An adaptive technique for measuring and correcting the effects of patient motion during magnetic resonance image acquisition was developed and tested. A set of algorithms that can reverse the effects of object displacements and phase shifts was used. These algorithms essentially transfer the frame of reference of the image reconstruction from the static frame of the imager couch to the moving "visceral frame." An accurate record of tissue motion during image acquisition is required. To achieve this, the authors used specially encoded "navigator" echoes that are interleaved with the imaging sequence. Postprocessing of the navigator echo data provides a highly detailed record of the displacements and phase shifts that occur during imaging. Phantom studies demonstrated that the technique can directly correct image degradation caused by motion. In contrast to conventional artifact reduction techniques, such as ordered phase encoding and gradient moment nulling, this new method has a unique capacity to reduce motion unsharpness. Preliminary in vivo studies have demonstrated that the technique can markedly improve images degraded by voluntary motion and shows promise for addressing the problem of respiratory motion in thoracoabdominal imaging.