Structural Alignment in Science Learning

• In Event: Domain-General Mechanisms of Science Learning and Reasoning

Abstract

Learning science involves thinking about the world in terms of relational structure, such as the anatomical links between species, and the planetary motion underlying events seen in the sky. Complex relational structures are typically conveyed through visual representations, like diagrams and models (Ainsworth, 2018). To make the relevant connections, students often must compare the parts, objects, or systems in these representations—a process of structural alignment in which correspondences are mapped between two representations on the basis of their shared relational structure (Gentner & Markman, 1997). Hence, support for structural alignment could promote science learning. We present two lines of research that substantiate this claim.

The first explores how the layout of an image can help or hinder comparison. Figure 1 shows the skeletal structures of human and bird forelimbs. When the limbs are horizontal, the correspondences between the human and bird limbs (represented by connecting lines) are clear and direct when one is placed above the other. However, when the two limbs are side by side, the same correspondences are impeded by nonmatching parts (thus the crossed connecting lines). In contrast, when the limbs are vertical, horizontal placement is direct, and vertical is impeded. Comparison of simple figures tends to be faster and more accurate when placement is direct (Matlen et al., 2020). We found similar patterns with more complex skeleton diagrams. College- and middle-school-aged participants had to find an anomalous bone in one of two highly-similar skeletons. The placement of the pairs varied across trials. Overall, accuracy was higher when placement was direct (M=0.81, SD=0.16) versus impeded (M=0.77, SD=.20; t=2.57, p<.05). This effect was particularly pronounced when diagrams were presented in uncommon spatial orientations. We also gauged the frequency of direct placements in educational materials by coding over 300 middle-school science textbook images. Visual comparisons were common (37% of images). Direct placement, however, occurred in fewer than half of the images, suggesting ample room for improvement.

In a second line of research, we explored students’ learning of scientific models. We view the causal integration of models and observations as a form of structural alignment in which temporal, spatial, and causal relations are mapped across a set of modeled and observable events. We will report research in which 3rd-graders were taught the scientific model of the day/night cycle (Jee & Anggoro, 2019). Some participants received lessons in which observable phenomena (the Sun’s location in the sky) were explicitly compared to the model (Earth’s rotation) (see Figure 2). We scored students’ pre- and posttest explanations along two dimensions: the frame of reference (Earth- vs. space-based), and the object in the causal role (Sun vs. Earth). Whereas participants in a “no instruction” (control) condition gave similar explanations at pre- and posttest, those who engaged in comparison dramatically shifted their explanations toward a space-based frame of reference with Earth motion as the central cause—consistent with the scientific model, X2 = 9.04, p < .05. Together these projects demonstrate the versatility and effectiveness of supports for structural alignment in science education.

Authors

• Benjamin D. Jee, Worcester State University
  Presenting Author

• Florencia K. Anggoro, College of the Holy Cross
  Non-Presenting Author

• Bryan Matlen, WestEd
  Non-Presenting Author

• Nina Simms, Northwestern University
  Non-Presenting Author

• Dedre Gentner, Northwestern University
  Non-Presenting Author