A wide range of results have been obtained over the years concerning levels of recall, levels of recognition, and levels of recognition given recall. Attention has focused on a regularity called the Tulving-Wiseman Law (e.g., Flexser & Tulving, 1970; Nilsson & Gardiner, 1993; Tulving & Wiseman, 1975). This concerns the relationship between probability of recognition p(Rn) and probability of recognition given recall p(Rn|Rc):
p(Rn|Rc) = p(Rn) + .5[p(Rn) - p(Rn)2]
This indicates a relatively weak relationship between the two. It turns out that ACT-R very closely predicts the Tulving-Wiseman function (for details see Anderson, Bothell, Lebiere, & Matessa, in press). The two measures-p(Rn) and P(Rn|Rc)-are somewhat related because of the shared permanent noise between recall and test, but they are substantially unrelated because of the uncorrelated temporary noise and variability in the Sji between the cues j and the chunks i.