HED Guidelines¶
HED (Hierarchical Event Descriptors) is a standardized vocabulary for describing events in behavioral experiments. This guide covers HED usage within AGI.
What is HED?¶
HED provides a hierarchical, controlled vocabulary for describing experimental events in a machine-readable format. Unlike free-text descriptions, HED annotations:
- Have consistent semantics across studies
- Are machine-parseable for automated analysis
- Support complex event structures via grouping
- Enable cross-study comparisons
HED Structure¶
Tags¶
HED uses a hierarchical tag structure:
Event
├── Sensory-event
│ ├── Visual-presentation
│ └── Auditory-presentation
├── Agent-action
│ ├── Participant-response
│ └── Experimenter-action
└── Data-feature
Basic Examples¶
| Description | HED Annotation |
|---|---|
| A visual stimulus appears | Sensory-event, Visual-presentation |
| Participant presses a button | Agent-action, Press, (Item-physical/Button) |
| Red circle on left side | (Red, Circle, (Left-side)) |
Grouping with Parentheses¶
Parentheses group related tags:
This means: a red circle on the left side of the screen.
Complex Events¶
For complex events, use multiple tag groups:
This describes: a visual presentation of a male face image that is the target.
HED Schema Versions¶
AGI tools support multiple HED schema versions:
| Version | Status | Notes |
|---|---|---|
| 8.3.0 | Stable | Default for most tools |
| 8.4.0 | Stable | Enhanced visual vocabulary |
| 8.5.0 | Development | In testing |
Specify the schema version in your requests:
Best Practices¶
Do¶
- Use specific tags when available (e.g.,
Faceinstead ofImage) - Group related tags with parentheses
- Include spatial information (Left, Right, Center)
- Add temporal information when relevant (Duration)
- Validate all annotations before use
Don't¶
- Use free text where HED tags exist
- Create inconsistent groupings
- Omit essential event features
- Mix schema versions within a dataset
Validation¶
Always validate HED strings:
Common Patterns¶
Visual Stimuli¶
# Image presentation
(Sensory-event, Visual-presentation, (Image))
# Specific image content
(Sensory-event, Visual-presentation, (Image, Face, (Male, Adult)))
# With spatial location
(Sensory-event, Visual-presentation, (Image, Face), (Left-side))
Participant Actions¶
# Button press
(Agent-action, Participant-response, Press, (Item-physical/Button))
# Keyboard response
(Agent-action, Participant-response, Press, (Item-physical/Key))
# Eye movement
(Agent-action, Participant-response, Move-eyes, (To, Item))
Temporal Events¶
# Onset
(Sensory-event, Visual-presentation, (Onset))
# Offset
(Sensory-event, Visual-presentation, (Offset))
# Duration
(Sensory-event, Duration/500 ms)
Resources¶
- HED Documentation: hedtags.org
- HED Schema Browser: hed-specification
- HED Validator: hed-python
- HEDit: AI-powered annotation generation