Spatial thinking skills are critical to success in many subdisciplines of the geosciences. We tested students' spatial skills in geoscience courses at three institutions (a public research university, a comprehensive university, and a liberal arts college, all in the midwest) over a two-year period. We administered standard psychometric tests of spatial skills to students in introductory geology, mineralogy, sedimentology and stratigraphy, hydrogeology, structural geology, and tectonics courses. In addition, in some courses we administered a related spatial skills test with geoscience content. In both introductory and upper level undergraduate geology courses, students' skills vary enormously as measured by several spatial thinking instruments. Additionally, students' spatial skills generally improve only slightly during one academic term, in both introductory and advanced geoscience classes. More unexpectedly, while there was a tendency for high-performing students to be adept at multiple spatial skills, many individual students showed strong performance on tests of one spatial skill (e.g., rotation) but not on others (e.g., penetrative thinking). This result supports the contention that spatial problem solving requires a suite of spatial skills, and no single test is a good predictor of "spatial thinking."