This paper makes an explorative study of part-of-speech (POS) motif and dependency motif using the treebanks of deaf students' writing in three learning stages. Firstly, the fitting to the POS motif data with Zip-Mandelbrot distribution yields a good result. However, the dependency motif distribution fits well by the Popescu-Altmann-Köhler function instead. Then, a further observation of the top-ranking motifs finds that the most frequent motifs are formed among the most frequently-used word classes and dependencies. Lastly, the hapax-type ratio for both motif types shows an extremely heavy proportion of hapax, which may indicate the discreteness of sentence structure. Through the three analyses, some changes of sentence structure and the development of syntax are witnessed across the language learning course.