Scientific reasoning abilities of nonscience majors in physics-based courses

We have found that non-STEM (science, technology, engineering, and mathematics) majors taking either a conceptual physics or astronomy course at two regional comprehensive institutions score significantly lower preinstruction on the Lawson’s Classroom Test of Scientific Reasoning (LCTSR) in comparison to national average STEM majors. Based on LCTSR score, the majority of non-STEM students can be classified as either concrete operational or transitional reasoners in Piaget’s theory of cognitive development, whereas in the STEM population formal operational reasoners are far more prevalent. In particular, non-STEM students demonstrate significant difficulty with proportional and hypothetico-deductive reasoning. Prescores on the LCTSR are correlated with normalized learning gains on various concept inventories. The correlation is strongest for content that can be categorized as mostly theoretical, meaning a lack of directly observable exemplars, and weakest for content categorized as mostly descriptive, where directly observable exemplars are abundant. Although the implementation of research-verified, interactive engagement pedagogy can lead to gains in content knowledge, significant gains in theoretical content (such as force and energy) are more difficult with non-STEM students. We also observe no significant gains on the LCTSR without explicit instruction in scientific reasoning patterns. These results further demonstrate that differences in student populations are important when comparing normalized gains on concept inventories, and the achievement of significant gains in scientific reasoning requires a reevaluation of the traditional approach to physics for non-STEM students.

Figure 1

Distribution of formal reasoning level for non-STEM students enrolled in conceptual physics or astronomy. Formal reasoning level was determined by the LCTSR. Concrete operational (CO), transitional (T), and formal operational (FO) reasoning levels are shown.

Figure 2

Population averages on specific scientific reasoning patterns as assessed by the LCTSR. Significant difficulty is observed with proportional reasoning, isolation of variables, and hypothetico-deductive reasoning.

Figure 3

Normalized learning gains on TUG-K versus LCTSR prescore. The solid line is the line of best fit ($\mathrm{\text{slope}}=0.64$ and $r=0.59$).

Figure 4

Normalized learning gains on DIRECT versus LCTSR prescore. The solid line is the line of best fit ($\mathrm{\text{slope}}=0.45$ and $r=0.50$).