How to use the catalog

This page is for a reader who wants to do something with the catalog: identify the task in an experiment, tag a dataset, look up a process, or propose a change.

Finding a task

Three routes lead to a task page.

  • By name or alias. The search box in the sidebar indexes canonical names, aliases, definitions and variation names, so “CWIT”, “Stroop” and “color-word interference” all reach the Stroop Color-Word Task.

  • By what the participant does. The task index groups the 103 tasks into paradigm families such as conflict and interference tasks, span tasks and economic games. Each family has a scope sentence and a table of its tasks.

  • By process. Each entry on a process category page lists the tasks that engage the process, and the task-process links page gives the whole matrix.

The alphabetical list is the fallback when you know the name but not the family.

Reading a task page

Take the Stroop Color-Word Task as the example.

Identifier and aliases. The identifier hedtsk_stroop_color_word is the value to record when tagging. The aliases are the other names the literature uses for the same procedure.

Inclusion test. The three rows decide whether a particular experiment is an instance of this task:

  • Procedure: Color words (RED, BLUE) are printed in incongruent ink colors; participants name the ink color while ignoring the word.

  • Manipulation: Congruency (congruent, incongruent, neutral); proportion congruent; response modality (vocal, manual); stimulus type (classic, spatial, numerical).

  • Measurement: Stroop interference effect (incongruent - congruent RT); facilitation (neutral - congruent); error rate; conflict adaptation (Gratton effect).

An experiment is an instance when its procedure matches, it manipulates at least one of the listed variables, and it records at least one of the listed measures. If the procedure does not match, it is a different task even if the construct is the same. If the procedure matches but the experiment differs in a way listed under Variations, it is an instance of that variation.

Variations. Each row names a version that changes what the participant experiences or does, describes it, and says why it counts. Variations have identifiers of the form hedvar_<task>__<variation>, so a dataset can be tagged with a variation when the distinction matters. Things that are deliberately not variations, such as adding a recording modality or testing a different population, are listed in the task criteria.

Cognitive processes. The processes the task is designed to engage, each linked to its definition. These are the process tags a dataset using this task would carry.

References. Key references established the paradigm; recent references are reviews or influential recent papers. Citations are reproduced as the source supplies them.

External links. The corresponding Cognitive Atlas entry, with a qualifier when the match is close rather than exact.

Reading a process page

Processes are presented one category to a page. The category page opens with the category’s scope, what is out of scope, any open issue about its boundaries, and a summary table of its processes. Each process entry then gives the identifier, aliases with a note on the terminological distinction, the definition, the tasks that engage it, and references. A process with no linked tasks says so; it is kept because the catalog may grow a task for it.

Tagging a dataset

The catalog’s identifiers are stable enough to record now, even though how they will be expressed in HED annotations and dataset metadata is still being decided. For a dataset:

  1. Identify the task with the inclusion test, and the variation if one applies.

  2. Record the task identifier (hedtsk_...) and, if applicable, the variation identifier (hedvar_...).

  3. Record the process identifiers (hed_...) listed on the task page. Add processes the task page does not list only if the experiment’s own design targets them; the task page lists what the paradigm is designed to probe, not everything it touches.

  4. If the experiment is a task the catalog does not have, or a variation a task page does not list, propose it (next section) rather than forcing a fit.

Resting runs, fixation blocks and questionnaire blocks are tagged the same way, with the identifier of the matching pseudo task: hedtsk_rest with the eyes-open or eyes-closed variation, hedtsk_fixate, or hedtsk_take_questionnaire. A pseudo task carries no process identifiers, so there is nothing to record for step 3.

Proposing a change

The catalog is curated continuously and depends on its users to grow. The repository’s contributing guide describes the process; in short, post an issue at https://github.com/hed-standard/hed-task/issues for any of the following.

  • A new task. Give the canonical name and aliases, a procedure, a manipulation and a measurement in the form the inclusion tests use, the processes it engages, and one or two references. Say why it is not a variation of an existing task.

  • A new variation. Name the parent task, describe what changes in what the participant experiences or does, and say why that is not one of the excluded kinds of change (measurement modality, analysis, population, stimulus swap, and so on).

  • A new process. Say when in a trial it happens, what elicits it, how it is measured, and which category it belongs in. Say why it is not an alias of an existing process.

  • A correction. Anything wrong on a page: a definition, a reference, a link, a family assignment, an Atlas mapping. Quote the page and the text.

  • A disagreement with a rule. The criteria pages state the rules as they stand; arguments for changing one are welcome.

The source for the site is at https://github.com/hed-standard/hed-task. The pages are generated from the catalog data, so corrections are applied to the data, not to the pages.