One group got five turns, where the number of boxes increased each turn. The subjects started choosing between two boxes, then the number went up to four, eight, 16, and 32. For the other group, the number of boxes decreased from 32 to two over the same five turns. With more boxes, the participants sampled the boxes more, but didn't increase the sampling proportionately to the number of boxes. They might do 12 samples for two boxes, but (only) 50 samples for 32 boxes.