Hematopoiesis, the formation of mature blood cells from stem (LTC-IC) and progenitor (CFU-GM) cells in the bone marrow, is a complex tissue-forming process that leads to many important physiological functionalities. Consequently, a functioning ex vivo hematopoietic system has a variety of basic scientific and clinical uses. The design and operation of such a system presents the tissue engineer with challenges and choices. In this study, three culture variables were used to control ex vivo human hematopoiesis. Systematic variation of inoculum density (ID), medium exchange interval (MEI), and the use of preformed stroma (PFS) showed that (1) all three variables significantly influenced culture performance, (2) the three variables interacted strongly, and (3) the variables could be manipulated to achieve the optimization of different performance criteria. Donor-to-donor variability in culture performance was great at low ID but was minimized at higher ID. PFS had a large positive effect on cell and CFU-GM output at low ID, but had minimal effect at higher ID. In fact, PFS caused a decrease in LTC-IC output at high ID. The effects of PFS indicated that stromal cell elements became more limiting than proliferative cell elements as ID was reduced. In cultures without PFS, maximum cell output was obtained with high ID using a short MEI, whereas the greatest cell expansion ratio was obtained at low ID with an intermediate MEI. Maximum CFU-GM output was obtained from cultures with high ID using a short to intermediate MEI, whereas the greatest CFU-GM expansion ratio was obtained at intermediate ID with an intermediate MEI. The addition of PFS altered the locations of these maxima. In general, PFS moved the maxima to lower ID, and culture output became more sensitive to MEI. Therefore, the optimization of one performance criterion always resulted in a decline of the others. This study demonstrates that ex vivo tissue function is sensitive to many culture variables in an interactive fashion and that systematic multivariable studies are required to characterize tissue function. Once the effects of individual variables and their interactions are known, this knowledge can be used to optimize tissue performance with respect to desired criteria.