This paper reports on the conversion of the TIGER treebank, a syntactically interpreted corpus of German newspaper texts, into a testsuite for a broad-coverage Lexical-Functional Grammar (LFG) for German. It presents the two major steps of the conversion, which consists of an XSL transformation of the TIGER XML representation into a relational Prolog-like representation and the subsequent application of term-rewriting rules as they are used in certain MT transfer components to that representation. Then some problems due to considerable differences in analysis or to information not encoded in the TIGER representation are discussed. The output consists of (partly ambiguous) f-structure charts, which can then be mapped against the grammar's output for evaluation purposes.