Does one change to the task change your speed?

Make your page run two conditions

1. In VS Code, make a folder week05-compare/ in your private repo. Copy your week02-rt/rt.html into it. Copy the Week 5 README into it too.
2. Before you change any code, fill in the Prediction section of the README: which condition will be slower, by roughly how many ms, and why. Commit it. The commit time is your proof it came first.
3. Pick one thing to change between A and B that you think will change speed: square size (72 vs 20 pixels), waiting time (always 1000 ms vs random 800–2200 ms), or brightness (bright vs dim). Not a colour you just like.
4. Edit the code. You need all five changes:

   // (a) near the top: which condition is running
   let condition = "A";
   
   // (b) in keyPressed(): A and B keys switch condition
   if (key === "a" || key === "b") {
     condition = key.toUpperCase();
     note = "Condition " + condition + ". Click Start.";
   }
   
   // (c) in draw() or beginTrial(): make B different. Example for size:
   const size = condition === "A" ? markSize : 20;
   rect(width / 2, height / 2 + 10, size, size, 10);
   
   // (d) add condition to BOTH results.push({...}) calls,
   //     to the CSV header, and to the .join(",") list
   condition: condition,
   
   // (e) in startBlock(): delete  results = [];
   //     so block B does not erase block A

5. Run it: Practice 6, press a, Start 20, press b, Start 20, Export. You should get one CSV with a condition column and at least 40 scored rows. Rename it rt-week05.csv and put it in week05-compare/.
6. Load it above. Value column rt, condition column condition. Copy the table into your README and write one sentence: did the data agree with your prediction?

The tool leaves out practice trials, anticipations, rows marked usable = 0, wrong answers (correct = 0), and times outside 150–2500 ms.

Make condition B a choice

In A, one square, one key (simple reaction time). In B, the square appears on the left or the right at random. Press f for left and j for right. This is the comparison Donders made in 1868; you will read about it next week.

1. In beginTrial(), pick a side: side = random() < 0.5 ? "left" : "right"; (declare let side; at the top).
2. In draw(), in B draw the square at width * 0.25 or width * 0.75 instead of the middle.
3. In keyPressed(), in B send f and j to handleResponse(key); in A keep the space bar.
4. In handleResponse(k), work out correct (always 1 in A). Add side and correct to the results and the CSV.
5. Run new blocks. Report how many errors you made in B. Errors are left out of the mean.

Do the comparison in Python

Copy week04-report/scripts/summarize.py to week05-compare/compare.py and change it so that it:

1. keeps only scored trials (same rules as this page);
2. splits the rows by the condition column and prints n, mean, and SD for A and for B, then B − A;
3. prints the overlap number: out of every possible pair of one A trial and one B trial, the percentage where B is slower;
4. saves compare.png with both conditions on one figure.

Check: your Python numbers must match this page. If they do not, find out why and say so in the README.