Differences in acquisition protocols or hardware result in measurable changes in image characteristics. These differences affect distributional properties and can also affect the spatial and temporal characteristics of the images. Likewise, the distribution of population characteristics can change between imaging centers, and the distribution of label characteristics depends on annotators. Shift in these three factors (image, population, and labels) typically yields inferior performance for machine learning algorithms. This chapter first defines these concepts, shows which magnetic resonance imaging parameters can cause shifts, and how shifts can be quantified. Then, domain adaptation and harmonization approaches that can minimize domain shift are presented, and finally, other approaches to improve generalization under domain shift are discussed.