This paper is concerned with whether deep syntactic information can help surface parsing, with a particular focus on empty categories. We consider data-driven dependency parsing with both linear and neural disambiguation models. We find that the information about empty categories is helpful to reduce the approximation error in a structured prediction based parsing model, but increases the search space for inference and accordingly the estimation error. To deal with structure-based overfitting, we propose to integrate disambiguation models with and without empty elements. Experiments on English and Chinese TreeBanks indicate that incorporating empty elements consistently improves surface parsing.