Spatial thinking is widely recognized as playing a critical role in developing geoscience expertise, yet less is known about how variation in student's spatial abilities impacts their learning of dynamic Earth systems. For a spatially complex geoscience process such as understanding stream processes, a learner could conceivably draw on spatial skills from the four typologies of Newcombe and Shipley (2014): intrinsic-static, intrinsic-dynamic, extrinsic-static, and extrinsic-dynamic. Physical models like Emriver's Em2 stream table may reduce the cognitive load by externalizing dynamic processes and enabling embodied manipulation. However, it remains unclear whether such tools benefit learners of varying spatial abilities equally. This study tests an Aptitude-Treatment Interaction (ATI) hypothesis: instructional effectiveness depends on learner characteristics, such that guidance may compensate for lower spatial ability while minimally guided exploration may better support higher spatial thinkers.

In Spring 2026, we are conducting an interaction test using a two-way factorial design in an upper-level hydrology course (n = 28), crossing spatial thinking (measured by the Spatial Thinking Aggregate Test, or STAT; Sabatini, 2024) with instructional condition (self-guided vs. instructor-guided stream table interaction). Based on initial STAT results, students were categorized as relatively high or low spatial thinkers then randomly assigned to instructional conditions. Following the intervention, students complete a spatially intensive task requiring them to modify model parameters to produce targeted erosion patterns. Performance is evaluated based on success, number of attempts, and time to completion. Students also complete structured reflections to examine differences in reasoning strategies and spatial problem-solving processes.

We hypothesize (1) main effects of instruction type and spatial thinking and (2) a significant interaction in which instructor guidance disproportionately benefits lower spatial thinkers, consistent with ATI and cognitive load theory. Findings will clarify when and for whom embodied physical models support learning of spatially complex, dynamic Earth systems in addition to being engaging.