Skip to main content

About trial

The Behaverse Trial Schema defines a set of tidy tables describing trial-level behavioral data — the task-specific aggregates derived from raw events — for cognitive tests and questionnaires. It is the Trials layer of the Behaverse Data Model (BDM).

Motivation

A trial is a single instance of a participant interacting with a task. The information about a trial naturally spans several related entities — the response, the stimuli shown, the options offered, the raw inputs, and the instrument used. Rather than flatten that into one wide table, the trial schema models it as a small set of tidy tables joined by _id foreign keys, so each concern stays normalized and analysable.

The tables

TableFieldsDescription
Response75Main table; one row per response in a trial.
Stimulus19Each stimulus shown during a trial.
Option18Each option a subject could choose from.
Input15Detailed log of inputs/clicks during the trial.
StimulusComponent13Components that make up a stimulus.
OptionComponent14Components that make up an option.
Instrument7The instrument (and its parameterizations) used for acquisition.

Each table is a group in the sidebar; expand one to browse its fields, each of which has its own page.

Conventions

  • A trailing _id denotes a foreign key into the table of that name (e.g. stimulus_id → the Stimulus table).
  • When several entities occur in one trial (e.g. multiple stimuli), their ids/values are concatenated into a single string on CSV export.

Relationship to events

Trial-level data is derived from the raw event stream (the Events layer of the BDM — see the event schema). Events capture what happened moment-to-moment; trials are the task-specific aggregates analysts work with.

Namespace & source

  • Namespace: https://behaverse.org/schemas/trial#
  • Source of truth: field-definitions.yaml (hand-maintained). The docs and the field-definitions.json render contract are generated from it; a per-table schema.json is planned.

Status

Work in progress, relocated from behaverse/data-model. Published as-is while a handful of typing and structure follow-ups are tracked.