A panel of 17 monoclonal antibodies (MAbs) recognizing various keratin polypeptides has been used to define their binding on non-epithelial elements in 28 bone-marrow samples and 14 lymph nodes, in order to establish their limitations for use as a possible tool for immunodiagnosis of carcinoma spread. lmmunocytochemical studies have shown that only 8 antibodies consistently exhibited no falsepositive staining of marrow cells. All the remaining MAbs labelled (mostly in a non-specific manner) a few cells of marrow samples derived from patients with either haematological disorders or malignant lymphomas. Fine granules and droplet-like cytoplasmic inclusions were predominant patterns of positive reactions. Homogeneous cytoplasmic staining reminiscent of specific keratin immunolabelling was occasionally seen as well. The positive cells could be also identified in some lymph nodes free of tumour infiltration. All antibodies visualized cytoplasmic droplets in scattered cells of lymph nodes taken from a patient with non-Hodgkin lymphoma. This type of positivity was mostly associated with positive histochemical reactions for iron. Quite significant was the detection of fibrillar positivity in the extrafollicular reticular cells in all nodes examined. Such a specific type of staining was exclusively induced by antibodies directed against epitopes of keratin 8 and 18, whereas those MAbs recognizing keratin 7 and I9 always gave negative results. Our data indicate that caution is required when such MAbs, considered as markers of specific cell types, are being used as an immunodiagnostic tool t o identify single carcinoma cells. A series of criteria, including morphological ones, must be utilized in order t o obtain meaningful results.