Mari Strand Cary, Carnegie Mellon University
David Klahr, Carnegie Mellon University
Abstract: The authors tested the effectiveness of two instructional methods on 3rd-5th grade children's acquisition of the Control of Variables Strategy (CVS; the idea of designing unconfounded experiments). It went beyond previous studies by including both immediate and delayed assessments; direct, near, and far-transfer assessments; pre-/post-tests based on standardized measures of concern to schools; and multiple definitions of mastery.
The major questions of interest were whether explicit instruction is a better, more efficient means of teaching CVS and whether the "path independence learning hypothesis" holds true. Though immediate assessments generally reflect patterns arising from the original study, data arising from different definitions of mastery and longer delays paint a different picture. Immediately after explicit training or exploratory learning, explicit instruction resulted in better performance. However, over the long-term, groups performed similarly except when perfect performance was the criterion, in which case there were more "perfect performers" in the explicit instruction group. Data also suggested that predictions of performance on near and far transfer tasks may not be meaningful if they are based on initial performance on the learning task. Instead, how well students retain or even consolidate skills appears to be a better predictor of their transfer performance on applied tasks and standardized test items. The path independence finding was replicated, but looked much more nuanced and interesting when the different delays and tasks were considered. The bottom line is that the assessment tool, the desired level of CVS expertise, and the time period over which such expertise is measured must be defined before one can decide which teaching method is "better."