The influence of physiological motion on magnetic resonance diffusion imaging techniques is studied. First, the applicability of the steady-state free precession (SSFP) imaging sequence for imaging of diffusion is studied with a novel, fast, and highly accurate computer simulation based on conditional random walk. By using measurements of in vivo brain motion, it is shown that the SSFP sequence cannot provide accurate measurements of in vivo diffusion attenuation. Next, the use of navigator echo for -correction of otion artifacts in the pulsed-gradient spin-echo sequence is studied. The navigator technique is also extended for the fast spin-echo (FSE/RARE) sequence. However, itis shown that the navigator technique cannot provide accurate measurements of diffusion in three orthogonal directions and needs cardiac gating to minimize the influence of spatially varying motion. Finally, the major result of this work is a novel line scan diffusion imaging technique (LSDI). The LSDI technique is inherently insensitive to motion artifacts because it does not use phase encoding. It is shown, that high quality diffusion maps can be obtained rapidly with the LSDI technique on both 0.5T and 1.5T systems without the use of head restraints or cardiac gating. Furthermore, it is shown that these diffusion maps compare very well with diffusion maps obtained using diffusion weighted echo-planar imaging (EPI) In contrast to the EPI technique, LSDI does not require modified gradient hardware and can easily be implemented on conventional scanners. The main disadvantage of the LSDI technique, is that multi-slice imaging is considerably slower than with EPI. However, the feasibility of using te LSDI technique for clinical evaluation of acute ischemic stroke in less than ten minutes is demonstrated. Thus, LSDI should increase the general availability of robust clinical diffusion imaging dramatically.