AIM: To evaluate scientific production among research fellows employed at the Zagreb University School of Medicine and identify factors associated with their scientific output. METHOD: We conducted a survey among research fellows and their mentors during June 2005. The main outcome measure was publication success, defined for each fellow as publishing at least 0.5 articles per employment year in journals indexed in the Current Contents bibliographic database. Bivariate methods and binary logistic regression were used in data analysis. RESULTS: A total of 117 fellows (response rate 95%) and 83 mentors (100%) were surveyed. The highest scientific production was recorded among research fellows employed in public health departments (median 3.0 articles, interquartile range 4.0), compared with those from pre-clinical (median 0.0, interquartile range 2.0) and clinical departments (median 1.0, interquartile range 2.0) (Kruskal-Wallis, P =0.003). A total of 36 (29%) research fellows published at least 0.5 articles per employment year and were considered successful. Three variables were associated with fellows' publication success: mentor's scientific production (odds ratio [OR], 3.14; 95% confidence interval [CI], 1.31-7.53), positive mentor's assessment (OR, 3.15; 95% CI, 1.10-9.05), and fellows' undergraduate publication in journals indexed in the Current Contents bibliographic database (OR, 4.05; 95% CI, 1.07-15.34). CONCLUSION: Undergraduate publication could be used as one of the main criteria in selecting research fellows. One of the crucial factors in a fellow's scientific production and career advancement is mentor's input, which is why research fellows would benefit most from working with scientifically productive mentors.