Cells from a cloned line of spontaneously transformed 3T3 cells had a colony-forming efficiency in agar (CFEag) of about 10-15% and induced poorly differentiated sarcomas when injected into nude mice. These tumor cells were recultured in vitro and tested for their ability to grow in suspension. Initially, the tumor cells had a CFEag which was a hundredfold to a thousandfold lower than the cells that had been grown only in vitro. After 5-8 further weekly passages, however, the tumor lines recovered their original ability to grow in agar. For determination as to whether this increased CFEag was due to selection of cells with a higher CFEag from the tumor cell population or to adaptation of many of the tumor cells to agar growth, clones were isolated directly from a primary tumor, and each was tested weekly for agar colony formation. All of the tumor clones, as well as the uncloned tumor line, were able to recover their original ability to grow in agar. However, one tumor clone had a relatively high CFEag in the first assay, so the selection hypothesis could not be totally excluded. The initial low CFEag of the recultured cells was not due to the presence of normal nude mouse cells in the population. Before in vivo growth no clones could be isolated from the sublines that had as low a CFEag as the tumor cells isolated after in vivo growth. Tumor cells that had been repeatedly passaged in vivo still had a much lower CFEag than the input cells upon explantation into culture. The results suggest that phenotypic alterations observed during tumor growth and subsequent cultivation have an epigenetic basis.