XML data is often modeled as node-labeled trees. In a structure-generic relational storage scheme, the structure of the XML data is shredded into pieces of a generic type, such as nodes or edges, and data representing these structural pieces is stored in relational tables. In this paper, we consider four structure-generic relational storage schemes. We describe the translation of XML queries into SQL queries for each scheme. We compare the performance of these storage schemes through experiments. Our study shows that each scheme performs well in some cases, and no single scheme is a clear winner in all cases.