I haven't read the paper, but James Hoffman and others regularly do blind tasting with 3 cups, testing things like extraction. In the case of extraction, for example, they'll measure extraction with precision scales, then have 2 cups at one level (either the "optimal" extraction or slightly below/above) and the third cup at the other level. They don't know which is the outlier, but they're to guess. They seem to be very open when they're not sure or the difference to them is marginal. They try to control variables like temperature.

I don't know if they make sure the person putting the cups down doesn't know which is which.