The article discusses the use of TLS (Translation, Libration, and Screw-rotation) parameters in macromolecular refinement to model anisotropic displacements. The authors present a detailed implementation of TLS parameterization in the macromolecular refinement program REFMAC, which uses fast Fourier transforms to efficiently calculate derivatives and improve refinement speed. The method is applied to two examples: glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and a transcription activator GerE, both with data to 2.0 Å resolution, where individual anisotropic refinement is not feasible. The inclusion of TLS parameters improves refinement statistics, particularly the R and free R values, by accounting for significant differences in displacement parameters between molecules. The study demonstrates that refining TLS parameters first, followed by refining atomic coordinates and individual B factors, is an effective strategy. The results show that the TLS model provides a good fit to the observed data and can account for the anisotropic component of atomic displacements, with the anisotropy values being reasonable. The article also highlights the utility of TLS parameters in modeling differences in displacement parameters between non-crystallographically related molecules, such as in the case of NCS restraints.The article discusses the use of TLS (Translation, Libration, and Screw-rotation) parameters in macromolecular refinement to model anisotropic displacements. The authors present a detailed implementation of TLS parameterization in the macromolecular refinement program REFMAC, which uses fast Fourier transforms to efficiently calculate derivatives and improve refinement speed. The method is applied to two examples: glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and a transcription activator GerE, both with data to 2.0 Å resolution, where individual anisotropic refinement is not feasible. The inclusion of TLS parameters improves refinement statistics, particularly the R and free R values, by accounting for significant differences in displacement parameters between molecules. The study demonstrates that refining TLS parameters first, followed by refining atomic coordinates and individual B factors, is an effective strategy. The results show that the TLS model provides a good fit to the observed data and can account for the anisotropic component of atomic displacements, with the anisotropy values being reasonable. The article also highlights the utility of TLS parameters in modeling differences in displacement parameters between non-crystallographically related molecules, such as in the case of NCS restraints.