Hi Alle Meije,
Try setting up a job with the same structure as this one (also see attached).

matlabbatch{1}.spm.util.defs.comp{1}.dartel.flowfield = {'/.../u_rc1OAS2_0080_MR1_Template.nii'};
matlabbatch{1}.spm.util.defs.comp{1}.dartel.times = [0 1];
matlabbatch{1}.spm.util.defs.comp{1}.dartel.K = 6;
matlabbatch{1}.spm.util.defs.comp{1}.dartel.template = {'/.../Template_6.nii'};
matlabbatch{1}.spm.util.defs.comp{2}.id.space = {'/.../OAS2_0080_MR1.nii'};
matlabbatch{1}.spm.util.defs.out{1}.pull.fnames = {'/.../spm12/tpm/labels_Neuromorphometrics.nii'};
matlabbatch{1}.spm.util.defs.out{1}.pull.savedir.savepwd = 1;
matlabbatch{1}.spm.util.defs.out{1}.pull.interp = 0;
matlabbatch{1}.spm.util.defs.out{1}.pull.mask = 1;
matlabbatch{1}.spm.util.defs.out{1}.pull.fwhm = [0 0 0];
matlabbatch{1}.spm.util.defs.out{1}.pull.prefix = '';

All the best,
-John


On 23 May 2016 at 08:20, Alle Meije Wink <[log in to unmask]> wrote:
Maybe it's better to describe the core problem:

I have an atlas in MNI space (sampled at 1.5 mm)
  /path/to/aal_atlas.nii
and a set of inverse warp parameters to put it in native space of a T1 image: (sampled at 1 mm)
  /path/to/iy_T1.nii

If I call a script that sets
  matlabbatch{1}.spm.spatial.normalise.write.subj(1).def = {'/path/to/iy_T1.nii'};
  matlabbatch{1}.spm.spatial.normalise.write.subj(1).resample = {'/path/to/aal_atlas.nii'};

and then run
  spm_jobman('run',matlabbatch)

my would-be native-space atlas has dimensions 79x95x79 and voxel size 2x2x2 mm.

Neither the bounding box nor the voxel size make sense?
The example scripts I know automatically use the T1's settings - why not now?

Many thanks!