Programming behavioral experiments · Meeting 5

Pointer data and deployment

A complete pilot, a verified save path, and realistic extensions

Bartosz Maćkiewicz

Concept · 05

Today's outcome

Record interpretable pointer events and distinguish a completed task from a confirmed saved result.

Artifact: a mouse-tracking pilot and a project release checklist.

Concept · 05

Keep the advanced feature small

We add one new measure to a familiar decision task.

Mouse tracking uses a supplied extension. Webcam gaze and community extensions are optional discussion.

Final projects do not need an advanced measure for full marks.

Concept · 05

What is the participant doing?

Click the center Start button.
Read a word.
Move to Living or Nonliving and click.
Return to the same start location for the next word.

A fixed starting action makes trajectories more comparable.

Concept · 05

Register the extension globally

const jsPsych = initJsPsych({
  extensions: [{
    type: jsPsychExtensionMouseTracking,
    params: { minimum_sample_time: 16 }
  }]
});

16 ms is a minimum interval, not a guaranteed sampling rate.

Concept · 05

Attach tracking to the decision trial

extensions: [{
  type: jsPsychExtensionMouseTracking,
  params: {
    targets: ["#left-btn", "#right-btn"],
    events: ["mousemove", "mousedown", "mouseup"]
  }
}]

Each selector should identify one element.

Experiment laboratory · 05

Button labels and response indices

Open the lab in its own tab · Predict → edit → run → inspect · The preview captures response keys.

button_html: (choice, index) => {
  const id = index === 0 ? "left-btn" : "right-btn";
  return `<button id="${id}" class="jspsych-btn">
    ${choice}</button>`;
}

button_html builds each button. The saved response is its zero-based index; IDs let the tracking extension find the targets.

Experiment laboratory · 05

Live checkpoint L1: enable tracking

Open the lab in its own tab · Predict → edit → run → inspect · The preview captures response keys.

Add the global extension and trial-level parameters. Use the supplied button IDs.

Move the pointer before clicking. Finish and download JSON.

Concept · 05

Read the measurement

mouse_tracking_data contains event records:

x, y: viewport coordinates in pixels;
t: time since the tracked trial started;
event: the recorded event type.

mouse_tracking_targets contains element rectangles.

Reveal the reasoning

x and y are viewport coordinates, and t is elapsed time within the tracked trial. A 16 ms minimum interval does not guarantee exactly 62.5 samples per second.

Try it · 05

Live checkpoint L2: make it interpretable

Find the decision row, not the start-click row.

Check events, time order, and the two target rectangles. Explain why an empty trajectory differs from a missing field.

JSON preserves nested arrays more naturally than a flat table.

Concept · 05

What this does not measure directly

Same response, direct path.

Same response, curved path.

Illustrations, not participant data. Geometry and device can change a path.

A pointer path is not a gaze trace or a direct reading of thought.

Control the start position, record target geometry, and consider the device.

Read the evidence · 05

A complete data workflow

RunData exist in browser memory.

SaveDownload locally or send through the runner.

RetrieveOpen the result and check the final row.

Completion is a screen state. Saving is an operation. Retrieval is evidence.

A static file server does not become a results database.

Concept · 05

Local downloads

The supplied completion panel offers CSV and JSON.

Download, reopen, and check the final trial. The files are on the current computer, not on a server.

Use this route when no classroom runner is available.

Concept · 05

JATOS: start only when ready

jatos.onLoad(() => {
  jsPsych.run(timeline);
});

Use the supplied JATOS entry page through the runner. The local entry page remains suitable for development.

Read the evidence · 05

JATOS: send data when the task ends

on_finish: () => {
  jatos.startNextComponent(jsPsych.data.get().json());
}

This sends the results and advances to the next component, or finishes the study. Let the platform manage the transfer.

Try it · 05

Live checkpoint L3: retrieve a pilot

Cognition.run: complete the task and open its saved results. JATOS: run the prepared component and retrieve its result.

Check the final trial and the expected task-row count. Local download is the fallback, not an upload to a platform.

Concept · 05

Cognition.run is a separate integration

Its documented workflow handles completion and result storage.

Follow its current platform instructions, check package versions, and retrieve a pilot. Do not paste JATOS-specific code into it.

No paid account is needed for the local course activities.

Build and verify · 05

Workshop: semantic mouse task

Enable tracking on four decision trials.
Verify zero-based choice scoring and item labels.
Preserve the common start action and geometry.
Save and inspect one complete result.

Then spend the protected clinic time on your final project.

Concept · 05

WebGazer: calibration is not validation

Camera initialization → calibration → validation → task.

A coarse area-of-interest question and letter-level gaze timing make very different demands on a measurement system.

Use the observation worksheet; no camera is required for credit.

Concept · 05

Community extension appraisal

Ask about compatibility, permissions, data fields, units, maintenance, licensing, and a fallback.

A demo that runs once does not establish measurement validity. Use the ecosystem worksheet for one concrete example.

Concept · 05

Final project review

Can another person:

run the extracted package;
explain one row and each condition;
reproduce a failure test;
retrieve the data;
identify the prototype’s limits?

Think first · 05

Exit ticket and completion

Explain the difference between a local download, a submitted request, and confirmed server storage.

Final project: 14 days after this meeting. Use the rubric, pilot log, and README template before submitting.