We investigated the potential of Second Life® (SL), a three-dimensional (3-D) virtual world, to enhance undergraduate students' learning of a vital chemistry concept. A quasi-experimental pre-posttest control group design was used to conduct the study. A total of 387 participants completed three assignment activities either in SL or using two-dimensional (2-D) images. Students were administered an 11-question chemistry achievement test and two measures of spatial ability (Purdue Visualization of Rotations Test, Card Rotations Test). Although analyses of covariance revealed no statistically significant differences between the two groups as a whole for any of the outcome measures, a subgroup analyses was conducted to decompose the relative impact of 3-D virtual reality instruction within SL. We found that students classified as having poor spatial ability showed significantly greater improvement in understanding the 3-D nature of molecules if they did relevant activities in a 3-D virtual world than those students who only worked with 2-D images.