[ACT-R-users] Ref for empirical tests of cognitive models by predicting task difficulty

[Ken Koedinger]

I'm looking for references for the following statement and figure the 
ACT-R community may have some:

"One way to empirically evaluate the quality of a cognitive model is to 
test whether it can be used to accurately predict differences in task 
difficulty."

I'm particularly interested in references to models that predict error 
rates (but reaction time prediction is ok too) across a number of 
related tasks.  Models that predict errors at steps in tasks and/or 
specific strategy or error differences are even better.  One such 
reference is our own tech report below -- see constraint C3 in Table 1.

Koedinger, K.R., & MacLaren, B. A. (2002).Developing a pedagogical 
domain theory of early algebra problem solving.CMU-HCII Tech Report 
02-100.[PDF 
<http://pact.cs.cmu.edu/koedinger/pubs/Koedinger,%20McLaren%20.pdf>]

Others?  Including your own work?

Thanks!
Ken