Table 1. Strategy for Biasing Responses Toward ‘old' or ‘new' Judgments.

Bias level	5	4	3	2	1
Cup height	Short	Short	Regular	Regular	Tall
Old reward	1/2	1/4	1/4	1/4	1/4
New reward	1/2	1/2	1	2	3

The numbers in the bottom two rows represent the cereal ring(s) awarded for correct ‘old' or ‘new' judgments. Each bias level has a distinct combination of cup height and food reward ratio, resulting in a distinct false alarm and hit rate (ie, one of five data points used to generate the ROC curves). Liberal criteria such as bias level 1 (digging is more difficult, thus biasing toward ‘old' judgments) result in high false alarm and high hit rates, whereas conservative criteria, such as bias level 5 (digging is easier, thus biasing towards ‘new' judgments) result in low false alarm and hit rates.