Skip to main content

See it in action: Serial Reaction Time

In a Serial Reaction Time task, a target appears in one of several locations and the participant presses the matching key. It's a classic implicit learning paradigm, and its raw data is a good example of columns that are hard to understand at a glance.

Look at a raw trial below: what is grid? What does target point at? Is rt measured from the start of the trial or the moment the target appears? On their own, the columns don't say.

The metadata tool answers those questions automatically. It reads each row's trial_type, fetches the matching jsPsych plugin (@jspsych/plugin-serial-reaction-time), and pulls the plugin's own documentation for every column into a Psych-DS dataset_description.json.

"Try this sample in the wizard" opens the real tool with this dataset already loaded, so no download or upload is required. It runs entirely in your browser.

Prefer the command line?

Download the sample data above (it saves as task-srt_subject-01.json), then run it through the CLI. Point it at the folder containing that file and you'll get the same dataset_description.json shown below.

Before and after​

On the left is one trial from the raw jsPsych export. On the right is what the tool generated for each of its columns. Almost every column gets a description pulled straight from the plugin, not written by hand. The one exception is plugin_version, which the plugin doesn't document, so it comes out unknown for you to fill in. (grid is shown on one line for readability, and the extra array columns the tool creates for grid and target are left out here; the full file has them.)

sample-data.json (raw export)
[
  {
    "rt": 1383,
    "response": "d",
    "correct": true,
    "grid": [[1, 1, 1, 1]],
    "target": [0, 1],
    "trial_type": "serial-reaction-time",
    "trial_index": 0,
    "plugin_version": "2.1.0",
    "time_elapsed": 1940
  }
]
dataset_description.json (generated)
[
  {
    "@type": "PropertyValue",
    "name": "rt",
    "description": "The response time in milliseconds for the participant to make a response. The time is measured from when the second stimulus first appears on the screen until the participant's response.",
    "value": "number",
    "minValue": 270,
    "maxValue": 1383
  },
  {
    "@type": "PropertyValue",
    "name": "response",
    "description": "Indicates which key the participant pressed.",
    "value": "string",
    "levels": ["d", "k", "s", "j"]
  },
  {
    "@type": "PropertyValue",
    "name": "correct",
    "description": "`true` if the participant's response matched the target.",
    "value": "boolean"
  },
  {
    "@type": "PropertyValue",
    "name": "grid",
    "description": "The representation of the grid. This will be encoded as a JSON string when data is saved using the `.json()` or `.csv()` functions.",
    "value": "array"
  },
  {
    "@type": "PropertyValue",
    "name": "target",
    "description": "The representation of the target location on the grid. This will be encoded as a JSON string when data is saved using the `.json()` or `.csv()` functions",
    "value": "array"
  },
  {
    "@type": "PropertyValue",
    "name": "trial_type",
    "description": "The name of the plugin used to run the trial.",
    "value": "string",
    "levels": ["serial-reaction-time"]
  },
  {
    "@type": "PropertyValue",
    "name": "trial_index",
    "description": "Join key referencing the position of an enclosing array element (0-based index).",
    "value": "number",
    "minValue": 0,
    "maxValue": 15
  },
  {
    "@type": "PropertyValue",
    "name": "plugin_version",
    "description": "unknown",
    "value": "string",
    "levels": ["2.1.0"]
  },
  {
    "@type": "PropertyValue",
    "name": "time_elapsed",
    "description": "The number of milliseconds between the start of the experiment and when the trial ended.",
    "value": "number",
    "minValue": 1940,
    "maxValue": 15064
  }
]

grid and target are the ones you'd never guess cold, and they're exactly the ones the tool describes for you. You can also open the full generated file.

Nested columns become their own files

grid and target are arrays, so the tool also extracts them into companion _measure-*_data.csv files alongside the main table, keeping the dataset flat and Psych-DS valid. Those companion files link back to each trial by trial_index, which is why trial_index is described as a join key above.

Try it on your own data​

Ready to do this with your own experiment? Open the wizard and point it at your data folder. Everything runs locally in your browser.