This study aims to explore the personalized recommendation needs (PRNs) of researchers from user's perspective.  Semi-structured in-depth interviews with 22 researchers were conducted, and NVivo11 was used for data analysis. Five PRNs were identified: content needs, interactive functional needs, interface layout needs, effectiveness needs and emotional needs. Furthermore, the PRNs hierarchical model indicates that content needs are basic needs, should be satisfied firstly, interactive functional needs, interface layout needs should be satisfied secondly, followed by effectiveness needs and emotional needs. In addition, tasks, interactive retrieval habits, and recommendation interpretation affect the PRNs of researchers. Based on the results, a PRNs hierarchical model is developed. This study has implications for incorporating personalized recommendation needs into algorithms. It adds new knowledge about personalized recommendation to the research community and informs personalized recommendation system design.