In a series of six experiments, prediction of exam performance from information about motivation and ability as well as about motivation alone or ability alone of students was studied. The factorial plot of the Motivation x Ability effect always yielded the parallelism pattern with subjects from both student and nonstudent populations. Manipulation of difficulty of exam did not alter this parallelism pattern. Results agreed with the hypothesis of cultural difference between the adding and constant-weight averaging rules disclosed a developmental trend: High school and undergraduate college students followed the averaging rule; post-graduate students followed the adding rule. Establishment of these rules allowed analyses of imputations about missing information. The conventional distinguishing tests which rely on just one of the two heterogeneous types of information were found to be more useful in analyses of imputation rules than in diagnosis of cognitive algebra. Manipulation of information reliability disclosed presence of two initial opinions, one about motivation and another about ability, contrary to the finding of one initial opinion in American students.