Instructional psychology: aptitude, adaptation, and assessment.

f igures to decide which response f igure completes the matrix. Tullos ( 1987) studied high school students , concluding that rule inference and application components were central; errors involved inferring wrong rules or omitting rules . Results supported previous findings that high-scoring students tend to construct answers analytically before searching response alternatives to find a match, whereas lower-scoring students more often use response­ elimination strategies (Bethell-Fox et al 1984). Carpenter et al ( 1990), fur­ thermore, presented a computer s imulation theory that Raven measures "the common ability to decompose problems into manageable segments and iterate through them, the differential ability to manage the hierarchy of goals and subgoals generated by this problem decomposition, and the differential ability to form higher level abstractions" (p . 429) . On this view, discovery instruc­ tion requires Or primar ily because it demands rule and subgoal analysis and management in working memory. The model is based on college student performance and leaves out visual encoding , strategic assembly , and rule induction processes as sources of individual differences. The Tullos study and other prior work would emphasize rule induction and assembly. Further work is needed to bridge the ability and education levels contrasted in these studies, but Carpenter et al note that direct training to promote analytic strategy use and reduce response elimination and other extraneous processes could make the measure and their theory more valid. They cite evidence that such training can increase correlations of Raven with other tests; presumably, training constrains individual differences in Raven performance to those sources represented in the computer model. This assumes that sources of var iance thus eliminated are irrelevant to understanding the correlation of Gf with learning from instruction. Research on Gr as aptitude for learning also needs to compare such models with the sorts of informal reasoning and knowledge used by beginners in a new domain (Voss et al 1989 ) . Studies contrasting Gr and domain knowledge differences have begun to show some instances where low Gr is unimportant relative to prior knowledge (Schneider et al 1990) and others where Gr is more important than prior knowledge (Langstaff 1989). But most interaction ques­ tions remain unaddressed (Schneider & Weinert 1990). New demonstrations that Of can be trained are plentiful (Budoff 1987; Campione & Brown 1987, 1990 ; DeLeeuw et a1 1987 ; Feuerstein et a1 1987 ; Klauer 1990). Training research on related thinking and reasoning skills is covered by Baron & Sternberg ( 1987), Segal et al ( 1985) , Chipman et al ( 1985b), Nickerson et al ( 1985), and Resnick ( 1987) . Visual spatial abilities (Gv) Gv stands for facility in visualizat ion of f igural­ spatial s ituations and mental operations applied to them (e.g. rotation , ref lec­ t ion, analysis , and synthesis) . Also included are skill in imagining spatial