Recent excitement about the impending completion of the human genome decoding tends to support the idea that a reductionist view of biology, disease, and subsequent therapy will result in substantial health care advances, and that molecular imaging will play an important part in the discovery and application of those advances. It is clear that the reductionist approach and methodology have greatly improved our understanding of discrete biologic processes. These advancements have led to the identification of unique molecular targets, the associated processes they modulate, and the development of drugs that affect those processes. STI-571 (Gleevec, Novartis) is a recent example of a small molecule that was specifically selected for its binding affinity to a molecular target; its resulting clinical effects in treating chronic myelogenous leukemia have been impressive. The clinical success of Gleevec gives hope that other targeted drugs, and by extension their related imaging agents, will be useful in the diagnosis and treatment of disease. However, although the ability of this relatively simple model to predict a result is high in controlled experiments, our ability to predict complex characteristics, such as a biologic phenotype, from molecular data in an in vivo model or patient is far less reliable. This difficulty lies in the complexity of determinism of biologic phenotypes, which in most cases represent the summation of multiple signals and processes that, despite similar input, may even change output under different conditions. Evolution has apparently endowed us with multiple pathways to accomplish the same thing and multiple modulators of all these pathways to protect us from a variety of noxious or toxic exposures. These conditions and the resulting signals are determined in many cases by the microenvironment, both at the cellular and at the extracellular levels. In relation to biologic phenotype, the microenvironment can be thought of as three overlapping compartments: interstitial, intracellular, and intercellular. Relatively discrete signaling pathways interlink these compartments. Cellular interactions are so complex and so numerous that modeling them is considered a monumental task that may be possible in a decade, given substantial funding and with use of sophisticated supercomputers. However, even a complete understanding of one of these compartments, such as the intracellular microenvironment, will represent a major feat. The addition of the interstitial and intercellular interactions may be far more complex than we can model in the foreseeable future. This situation is a major problem for drug development. Reductionist analysis has provided a plethora of individual targets for which drugs can be rationally designed, but the key to success, as well as the approval, of a drug is demonstration of effectiveness. Almost without exception, the basis of a drug’s effectiveness is alteration of a pathologic phenotype. In the absence of substantial mitigation or reversal of such a phenotype, substantial health benefit is not likely to occur. For this reason, there is increasing interest by drug companies in an integrationist approach to drug selection in the early stages of development. Even with conventional cancer therapy, it has long been recognized that the tissue environment can greatly influence therapeutic efficiency. Tissue perfusion, oxygen tension, and pH are well-described examples. With molecularly targeted drugs, this influence appears more complex and, in fact, may force us to look at smaller and smaller components of the microenvironment. Although there is hope that all the various components of the microenvironment can eventually be modeled, the current and most practical solution for drug developers is in vivo quantitative imaging of specific phenotypic changes associated with disease outcome. With in vivo detection and quantitative monitoring of these phenotypic changes, which are the result of multiple inputs from both inside and outside the Acad Radiol 2001; 8:1192–1193