Dear Tine, as it is about the 'bulk' density, you should not correct for stones, roots or whatever. It is an all inclusive value. You should however take a REV, representative elemetary volume, thus a representative sample, not dominanted by a single stone, root, etc. With such a domination, you should take a bigger sample, or take it elsewhere. Another option is to take that many samples that your average does not change anymore. Take care btw to saturate your sample first before leveling it to your volume, otherwise, with swelling soils, your repeatability will be poor.

Dear Tine, as the roots are mostly water - when you dry the sample (to measure the total dry weight) you will have removed most of the mass associated with the roots anyway. When you have collected the samples (ie from a fixed volume) and you are preparing them for drying, you could try to remove the larger roots without taking any attached soil - this may not be easy or possible if the soil has a lot of clay in it. The error of including fine roots is likely to be less than that of the actual sampling method (particularly the errors associated with compaction when driving cores into the ground or the trimming of samples once removed). I would like to follow up on a comment from Gerben. It is worth calculating the density of the fine earth fraction (ie the density of the soil between stones as this gives you an idea of the relative density that roots experience (ie related to mechanical impedance) if you are comparing sites with and with out stones - of layers with different stone content. This may not strictly be the "bulk" density of the overall soil, but it is important information depending on your reason for measuring it is. If you are measuring bulk density to convert carbon content of mass/mass basis to mass/area basis, then you will need to know the volume of fine fraction (whole sample minus stone volume) to be able to calculate the mass/area.