In two cross-linguistic priming experiments with native German speakers of L2 English, this study explored the role of constituent order and level of embedding in cross-linguistic structural priming. In both experiments, significant priming effects appeared only if prime and target were similar concerning constituent order and also situated on the same level of embedding. The study debates its results on the basis of two current theoretical accounts of cross-linguistic priming, and concludes that neither an account based on combinatorial nodes nor an account adopting that constituent order is directly responsible for the priming effect can fully clarify the data pattern. The study proposes an account that clarifies cross-linguistic priming via a hierarchical tree representation. This representation is computed during processing of the prime, and can affect the formulation of a target sentence only when the structural features detailed in it are grammatically correct in the target sentence.