMRI dataset.

Run Parameters

Typical Run Specifications:

  • Duration: 6-10 minutes per run

  • Number of trials: varies by experiment (60-120 trials)

  • Repetition Time (TR): specified in acquisition parameters

  • Volumes per run: varies by duration and TR

Trial Structure

Individual trial timing:

Trial Timeline:
├── Fixation: 500ms
├── Stimulus: 500-2000ms (varies by experiment)
├── Response window: 1000-2000ms
└── Inter-trial interval: 2000-6000ms (jittered)

Experimental Conditions

Condition Types

Experiments include various conditions:

Stimulus Categories:

  • Category A (e.g., faces)

  • Category B (e.g., scenes)

  • Category C (e.g., objects)

  • Baseline/Fixation

Task Demands:

  • Passive viewing

  • One-back memory task

  • Category detection

  • Feature discrimination

Attention Conditions:

  • Attended stimuli

  • Unattended stimuli

  • Divided attention

Counterbalancing

Experimental design includes counterbalancing across:

  • Stimulus presentation order

  • Category assignment to conditions

  • Run order across sessions

  • Left/right button mapping

Events Files

Format

Event timing information is stored in TSV (Tab-Separated Values) format following BIDS conventions:

sub-01/func/sub-01_task-nback_run-01_events.tsv

Example Events File

onset    duration    trial_type    stimulus_id    response_time    accuracy
2.0      0.5         face          stim_0001      0.823            1
6.5      0.5         scene         stim_0045      0.912            1
11.2     0.5         object        stim_0123      0.0              0
15.8     0.5         face          stim_0002      0.756            1

The respective events.json files provide detailed descriptions of each column.

{
    "onset": {
        "LongName": "Event onset time",
        "Description": "Time in seconds from the start of acquisition when the event begins",
        "Units": "seconds"
    },
    "duration": {
        "LongName": "Event duration",
        "Description": "Duration of the stimulus presentation",
        "Units": "seconds"
    },
    "trial_type": {
        "LongName": "Trial type",
        "Description": "Category or type of stimulus presented in this trial",
        "Levels": {
            "face": "Human face stimulus",
            "scene": "Natural or indoor scene stimulus",
            "object": "Object stimulus",
            "baseline": "Fixation or rest period"
        }
    },
    "stimulus_id": {
        "LongName": "Stimulus identifier",
        "Description": "Unique identifier for the specific stimulus image presented, references stimuli.tsv"
    },
    "response_time": {
        "LongName": "Response time",
        "Description": "Time in seconds from stimulus onset to participant response. 0 indicates no response or missed trial",
        "Units": "seconds"
    },
    "accuracy": {
        "LongName": "Response accuracy",
        "Description": "Whether the participant's response was correct",
        "Levels": {
            "1": "Correct response",
            "0": "Incorrect response or miss",
            "n/a": "No response required for this trial type"
        }
    }
}

Experimental Protocols

Main Experiment

Objective: Investigate neural responses to different visual categories

Procedure:

  1. Participant positioned in scanner

  2. Structural scan acquisition

  3. Functional localizer run

  4. Main experimental runs (typically 4-8 runs)

  5. Validation run

Instructions to Participants:

Participants were instructed to:

  • Maintain fixation on central cross

  • Respond to target stimuli (when applicable)

  • Stay alert throughout the run

  • Minimize head movement

Localizer Experiment

Purpose: Identify category-selective brain regions

Design:

  • Block design with alternating categories

  • Each category block: 16-20 seconds

  • Multiple repetitions per category

N-back Task

N-back task design overview

N-back task design overview in the LAION-fMRI dataset.

Purpose: Assess working memory and attention

Timing and Synchronization

Stimulus Presentation

Stimuli were presented using:

  • Software: PsychoPy, Presentation, or similar

  • Display: MRI-compatible LCD screen

  • Resolution: 1920x1080

  • Refresh rate: 60 Hz

  • Viewing distance: ~120 cm

Scanner Synchronization

Stimulus presentation was synchronized with MRI acquisition:

  • TTL pulse from scanner triggers experiment start

  • Each volume acquisition logged for timing verification

  • Drift correction applied if necessary

Quality Assurance

Experimental Monitoring

During scanning:

  • Real-time monitoring of participant responses

  • Head motion tracking

  • Attention monitoring via task performance

Post-Scan Verification

After each session:

  • Behavioral data completeness check

  • Timing accuracy verification

  • Response file validation

  • Event file generation and validation