Hi Amelia, 1. As mentioned in a recent post, the dispersion is equal to 1 minus the average dot product of every sample orientation with the main dyad. As the bedpostx effectively produces in each voxel a distribution for any modelled fibre orientation, the dispersion simply characterizes how wide this distribution is around the dyad, which represents the mean of the distribution. 2. Have you been using different FSL versions for different subjects? The dispersion file is only produced by relatively recent FSL versions. 3. I am not sure I understand the question here. After running probtrackx, you should observe more coherent results when tracking through regions with low dispersion. Going through regions with high dispersion will produce more spatially distributed results and potentially smaller connection probabilities. Hope that helps, Stam ----- Original Message ----- From: "Versace, Amelia" <[log in to unmask]> To: <[log in to unmask]> Sent: Sunday, August 01, 2010 1:39 PM Subject: [FSL] Dyads Dispersion Dear FSL experts, I ran probabilistic tractography on a quite large sample of HC and I realized that in some of them the derived fdt_paths are much more informative (more vs few fibers reconstructed) than in others. So, going back to the bedpost folders, I realized that some of my subjects, but not all of them, have an extra output, i.e.= dyads*_ dispersion. I searched the FSL archive, but I found just one post about this. So, can anybody 1. provide a better description of or references to learn more about this output? 2. suggest a possible explanation of why just some of my subjects have this output (is this noise-related??) and confirm that the presence of dyads*_ dispersion can explain the difference seen in the estimated fdt_paths 3. suggest a way to implement the use of this output in the estimation of my fdt_paths, I order to confidently include these "different" subjects in one group Thank you very much!!! Kind Regards Amelia