Ductal carcinoma in situ (DCIS), a known precursor lesion of invasive cancer of the female breast, is surrounded by a thick basement membrane and a layer of myoepithelial cells. For DCIS to become invasive, both these barriers must be breached by cancer cells. It has been repeatedly suggested that proteolytic enzymes are somehow involved in this process but a direct proof of this event has never been provided. It is our hypothesis that invasion of the DCIS by capillary vessels derived from the periductal necklace of vessels is the most likely mechanism of breaching the basement membrane, providing an escape hatch for cancer cells. This hypothesis was initially tested on ten randomly selected cases of DCIS, with or without invasion. Capillary vessels were visualized by staining histologic sections with an antibody to CD 34 and, in three cases, by combined stain for CD 34 and collagen IV. In five of the 10 cases of DCIS, the presence of discrete capillary vessels invading DCIS could be documented. In two of these five cases, the vessels subdivided the cancerous ducts into territories of unequal sizes. Vascular invasion of DCIS is a plausible mechanism of breaching the basement membrane in DCIS as a prelude to invasion. This hypothesis must be further tested on a much larger number of cases. The hypothesis, if confirmed, may suggest that invasive cancer derived from DCIS may be prevented by antiangiogenic therapy.