Performing the test in fixed order creates a risk of bias due to either training or exhaustion. With a sufficiently big data set, it would be interesting to sho

Performing the test in fixed order creates a risk of bias due to either training or exhaustion. With a sufficiently big data set, it would be interesting to show the test to different groups in different orders, or at least repeat one of the "easy" tests at the end. I noticed that my performance significantly dropped towards the end because my brain was just fried. (Also, who else did the test while task-switching to HN while waiting for something to finish?)