Background: Microscopic human cancers can remain dormant for life. Tumor progression depends on sequential events, including a switch to the angiogenic phenotype, i.e., initial recruitment of new vessels. We previously demonstrated that human tumors contain tumor cell populations that are heterogeneous in angiogenic activity. Here, we separated angiogenic from nonangiogenic human tumor cell populations and compared their growth. Methods: Severe combined immunodeficient (SCID) mice were inoculated with nonangiogenic human MDA-MB-436 breast adenocarcinoma, KHOS-24OS osteosarcoma, or T98G glioblastoma cells. Most of the resulting tumors remained microscopic (<1 mm diameter), but some eventually became angiogenic and enlarged and were used to isolate angiogenic tumor cells. Angiogenic and nonangiogenic tumor cells were inoculated into SCID mice, and time to the development of palpable tumors was determined. Cell proliferation was assayed in vitro by growth curves and in vivo by staining for proliferating cell nuclear antigen or Ki67. Microscopic tumors from both tumor cell populations were examined for histologic evidence of vascular development 14 days after inoculation in mice. Expression of the angiogenesis inhibitor thrombospondin-1 was examined by immunoblotting. Results: Nonangiogenic tumors of each tumor type developed palpable tumors after means of 119 days (range: 53–185 days) for breast cancer, 238 days (184–291 days) for osteosarcoma, and 226 days (150–301 days) for glioblastoma. Angiogenic cells developed palpable tumors within 20 days after inoculation. However, nonangiogenic and angiogenic cells of each tumor type had similar proliferation rates. Fourteen days after tumor cell inoculation, tumors from angiogenic cells showed evidence of functional vasculature. In contrast, nonangiogenic tumors remained microscopic in size with absent or nonfunctional vasculature. Thrombospondin-1 expression was statistically significantly lower (by five- to 23-fold, depending on tumor type) in angiogenic than nonangiogenic cells. Conclusions: This model provides a conceptual framework and a reproducible in vivo system to study unresolved central questions in cancer biology regarding the initiation, reversibility, and molecular regulation of the timing of the angiogenic switch.