Purpose: To test the extent of variation among nuclear medicine physicians with respect to staging non-small cell lung cancer with positron emission tomography (PET). Procedures: Two groups of nuclear medicine physicians with different levels of PET experience reviewed 30 PET scans. They were requested to identify and localize suspicious mediastinal lymph nodes (MLN) using standardized algorithms. Results were compared between the two groups, between individuals, and with expert reading. Results: Overall we found good interobserver agreement (kappa 0.65). Experience with PET translated into a better ability to localize MLN stations (68% vs. 51%, respectively), and experienced readers appeared to be more familiar with translating PET readings into clinically useful statements. Conclusions: Although our results suggest that clinical experience with PET increases observers' ability to read and interpret results from PET adequately, there is room for improvement. Experience with PET does not necessarily improve the accuracy of image interpretation.