Many new developments in tissue engineering rely on the culture of primary tissues which is composed of parenchymal and mesenchymal (stromal) cell populations. Because stroma regulates parenchymal function, the parenchymal:stromal cell (P:S) ratio will likely influence culture behavior. To investigate parenchymal-stromal cell interactions, the P:S ratio was systematically varied in a human bone marrow (BM) culture system, measuring the output of mature cells, immature progenitors (colony forming units-granulocyte/macrophage [CFU-GM]), and primitive stem cells (long-term culture-initiating cells [LTC-IC]). When parenchymal CD34-enriched cells were grown without stroma, cell and CFU-GM output increased linearly as inoculum density was increased, resulting in constant cell and CFU-GM expansion ratios. On irradiated preformed stroma (IPFS), culture output was significantly higher and less dependent on CD34-enriched cell inoculum density, resulting in greater expansion ratios at lower inoculum densities. The number of IPFS cells required to support CD34-enriched cells was independent of the CD34-enriched cell number, suggesting that IPFS did not provide discrete niches, but instead acted through soluble signals. Experiments using conditioned medium (CM) from IPFS confirmed the presence of soluble signals, but CM did not completely substitute for direct contact between CD34-enriched cells and IPFS. Because of known differences between IPFS and stroma growing within BM mononuclear cell (MNC) cultures, experiments were next performed using mixtures of CD34-enriched and CD34-depleted fractions of MNC. When inoculated with a fixed CD34+lin− cell number, culture output was optimal near the P:S ratio of the unmanipulated MNC sample and declined as CD34− cell number was increased or decreased. In cultures inoculated with a fixed total cell number, CFU-GM output increased as CD34+lin− cell number was increased, whereas LTC-IC output reached a plateau. These data suggest that a limited number of LTC-IC supportive niches were present in MNC stroma, whereas IPFS lacks these niches and acts predominantly through a less potent soluble mechanism. These studies underscore the importance of parenchymal-stromal cell interactions in the ex vivo reconstitution of tissue function and offer insight into the nature of these interactions in the human BM culture system.