The separation of benign from malignant mesothelial proliferations has emerged as a major problem in the pathology of the serosal membranes. For both epithelial and spindle cell mesothelial processes, true stromal invasion is the most accurate indicator of malignancy, but stromal invasion is often difficult to assess, especially in small biopsies. In the pleural cavity, deep penetration of a thickened and fibrotic pleura or penetration of mesothelial cells into the fat of the chest wall are good indicators of malignancy; however, superficial entrapment of mesothelial cells and glands by organizing effusions is common in benign reactions and needs to be distinguished from invasion. In the peritoneal cavity, invasion of fat or of organ walls is again the most reliable indicator of malignancy, but entrapment of benign cells in organizing granulation tissue or between fat lobules is frequent and confusing. Proliferations confined to the pleural or peritoneal space, particularly linear arrays of atypical mesothelial cells on the free surface, should not be called malignant in the absence of unequivocal invasion. Cytologic atypia is often not helpful in separating benign from malignant reactions, because benign processes are commonly atypical and mesotheliomas are often deceptively monotonous. Densely packed mesothelial cells within the pleural space are frequent in benign reactions, but densely packed mesothelial cells within the stroma favor a diagnosis of malignancy. Organizing effusions (fibrous pleurisy) typically show zonation with high cellularity and cytologic atypia toward the pleural space and increasing fibrosis with decreasing cellularity and lesser atypia toward the chest wall, whereas sarcomatous (including desmoplastic) mesotheliomas do not demonstrate this type of zonation. Elongated capillaries perpendicular to the pleural surface are seen in organizing effusions but are not a feature of sarcomatous mesotheliomas. The combination of a paucicellular storiform pattern, plus invasion of the stroma (including fat and adjacent tissues), or bland necrosis, overtly sarcomatous foci, or distant metastases, is required for the diagnosis of desmoplastic mesothelioma. Necrosis is usually a sign of malignancy but is occasionally seen in benign mesothelial reactions. Keratin staining is useful in indicating the distribution of mesothelial cells, and particularly in demonstrating penetration of mesothelial cells into the stroma or adjacent structures, but is of no help in separating benign and malignant proliferations because both are keratin-positive. Although both p53 and EMA staining have been proposed as markers of mesothelial malignancy, in our experience they are not helpful for the individual case.