Frank von Delft wrote:

Jianghai Zhu wrote:

Adding riding hydrogens into the structure was a powerful validation 
tool.  But now we all add riding hydrogens into the refinement since it 
makes the refinement better.  At the same time, we lose some power in 
the validation.  We still think it is worth doing so, don't we?


No:  they are RIDING hydrogens, they do not influence the refinement 
(i.e. bumps), only the Rfactors/maps.  Which is not the same as you ask 
for with the phi/psi angles.

Not at all. Adding the hydrogen atoms from high to low resolution can be viewed as adding additional anti-bumping restraints. The improvement in the R/Rfree factors in this case is mostly attributed to overall model geometry improvement and less attributed to adding additional scattering power to the structure (depending on resolution).
Saying differently, the X-ray gradients for H atoms are zero, H atoms come to Fcalc and hence to X-ray target. However, the geometry restraints target "sees" hydrogens as good as other atoms.

Cheers,
Pavel.