Background: Spatial ability (SA) was shown to be important for success in different fields, including STEM. Recent research suggested that SA is a unitary construct, rather than a set of related skills. However, it is not clear how individual differences in different facets of SA emerge, and how they relate to variance in general cognitive ability. Aims: The aim of the present study was threefold: (1) to examine the structure of SA testing nine theoretical models; (2) to explore the relation between 16 different facets of SA with general cognitive ability; and (3) to identify central facet(s) within the network of SA -- with most links and/or strongest links to other facets. Sample: The study participants were 958 university students from Russia. Methods: The study used a comprehensive battery of 16 SA tests and a verbal ability measure. Results: Results supported previous research, suggesting moderate overlap between all SA facets. Factor analysis suggested several potential structures, with similar fit indices for five different theoretically driven models, including split into small- and large scale; partially independent manipulation, visualization and navigation facets. Confirmatory factor analysis, mediation and network analyses showed spatial ability being largely independent from verbal ability. In addition, network analysis showed that navigation according to directions is in the centre of the network, potentially linking all SA facets. Conclusion: The results have potential implications for identifying the best targets for SA interventions. The next step in research is conducting experimental studies to evaluate effectiveness of interventions targeting navigation in comparison with other facets of SA.