The geoscience education research conducted for this study was designed to recognize, measure, and better characterize the three-dimensional spatial abilities of introductory geology lab students. The specific visualization skill under investigation is the ability to conceptualize the detailed three-dimensional underground structure of folded sedimentary rocks from only surface clues - a basic skill any geology undergraduate student must master, known within the geoscience community as Visual Penetrative Ability (VPA). While working on classroom laboratory exercises, students in two of our sample of four laboratory sections were given Geo 3D, a computer-based educational program, while the other two sections were provided with the standard, non-experimental paper-and-pencil lab-based learning exercises traditionally used in the laboratory course. The purpose was to see if spatial abilities as measured by the Geo SAT instrument after the exercises differed from pre-test results and if so, how they differed. Post-instruction problem-solving interviews were also conducted with students representing the range of performance on the Geo SAT. Within our sample of four laboratory sections, the post-test performance of both the experimental and traditional groups was essentially the same, indicating no obvious advantage to our relatively short and prescribed exercise with Geo 3D. Pre-test results also showed that beginning geology students, mostly non-majors, routinely produced over 50% correct responses on the Geo SAT instrument. This suggests a more difficult or refined version of this instrument is required for measuring more advanced spatial abilities in college level geology students. Qualitative analysis of problem-solving interviews yields insight into the nature of the difficulties faced by students in solving this style of spatial problem and explained the origin of many of the common incorrect responses seen by previous workers using the Geo SAT instrument. Students with high VP A appear to rapidly construct a complete, three-dimensional internal visual model, and after some work can draw cross-sectional and face-completion diagrams of cubic slices of geologic structures based only on surface information. Students with poor VP A tend to view external information as merely a type of "gift wrapping" over the cube surface and do not perceive the internal structure presented. Their attempts at completing cross-sections or blank faces illustrate how many common non-penetrative errors shown by previous research is generated. We construct a process model for visual problem solving of this type, which shows the origin of commonly observed errors as a function of a spectrum of varying VPA ability at specific crucial steps, and associated accommodation strategies in problem solving related to this type of spatial visualization.