The frequency of metastatic involvement of 8 target organs was examined in people dying as a consequence of squamous cell carcinomas of the lower third of the esophagus or adenocarcinomas of the upper third of the rectum. In both groups, there was a stronger association of metastatic incidence in the target organs with organ blood flow per gram in patients with liver metastases than without. This association, which is consistent with the mechanical hypothesis of metastatic pattern, is apparently only seen in people with advanced disease. In the others temporal limitations imposed by death or cure prevent full development of the pattern. The data suggest that in accord with the cascade theory, cancer cells must first form metastases in the liver, that cells from these metastases, not the primary cancer, then form metastases in the lungs; cancer cells from the lung metastases then subsequently disseminate with a trend towards an arterial pattern. Data from animal experiments suggest that owing to the inefficiency of the metastatic process in terms of cancer cells, this sequential process is slow relative to tumor development and that this time base effects expression of metastatic pattern. This proposed scheme for metastatic pattern development provides evidence in favor of Ewing''s mechanical hypothesis for 2 histologic types of cancer, each growing in one designated primary site. It does not exclude a role for seed and soil effects in these and other cancers.