Improving achievement in Science, Technology, Engineering, and Mathematics (STEM) is essential for elevating individual success and solving societal challenges. Yet national and international assessments paint a poor picture of students' STEM competencies (NAEP, 2019; OECD, 2018). An untapped avenue to improve STEM education lies in developing students' spatial skills. Spatial thinking is a collection of mental skills that allow one to reason about real and imagined spaces, the locations of objects, their shapes, their relations to each other, and the paths they take as they move (Newcombe, 2010). These skills are not only essential for daily life, research has shown that having good spatial skills leads to success in STEM courses and careers, and improving skills leads to improved STEM performance (Wai et al, 2009). These data have led to efforts to develop training programs to build spatial skills (Burte et al, 2017; Cheng, & Mix, 2014; Lowrie et al, 2017). These efforts have focused on dedicated training to develop skills, yet this approach is often infeasible for public schools in the United States facing budget cuts and pressure to perform high-stakes assessments. Recognizing these challenges, we partnered with a large, urban school district to evaluate a never-before-tested approach to developing spatial thinking skills in elementary-aged students and their teachers. The district serves predominantly students who are historically underrepresented in STEM in the United States. Together with the district's curriculum development team, we created a "spatially enhanced (SE)" science curriculum for 3rd grade. This SE curriculum incorporated five spatial tools (spatial language, gesture, spatial comparison, sketching, and visualizations) into teachers' pedagogy and materials, and student activities throughout the academic year. These tools were research-informed and have been shown to improve spatial thinking skills in previous research (see Gagnier & Fisher, 2020). Based on this work, we hypothesized that incorporating these five enhancements into teachers' and students' classroom activities would improve students' academic achievement in science. Our research question was whether the SE curriculum led to improved academic achievement in science and math. To examine this question, we compared student grades and performance on standardized tests for students taught with the SE curriculum compared to those taught with two control curricula. Using a cluster-randomized controlled trial and a stratified sampling technique, we divided schools into separate strata based on student demographic characteristics (socioeconomic status, ethnicity). Within each stratum, schools were randomly assigned to one of three conditions: (1) In the Business-As-Usual Condition (BAU), teachers taught the standard district-provided curriculum; (2) In the NGSS Condition, the teachers were provided with a curriculum that was aligned with the Next Generation Science Standards (NGSS) and included detailed lesson plans, prompts for teachers to guide discussion, and guiding slide decks to facilitate the lesson implementation; and (3) In the Spatially-Enhanced-NGSS condition (SE-NGSS), our research team modified the NGSS-aligned lesson plans, teacher instructional approaches and prompts, and teacher materials to teach students science using five spatial enhancement tools (spatial language, gesture, spatial comparison, sketching, and visualizations; Gagnier & Fisher 2020). In this way, we spatially enhanced science instruction by infusing spatial thinking practices into daily science instruction. Our student sample was 50% male and 50% female with a mean age of 9.3 years (SD=0.36). Our sample was diverse; over 87% of the students were from backgrounds historically underrepresented in STEM (see Table 1). Additionally, 8.8% of students were identified by the district as having an individualized education plan (IEP), 15.1% as being an English Language Learner (ELL), and 50.7% qualified for free or reduced lunch. Our analytic approach proceeded in 3 steps. In Step 1, we examined baseline equivalence using 2nd-grade science grades. In Step 2, we established the baseline module including which covariates to include. Step 3 occurred in three phases. First, we partitioned the variance in each outcome based on within and between teacher components. Then we added student-level covariates (free-or-reduced lunch qualification, gender, IEP status, 2nd-grade science) and teacher-level covariates (years of experience, gender). Finally, in Phase 3 we added student-level moderators. We measured academic achievement using teacher-assigned grades (N=1,411 students) in English language arts (ELA), science, and mathematics and state standardized assessments in ELA and mathematics (N=1,322 students) using the Maryland Comprehensive Assessment Program (MCAP). We met baseline equivalence for grades. Figure 1 shows mean group differences in year-end grades by condition. As can be seen in the Figure, while there were no differences in ELA, math and science grades were significantly higher in the SE-NGSS condition. These effects, however, were not robust enough to hold when student and teacher covariates were entered into the model. Yet, we have promising moderation effects. The SE-NGSS appears to have a differential effect on math and science grades for students who ended 2nd grade lower in their science achievement (p<0.001; see Table 2). Additionally, there is a significant difference between students who qualify for free or reduced lunch in both the BAU (p=0.035) and NGSS (p<0.001) conditions but not in the SE-NGSS condition (p=0.564; see Table 3). For state standardized assessments, there were no effects for ELA MCAP. For MCAP math scores, we found that the SE curriculum mitigated factors that cause a test difference in math scores between girls and boys. Whereby boys assigned to the BAU condition did significantly better on their standardized math (p=0.017) tests than girls in the BAU condition and we see no such difference among kids assigned to the NGSS-SE condition (see Table 4) and this pattern holds after controlling for 2nd-grade science grades (p=0.035). These data suggest that spatially enhancing the science curriculum is a viable method for supporting students' achievement in science and math. Findings will be discussed in the context of the generalizability and scalability of spatially enhancing curricula for improving academic achievement in STEM and as a potential lever for reducing achievement gaps.