Magnetic resonance imaging techniques designed for sensitivity to microscopic motions of water diffusion and blood flow in the capillary network are also exceptionally sensitive to bulk motion properties of the tissue, which may lead to contrast artifact and large quantitative errors. The magnitude of bulk motion error that exists in human brain perfusion/diffusion imaging and the inability of cardiac gating to adequately control this motion are demonstrated by direct measurement of phase stability of voxels localized in the brain. Two methods are introduced to reduce bulk motion phase error. The first, a postprocessing phase correction algorithm, reduces coarse phase error but is inadequate by itself for quantitative perfusion/diffusion MRI. The second method employs orthogonal slice selection gradients to define a column of tissue in the object, from which echoes may be combined in a phase-insensitive manner to measure more reliably the targeted signal attenuation. Applying this acquisition technique and a simplistic model of perfusion and diffusion signal attenuations yields an estimated perfusion fraction of 3.4 +/- 1.1% and diffusion coefficient of 1.1 +/- 0.2 x 10(-5) cm2/s in the white matter of one normal volunteer. Successful separation of perfusion and diffusion effects by this technique is supported in a dynamic study of calf muscle. Periods of normal blood flow, low flow, and reactive hyperemia are clearly distinguished in the quantitative perfusion results, whereas measured diffusion remained nearly constant.