Evidence Trumps Belief in Science Education: In Both Content and Method

Our research explores the relative effectiveness of different mixes and sequences of direct instruction, discovery learning, and hands-on science methods.  The most fundamental step in this type of research is to provide clear, unambiguous operational definitions of what is meant by these commonly used, but poorly defined, approaches to early science instruction. The second step is to create well designed studies that can identify causal paths between different types of instruction and well-defined learning measures. Only with these steps in place can we be sure that decisions about science education are based on the same kinds of robust evidence that scientists demand in the very disciplines being taught.   A third step is to take demonstrably effective instructional methods and incorporate them into a computer based adaptive tutor:  in effect, transforming labor into capital in early science instruction.  I will describe one such system that is being developed to teach middle school children about how to design simple unconfounded experiments.