The missing self model predicts that NK cells adapt somatically to the type as well as levels of MHC class I products expressed by their host. Transgenic and gene knock-out mice have provided conclusive evidence that MHC class I genes control specificity and tolerance of NK cells. The article describes this control and discusses the possible mechanisms behind it, starting from a genetic model to study how natural resistance to tumors is influenced by MHC class I expression in the host as well as in the target cells. Data on host gene regulation of NK-cell functional specificity as well as Ly49 receptor expression are reviewed, leading up to the central question: how does the system develop and maintain "useful" NK cells, while avoiding "harmful" and "useless" ones? The available data can be fitted within each of two mutually none-exclusive models: cellular adaptation and clonal selection. Recent studies supporting cellular adaptation bring the focus on different possibilities within this general mechanism, such as anergy, receptor calibration and, most importantly, whether the specificity of each NK cell is permanently fixed or subject to continuous regulation.