It is proposed to convert nuclear Overhauser effects (NOES) into relatively precise distances for detailed structural studies of proteins. To this purpose, it is demonstrated that the measurement of NOE buildups between amide protons in perdeuterated human ubiquitin using a designed N-15-resolved HMQC-NOESY experiment enables the determination of H-1(N)-H-1(N) distances up to 5 angstrom with high accuracy and precision. These NOE-derived distances have an experimental random error of similar to 0.07 angstrom, which is smaller than the pairwise rmsd (root-mean-square deviation) of 0.24 angstrom obtained with corresponding distances extracted from either an NMR or an X-ray structure (pdb codes: 1D3Z and 1UBQ), and also smaller than the pairwise rmsd between distances from X-ray and NMR structures (0.15 angstrom). Because the NOE contains both structural and dynamical information, a comparison between the 3D structures and NOE-derived distances may also give insights into through-space dynamics. It appears that the extraction of motional information from NOES by comparison to the X-ray structure or the NMR structure is challenging because the motion may be masked by the quality of the structures. Nonetheless, a detailed analysis thereof suggests motions between beta-strands and large complex motions in the alpha-helix of ubiquitin. The NOE-derived motions are, however, of smaller amplitude and possibly of a different character than those present in a 20 ns molecular dynamic simulation of ubiquitin in water using the GROMOS force field. Furthermore, a recently published set of structures representing the conformational distribution over time scales up to milliseconds (pdb: 2K39) does not satisfy the NOES better than the single X-ray structure. Hence, the measurement of possibly thousands of exact NOES throughout the protein may serve as an excellent probe toward a correct representation of both structure and dynamics of proteins.