PURPOSE: To investigate the effect of spatial resolution and parallel imaging acceleration factor on the quality of the motion estimates derived from image navigators with a three-dimensional (3D) gradient-recalled echo (GRE) acquisition with fat excitation (3D FatNavs) for neuroimaging at 7T. METHODS: Six healthy subjects were scanned for 10 min, during which time repeated GRE volumes were acquired during small movements-alternating between fat and water excitations (WaterNavs)-allowing retrospective decimation of the data to simulate a variety of combinations of image resolution and acceleration factor. Bias and error in the motion estimates were then compared across navigator parameters. RESULTS: The 2-mm, 4 x 4 accelerated data (TRvolume = 1.2 s) provided motion estimates that were almost indistinguishable from those from the full original acquisition (2 mm, 2 x 2, TRvolume = 5.2 s). For faster navigators, it was found that good accuracy and precision were achievable with TRvolume = 144 ms, using a lower spatial resolution (4 mm, 6 x 6 acceleration) to avoid the bias observed at exceptionally high acceleration factors (8 x 8 or higher). Parameter estimates from WaterNavs and FatNavs showed close agreement with FatNavs, with better performance at exceptionally high acceleration factors. CONCLUSION: Our data help to guide the parameter choice for 3D FatNavs when a compromise must be reached between the quality of the motion estimates and the available scan time.