Breath-holding during MR imaging eliminates respiratory motion artifacts but places a major time constraint on data acquisition. This constraint limits image signal-to-noise ratio and hence spatial resolution. A new method, multiple breath-hold averaging, is presented that overcomes these time limitations. Several images are acquired in sequential breath-hold periods, separated by periods of normal breathing, and averaged. This averaged image shows the expected increase in SNR with surprisingly little blurring due to misregistration. SNR improvements can be traded for increased spatial resolution. The MBA methodology can also be applied to 3D data acquisitions, dynamic contrast acquisitions, and image subtractions.