Jump to: navigation, search


HED 2.0

Current page describes HED 1.0 which is now deprecated and replaced by HED 2.0. Please visit www.hedtags.org for latest HED information.


There have been multiple efforts to standardize terminologies used in cognitive paradigms by developing formal database ontologies (NEMO, BrainMap, CogPO and NeuroLex) but none of these is very often used to describe EEG events in research and publications. We believe a key missing factors is the ease of use of these ontologies: although formal ontologies in OWL format are elegant and can be readily processed by computers, their apparent complexity to human eyes does not encourage their casual use by (non-programmer) researchers.

To address this problem we have built upon the popular tagging method, extensively used on the web (particularly on social sites, e.g. image tags on Flicker, video tags on YouTube...) -- a hierarchy of standard, but extensible, descriptors for EEG events. Our Hierarchical Event Descriptor (HED) ontology generation system is a base set of descriptor tags, in part adapted from the BrainMap and CogPO ontologies, and organized hierarchically. HED tags can be used to describe many types of EEG experiment events in a uniform, easily extensible, and both human and machine readable manner. The main contribution of HED is to offer a user-friendly interface for using the underlying ontology in EEG acquisition and analysis workflow.

HED Resources

HED 1.31 Specification (PDF)

You can download a collection of HED tools from HeadIT Tools repository on GitHub.

Prof. Kay Robbins and her group at UTSA have created a nice HED tagging GUI and EEGLAB plugin, named CTAGGER Website, Matlab code repository, Java layer code repository

For a well-documented collection of EEG studies that include HED and ESS meta-data please visit http://headit.org

HED whitepaper

Hierarchical Event Descriptor (HED) Tags

In the same way that we tag a picture on Flicker, or a video clip on Youtube (e.g. cat, cute, funny), we can tag EEG experimental event types used in event-related EEG research. In this document we introduce Hierarchical Event Descriptors (HED), a set of descriptor tags partially adopted from the BrainMap ontology (brainmap.org) and organized hierarchically. HED tags can be used to describe many types of EEG experiment events in a uniform, extensible, and machine readable manner.

The hierarchical structure of the HED tags makes it easy to search across variations of the same type of event across studies (supporting EEG data meta-analysis), while preserving a more detailed description of each event type. For example, an event marking the presentation of a visual feedback stimulus, in EEG study A, may present a red circle to the participant on a black screen background, while in study B the visual feedback stimulus is a blue rectangle on a white screen background. In HED syntax, these event types can be described as:

Study A:
 Stimulus/Visual/Background/Uniform Color/Black
Study B:
 Stimulus/Visual/Background/Uniform Color/White

These descriptors explicitly capture some salient commonalities across and differences between the two event types. If these feedback events are accompanied by an auditory beep at 500 Hz with 25-dB amplitude, the following tag may be added:

Stimulus/Auditory/Tone/Ramp Down/10-ms.

While higher levels of the HED hierarchy are fixed (revised infrequently in discrete version levels based on community feedback), lower levels can be extended, with no restriction, to describe any event type in the desired details. For example


may be extended to


to provide more information about the circle.

HED can be used to describe EEG ‘conditions’ (i.e., sets of similar events) extracted from experimental paradigms used in EEG studies. In addition to describing stimulus-presentation events, HED can also be used to describe subject motor responses (e.g., button press, swipe, etc.), subject states (e.g., Drowsy, Attending Visual), and combinations of these two (e.g., button press while subject is drowsy). Since HED can be easily interpreted by computers, it can facilitate search and meta-analysis of event-related EEG dynamics across multiple studies.

HED tags may also be used to organize events from an study in a logical hierarchy so they can be more easily analyzed. For example several event subtypes may be aggregated into a more general type and then compared to other event types.

We have developed a companion XML-based specification, called ESS (EEG STUDY Schema) to hold all the information necessary to analyze an EEG study, e.g subject gender, handedness, age and group associations, task and paradigm description..., in a format that is both machine and human understandable (the XML file is automatically formatted and become a readable report about the EEG study). To achieve this goal, ESS relies on HED descriptions of experimental events and tasks that are embedded into ESS XML document.

How to use HED

To describe a ‘time-locking event’ (an EEG experiment event type with a well-defined onset latency), look under Time-Locked Event section below and find nodes that best describe the event. Form each tag by separating different levels with a forward-slash (‘/’) character. Add any number of sub-level descriptions to these tags to make them more precise (as exemplified above). Placement of / at the beginning and end of HED tags is optional (Stimulus/Visual = /Stimulus/Visual = /Stimulus/Visual/). Finally, join your tags with commas ‘,’ or semicolons ‘;’ to form a full HED string.

There are few locations in the hierarchy in which new nodes may be added at the same level of the declared nodes (otherwise, new nodes cannot be added at the same level). For example you can extend Stimulus/Visual/Shape/Ellipse/Circle/ to Stimulus/Visual/Shape/Ellipse/Circle/Filled but you should not extend Stimulus/Visual/Shape/Ellipse/ to Stimulus/Visual/Shape/Egg (Egg does not exist under Stimulus/Visual/Shape and this tag does not allow for custom first-level child nodes]. On the other hand, Stimulus/Auditory/Tone does allow adding nodes immediately under, so Stimulus/Auditory/Tone/Square Wave is valid.

To describe subject states (e.g. drowsy, tired) start from /State node. You may combine subject state tags with Time-locked tags (e.g. visual stimulus when subject was drowsy).

To describe an experimental paradigm, use any of the nodes under /Paradigm. Add tags that are common across all events under /Context. For example, Context/Afternoon [for experiments always conducted in the afternoon] Context/Moving Vehicle [for experiments always conducted in a moving vehicle]

The HED Node Hierarchy section below shows the current fixed top-level HED nodes. The root (Time-Locked Event) node is assumed by default and does not need to be included. Additional explanation is provided for some nodes and placed in square brackets (‘[ ]’).