The objectives: (1) measuring spatial ability and geometric thinking level of Prospective Elementary School Teachers (PGSD students), (2) analyzing difficulties faced on a certain spatial ability. The study used a mixed-method quantitative and qualitative. Researchers developed test items to measure the spatial ability and geometric thinking of 139 Prospective Elementary School Teachers (PGSD students), Universitas Negeri Semarang. The sampling method employed was cluster random sampling, while three students selected for the interview were based on the 'poor' and 'very poor' category. Data analysis with the Rasch Model showed that the item reliability, item validity, item separation, and unidimensionality were classified as good. The results revealed that spatial ability and geometric thinking of Prospective Elementary School Teachers were classified as fair (62.5%) and good (31.7%). The difficulties in completing tasks of spatial ability primarily on the perception and relation, which happened due to their lacking of imagination.