Assignment 2: Conditions and data: a randomized interference pilot
Weight: 10 course points. Due: Before Meeting 3. Expected work: 3 hours.
Task
Extend the Meeting 2 Stroop starter to 16 trials: eight congruent and eight incongruent. Each ink color must occur twice in each condition. Give every item a unique ID, randomize the order without changing these counts, and store word, ink, condition, expected response, response, omission, and correctness. If a color task is unsuitable, use left/right central arrows with congruent/incongruent flankers and balanced target directions; document the substitution.
Show total trials, omissions, accuracy over all task trials, and mean RT for correct finite responses, overall and by condition. N/A must appear when an eligible subset is empty. Download CSV and JSON.
Required checks
Hand-calculate accuracy and mean correct RT for a four-row subset, then compare with your code. Run an all-timeout pilot to show that no NaN/Infinity or 0 ms mean is fabricated. Run twice and report the observed order; different order is plausible but not guaranteed by randomization. Counts must remain balanced regardless of order.
Submit
Follow the common submission requirements. Include the runnable experiment, a synthetic/self-pilot export, and a short note with your checks. Keep a failed run when it demonstrates an intended branch; label it clearly. Do not collect research participants for this assignment.
Rubric
| Criterion | Evidence | Points |
|---|---|---|
| Design and item balance | 16 items; eight per condition; target colors/directions balanced within condition | 2 |
| Reusable procedure | Correct v8 timeline variables, fixation, randomized order | 2 |
| Scoring and labels | Correct mappings, omission handling, independent answer-key check | 2 |
| Summaries | Explicit denominators, correct finite RT filter, empty-subset guard | 2 |
| Pilot and explanation | Exports, two observed orders, hand-calculated subset, attribution | 2 |
| Total | 10 |
Optional extension (no extra points required for full credit)
Describe a possible congruency contrast without treating this pilot as evidence for an effect. Explain why full randomization does not guarantee alternation and why equal condition counts do not remove item confounds.