Meeting 4 reference: Reliability and participant experience

What you should be able to do

Four-image shape pilot plus a documented normal and failure path. Explain the logic and verify at least one normal and one edge case, using the actual saved data.

Concepts in context

Participant instructions should specify the target, response mapping, speed/accuracy emphasis, deadline, practice, and how to stop. Ask another person to explain the task without coaching. A layout can look polished while still communicating the wrong decision.

Use semantic HTML and a restrained CSS style. Control image size and stimulus geometry deliberately; changing them between conditions may introduce a confound. Pixels are not visual angle and identical CSS does not make different displays physically identical. Keep keyboard focus visible and provide alternatives where color or input method is not essential to the question.

Preload an explicit list derived from the stimulus table when image paths are dynamic. The preload plugin should run before image trials, have a finite maximum loading time, and stop helpfully on failure. The example does not continue after an error. Test a broken path and restore it. Relative paths resolve from the page; source packaging must preserve or rewrite them.

The browser check tests a minimum viewport and excludes detected mobile devices. It does not prove that the participant has a keyboard, is attentive, or meets every inclusion criterion. Select thresholds for the actual task, explain exclusion, and test resize behavior. Do not copy a threshold into every project without a reason.

Preloading addresses asset readiness, not every source of timing variability. Hardware, refresh rates, browsers, operating systems, and input devices matter. A requested duration in code is not a direct measurement of physical presentation.

A pilot log records expected behavior before the test, actual observations, evidence, and fixes/retests. A failed test can be valuable evidence if honestly diagnosed. Open exported files and inspect their final rows; a download notification alone is not a complete data check.

What to check in your implementation

Four image-task rows in the workshop; circle/ellipse map to F, square/triangle to J. Two rows in the live demonstration (circle and triangle). Preload failure must not silently continue. All-timeout summaries have 0% accuracy and N/A correct RT.

Use the live exercises and workshop sheet as the practical sequence. Compare with the released reference solution after an attempt. The debugging guide provides a symptom-to-check table. Try one editable example and its practice check before the full task.

Questions to answer without copying the code

1. Does preload guarantee timing accuracy?
2. Does mobile=false prove a keyboard?
3. What makes a useful pilot log?

Check your reasoning by running a small example; the instructor guide contains the expected explanations.

Reading (about 15 minutes)

Read media preloading, browser-check, and timing accuracy. Before Meeting 5, read the course deployment guide.

Preparation for the next checkpoint

Show one deliberately tested failure and the evidence that it was handled. See assessment and deadlines.