The ability of tumor cell lines to form experimental pulmonary metastases is determined in part by characteristics which are stable over many cell generations; in part by characteristics that are acquired by adaptation or phenotypic instability; but also in part by characteristics which may change over less than one cell generation. This study was designed to examine the hypothesis that tumor cells secrete and respond to paracrine factors which can reversibly modulate metastasis. The number of experimental lung metastases increased for 13762NF rat mammary adenocarcinoma cell clones MTF7 and MTLn3 as they approach 100% confluence. This observation corresponded to increased attachment to bovine brain capillary and bovine corneal endothelial monolayers and to ability of tumor cells to invade reconstituted basement membrane barriers in the Membrane Invasion Culture System (MICS), but did not correspond to cell cycle distribution, susceptibility to NK or PMN cell killing or average cell size/Coulter volume. While changing confluence did not qualitatively alter metastatic potential, modification of metastasis in a quantitative manner suggested that some properties pertinent to metastasis are transient and manipulatable. Tumor cell-conditioned medium (CM) collected from donor cells grown to defined levels of confluence when placed onto recipient cells reversibly raised or lowered metastatic potential depending upon the medium source and confluence of the recipient cells. CM from 20% confluent donor cultures reduced recipient cell metastatic potential. In contrast CM from 100% confluent cultures increased metastatic potential of subconfluent cells. Replacement with fresh unconditioned medium or leaving the medium unchanged did not alter experimental metastasis. These data suggest that metastasis involves steps which may be influenced by paracrine factors elaborated by tumor cells.