UC Berkeley

Humans excel at detecting patterns in information, abstracting rules, and making inferences. Underlying these skills is relational reasoning: the cognitive ability to identify and map abstract, generalizable relations between pieces of information. Though this powerful ability supports higher-order cognition, it can also be a processing bottleneck when the complexity of information is too high for our limited cognitive resources. My dissertation explores this dynamic—relational reasoning as both a cognitive tool and bottleneck for learning and reasoning—and how we overcome cognitive limitations by offloading relations to external spatial representations, such as visuospatial tools (e.g., graphs and diagrams). I focus much of my work on STEM outcomes in children because many important concepts and skills in these disciplines are relational in nature and difficult for students to learn, making them an ideal testbed for these empirical questions.

In Chapter 1, I take a high-level view of the relationship between reasoning and education and review evidence that education hones reasoning ability. I find significant evidence that the protracted, immersive experience of formal schooling taxes, and therefore improves, general reasoning skills, such as relational reasoning.

In Chapter 2, I establish that there is a unique role for relational reasoning in learning that is distinct from other cognitive skills. In two empirical studies, I use the case of fraction learning to investigate the main executive functions involved in fraction processing, and then show that relational reasoning predicts fraction performance over and above these other strong domain-general predictors.

In Chapter 3, I investigate how we learn to offload relations to physical space. In an empirical study with the Tsimane’, an indigenous farmer-forager people from the Amazon basin, I find that individuals spontaneously offload to-be-remembered relations to space, including individuals who report no formal schooling and are not literate. This study demonstrates that offloading relations to external representations is part of a foundational cognitive toolkit and is separate from the regular use of visuospatial tools.

Finally, in Chapter 4, I show that scaffolding relational reasoning during learning can improve understanding, focusing on the case of graph comprehension. I first propose a relational reasoning perspective on the difficulties that children and adults have with graph comprehension. Then, I empirically test aspects of this approach in a preliminary intervention study that manipulates the extent to which relational reasoning is engaged during a lesson on interpreting graphs of linear functions. I find that having students focus on patterns—both visual and conceptual—and make comparisons supports learning.

Taken together, this body of work simultaneously contributes to our understanding of the role of relational reasoning in higher-order cognition and to applications of relational reasoning for improving STEM education. In particular, my research provides a generative framework for identifying the STEM content that students may find especially difficult, as well as for informing the design of pedagogical approaches for helping students overcome these obstacles.