jsPsych / Behavioral experiments · 03
Programming behavioral experiments · Meeting 3

Control flow that matches the design

Mappings, practice feedback, bounded loops, and conditional test

Bartosz Maćkiewicz

Concept · 03

Today's outcome

Build a task that takes the right route after success, failure, or a change of response mapping.

Artifact: a flanker pilot with bounded practice and a test gate.

Concept · 03

Start from the participant's journey

Did this attempt meet the criterion?
Yes → test, without practice feedback
No, attempts remain → repeat practice
No, maximum reached → stop and explain

Instructions → practice → check the current attempt.

A bounded loop and a test gate solve different problems.

Concept · 03

What is the correct answer?

←←←←←    left
→→→→→    right
←←→←←    right
→→←→→    left

Respond to the central arrow. Ignore the neighbors. Check every answer key independently.

Experiment laboratory · 03

A mapping is not a direction

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

Group A: left = F; right = J. Group B: left = J; right = F.

Store the target direction and the expected key separately. Instructions, scoring, and saved group must agree.

Concept · 03

Choose a response mapping

const requested = new URLSearchParams(location.search)
  .get("group");
const group = requested === "B" ? "B" : "A";

Use ?group=B for the reversed mapping; otherwise use A. The example reads one parameter, then uses one mapping.

Concept · 03

Counterbalancing needs an allocation rule

A URL parameter carries an assignment; it does not create balanced groups by itself.

For a classroom demonstration, allocate A/B deliberately. A real study needs a documented allocation procedure and checks of actual group counts.

Concept · 03

Label phases explicitly

const practiceItems = PRACTICE_ITEMS.map((item) =>
  ({ ...item, block: "practice" }));

Record task, block, group, condition, item, and attempt. Avoid inferring a phase later from row position alone.

Concept · 03

A practice procedure

const practiceBlock = {
  timeline: [fixation, trial, feedback],
  timeline_variables: practiceItems,
  randomize_order: true
};

The test procedure uses [fixation, trial].

Concept · 03

Feedback reads the last task response

const last = jsPsych.data.get()
  .filter({ task: "flanker" })
  .last(1).values()[0];

Distinguish correct, incorrect, and no response. Do not accidentally read a fixation or feedback row.

Try it · 03

Live checkpoint L1: feedback only in practice

Add feedback to the practice procedure. Test a correct answer, a wrong answer, and a timeout.

Check that feedback has its own task label and does not enter the accuracy denominator.

Concept · 03

Measure the current attempt

const recent = jsPsych.data.get().filter({
  task: "flanker", block: "practice",
  practice_attempt: practiceAttempts
});

Pooling all previous attempts can keep a successful learner below criterion unnecessarily.

Concept · 03

A bounded loop

const loopedPractice = {
  timeline: [practiceBlock, practiceSummary],
  loop_function: () => {
    const repeat = !practicePassed && practiceAttempts < 3;
    if (repeat) practiceAttempts++;
    return repeat;
  }
};

Both the criterion and the maximum matter.

Experiment laboratory · 03

The threshold is a boundary case

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

Four trials, criterion 75%.

  • 2 correct: fail.
  • 3 correct: pass.
  • 4 correct: pass.

Use >= 0.75, not > 0.75. Test the boundary explicitly.

Try it · 03

Live checkpoints L2 and L3

L2: run both ?group=A and ?group=B.

L3: implement bounded repetition and inspect the attempt ID.

Use the arrow on screen to decide the correct key; do not let the saved answer key be your test oracle.

Concept · 03

Conditional entry to test

const testIfPassed = {
  timeline: [beginTest, testBlock, summary],
  conditional_function: () => practicePassed
};

Stopping the loop after three failures is not permission to enter test.

Reveal the reasoning

Exiting a loop after the final allowed attempt does not imply a pass. The conditional gate must still check practicePassed.

Concept · 03

The failed-practice route

Show an explanatory stop screen and allow data export.

Tell the participant what happened and how to get help. No test rows should be created on this route.

The task must not loop forever or silently lower the criterion.

Concept · 03

First platform step: reuse a working task

Keep the Meeting 2 experiment unchanged while learning JATOS.

  • Start the prepared local JATOS installation.
  • Import the supplied study archive.
  • Run the study and retrieve its results.

Follow the local JATOS worksheet; setup trouble has a demo fallback.

Concept · 03

Move a study, then run it again

Local filesEdit and test the task.
Study .jzipTransfer assets + configuration.
New instanceRun again; retrieve a new result.

Export a .jzip; import it into a different instance; run and retrieve a new pilot.

Study archives do not transfer previous results.

Concept · 03

Three files, three purposes

File Purpose
Source ZIP Edit HTML, JavaScript, styles, and assets
Study .jzip Import a configured study into JATOS
Results export Inspect the recorded runs

Cognition.run is an alternative with its own editor and saving.

Build and verify · 03

Workshop: assemble the complete flow

  1. Check the mapping and item keys.
  2. Add practice-only feedback and bounded repetition.
  3. Gate test on passing practice.
  4. Verify normal and failure routes.

The starter contains the structure; complete the marked parts.

Read the evidence · 03

Expected task-row counts

Route Practice Test
Pass first attempt 4 8
Fail, then pass 8 8
Fail all three 12 0

Fixation, instructions, and feedback are additional rows.

Concept · 03

A design decision with consequences

Excluding people who fail practice changes who reaches test.

Document the criterion, maximum attempts, and number stopped. Do not use a practice rule as an invisible post-hoc exclusion.

Try it · 03

Project checkpoint

Complete the paper-to-prototype worksheet.

Name one manipulation, one response, one data row, and one failure route. Choose the smallest coherent version.

Proposal and Assignment 3 are due before Meeting 4.

Think first · 03

Exit ticket

Explain why both a loop bound and a test condition are needed.

Show a row from attempt 2 and explain its mapping. How would you prove that failed practice never reaches test?

← → Navigate · Esc Overview · S Presenter notes · Click outside the lab to resume navigation