Nuclear magnetic relaxation provides a powerful method giving insight into molecular motions at atomic resolution on a broad time scale. Dynamics of biological macromolecules has been widely exploited by measuring (15)N and (13)C relaxation data. Interpretation of these data relies almost exclusively on the use of the model-free approach (MFA) and its extended version (EMFA) which requires no particular physical model of motion and a small number of parameters. It is shown that EMFA is often unable to cope with three different time scales and fails to describe slow internal motions properly. In contrast to EMFA, genuine MFA with two time scales can reproduce internal motions slower than the overall tumbling. It is also shown that MFA and simplified EMFA are equivalent with respect to the values of the N-H bond length and chemical shift anisotropy. Therefore, the vast majority of (15)N relaxation data for proteins can be satisfactorily interpreted solely with MFA.