Purpose -The present study aims to investigate how textual features, depth of citation treatment, reasons for citation, and relationships between citers and citees predict author-rated citation importance. Design/methodology/approach -A total of 49 biology and 50 psychology authors assessed the importance, reason for citation, and relationship to the cited author for each cited reference in his or her own recently published empirical article. Participants performed their evaluations on individualized web-based surveys. Findings -The paper finds that certain textual features, such as citation frequency, citation length, and citation location, as well as author-stated reasons for citation predicted ratings of importance, but the strength of the relationship often depended on citation features in the article as a whole. The relationship between objective citation features and author-rated importance also tended to be weaker for self-citations.Research limitations/implications -The study sample included authors of relatively long empirical articles with a minimum of 35 cited references. There were relatively few disciplinary differences, which suggests that citation behavior in psychology may be similar to that in natural science disciplines. Future studies should involve authors from other disciplines employing diverse referencing patterns in articles of varying lengths and types. Originality/value -Findings of the study have enabled a comprehensive, profound level of understanding of citation behaviors of biology and psychology authors. It uncovered a number of unique characteristics in authors' citation evaluations, such as article-level context effects and ruleversus affective-based judgments. The paper suggests possible implications for developing retrieval algorithms based on automatically predicted importance of cited references.