Steady-state free precession (SSFP) pulse sequences employing gradient reversal echoes and short repetition time (TR) between successive rf excitation pulses offer high signal-to-noise ratio per unit time. However, SSFP sequences are very sensitive to motion. A new SSFP method is presented which avoids the image artifacts and loss of signal intensity due to motion. The pulse sequence is designed so that the time integral of each of the three gradients is zero over each TR time interval. The signal then consists of numerous echoes which are superimposed. These echoes are isolated by combining the data from N different scans. In each scan a specific phase shift is added during every TR interval. Each of these N isolated echoes produces a motion-insensitive, artifact-free image. Because all the echoes are sampled simultaneously, the signal-to-noise ratio per unit time in this SSFP method is higher than in existing SSFP techniques which sample only one echo at a time. The new method was implemented and used to produce both two- and three-dimensional images of the head and cervical spin of a human patient. In these images the high signal intensity of cerebrospinal fluid is preserved regardless of its motion. Further work is required to evaluate the imaging parameters (TR, TE, rf tip angle) so as to give optimal tissue contrast for the various echoes.