The serial levels of fibronectin (FNp) in plasma from 65 patients with benign and malignant breast disease and from 74 healthy control women were assayed by the use of the rocket immunoelectrophoresis procedure. Mean FNp levels in patients with breast cancer were age-matched with control subjects, but no clear correlation was found between FNp levels and the presence of primary tumor. Mean FNp values for fibroadenoma patients did not differ either from controls or from patients with malignant disease. Patients were also categorized according to TNM classification, to the number of positive axillary nodes, to the histologic grade of malignancy, and to the presence of estrogen receptors. Although differences were not significant, a higher number of patients with levels greater than the normal 95% percentile were found only in the group with four or more positive axillary nodes and in the group with a greater number of histologic malignancies. Marked fluctuations in FNp concentrations were found in individual patients during the follow-up period, independently of the treatment received. FNp seems unsuitable as a tumor marker because, besides its apparent lack of specificity for cancer, it reflected neither the host-tumor burden nor the response to treatment.