Reproducibility of cytologic diagnoses of malignancy and accuracy and variability of diagnoses of malignant cell types were assessed in a study of pulmonary aspirates from 100 patients. Original cytologic diagnoses were positive for 60 of the 80 patients who had malignancy (true-positive rate 75%). No false-positive diagnosis was made. Original cytologic diagnoses were negative for 15 of the remaining 20 patients (true-negative rate 75%). In "blindly" reviewing the slides, the pathologist who had made the original diagnoses (Observer A) and another pathologist (Observer B) confirmed the 60 original positive diagnoses. Observer A added four cases, Observer B 11 cases to the positive category. Cell types of original positive cytologic diagnoses and those of subsequent histologic diagnoses agreed for 23 of 33 malignancies (70%). Although two observers increased the proportion of positive cytologic diagnoses, the percentages of agreement with respect to histologic malignant cell types remained similar. For cytologic specimens, intraobserver agreement as to malignant cell types was 83% (50 of 60); interobserver agreement was 80% (48 of 60).