This paper studies literature recommendation approaches using both content features and coauthorship relations of articles in literature databases. Most literature databases allow data access (via site subscription) without having to identify users, and thus task-focused recommendation is more appropriate in this context. Previous work mostly utilizes content and usage log for making task-focused recommendation. More recent works start to incorporate coauthorship network for recommendation and found it beneficial when the specified articles preferred by authors are similar in their content. However, it was also found that recommendation based on content features achieves better performance under other circumstances. Therefore, in this work we propose to incorporate both content and coauthorship network in making task-focused recommendation. Three hybrid methods, namely switching, proportional, and fusion are developed and compared. Our experimental results show that in general the proposed hybrid approach achieves better performance than approaches that utilize only one source of knowledge. In particular, the fusion method tends to have higher recommendation accuracy for articles of higher ranks. Besides, the content-based approach is more likely to recommend articles of low fidelity, whereas the coauthorship network-based approach has the least chance.