Science centers such as museums and planetariums have used stereoscopic ("three-dimensional") films to draw interest from and educate their visitors for decades. Despite the fact that most adults who are finished with their formal education get their science knowledge from such free-choice learning settings very little is known about the effect of stereoscopic film presentation on their science learning. We explored this issue by designing a quasi-experimental field trial with a short film about the shape of the Milky Way galaxy. The film was produced based on a set of stereoscopic design principles derived from spatial cognition and cognitive load literature with the goal of lowering the audience's extraneous cognitive load. The film was randomly shown in either two-dimensional (2D) or stereoscopic format to 498 adults who visited a large, urban planetarium. To investigate the extent of audience's change related to galaxy-related spatial concepts, an identical set of questions was asked on iPads before and after the film was shown. A delayed posttest was given to 123 of those adults approximately 6 months later. Test performances were analyzed using repeated measures analysis of covariances (ANCOVAs) with demographic and spatial visualization ability measures as covariates. Results show identical short-term learning gains in both the 2D and stereoscopic groups. However, only the stereoscopic group exhibited long-term learning gains. Findings were interpreted through the lenses of cognitive load theory and the limited capacity model of mediated message processing.