By default (including auto-generated preprocessing scripts), the smoothing kernel in 3dMerge is 4mm. The example in the 3dMerge documentation uses this size kernel for voxels that are 3.75x3.75x3.5. The voxels in my data are slightly smaller (3x3x4). A reviewer has raised the question as to whether the assumptions of Gaussian Random Field theory are violated when using a smoothing kernel this small. Since 4mm is the default, I presume this is the recommended kernel size, for averaged sized voxels. Can someone please help address the concern?