nism for providing firsthand experience and are therefore assumed to improve students’ understanding of the nature of science.

Research on student understanding of the nature of science provides little evidence of improvement with science instruction (Lederman, 1992; Driver et al., 1996). Although much of this research historically did not examine details of students’ laboratory experiences, it often included very large samples of science students and thus arguably captured typical laboratory experiences (research from the late 1950s through the 1980s is reviewed by Lederman, 1992). There appear to be developmental trends in students’ understanding of the relations between experimentation and theory-building. Younger students tend to believe that experiments yield direct answers to questions; during middle and high school, students shift to a vague notion of experiments being tests of ideas. Only a small number of students appear to leave high school with a notion of science as model-building and experimentation, in an ongoing process of testing and revision (Driver et al., 1996; Carey and Smith, 1993; Smith et al., 2000). The conclusion that most experts draw from these results is that the isolated nature and rote procedural focus of typical laboratory experiences inhibits students from developing robust conceptions of the nature of science. Consequently, some have argued that the nature of science must be an explicit target of instruction (Khishfe and Abd-El-Khalick, 2002; Lederman, Abd-El-Khalick, Bell, and Schwartz, 2002).

As discussed above, there is reasonable evidence that integrated instructional units help students to learn processes of scientific inquiry. However, such instructional units do not appear, on their own, to help students develop robust conceptions of the nature of science. One large-scale study of a widely available inquiry-oriented curriculum, in which integrated instructional units were an explicit feature, showed no significant change in students’ ideas about the nature of science after a year’s instruction (Meichtry, 1993). Students engaged in the BGuILE science instructional unit showed no gains in understanding the nature of science from their participation, and they seemed not even to see their experience in the unit as necessarily related to professional science (Sandoval and Morrison, 2003). These findings and others have led to the suggestion that the nature of science must be an explicit target of instruction (Lederman et al., 2002).

There is evidence from the ThinkerTools science instructional unit that by engaging in reflective self-assessment on their own scientific investiga-