Tumor cells are generally regarded as poor stimulators of naive T cells. In contrast, dendritic cells (DCs) are highly specialized in this function, and are therefore likely to be important intermediaries in the process of stimulating T cell responses to tumors. While providing solid evidence that DCs participate in antitumor immunity has proved difficult, several lines of evidence point in this direction. First, animal models involving bone marrow chimeras have shown that cells of hematopoeitic origin are required to elicit T cell responses to whole-tumor vaccines. Second, compared with other cells of hematopoeitic origin, DCs are particularly well-equipped to cross-present exogenous antigens to CD8+ T cells, a critical function if intermediary cells are involved. Third, tumor-infiltrating DCs purified from tumor samples have the capacity to cross-present tumor antigens in vitro. Finally, priming of anti-tumor T cell responses can be abrogated in new in vivo models in which DCs can be specifically depleted. It is therefore significant that DCs in cancer patients are often kept in an immature or dysfunctional state, thereby preventing stimulation of tumor-specific T cells. This review describes the different steps required for DCs to elicit T cell responses to tumor-associated antigens, and highlights processes that are amenable to intervention as therapy. We conclude that effective anti-tumor activity may be dependent on the ability to re-program DCs resident in the host, perhaps even when transferred autologous DCs generated ex vivo are used as vaccines. In this context, recruiting the activity of cells of the innate immune system to condition host DCs may help elicit more effective T cell-mediated responses.