A better comprehension of the rheology of the lithosphere is required to relate long and short term deformation regimes and describe the succession of events leading to earthquakes. But our vision of the rheology is blurred because gaps exist between visions of geologists, experimentalists and modellers. Geologists describe the evolution of a structure at regional-scale within geological durations. Specialists of experimental rheology control most parameters, but laboratory time constants are short and they often work on simple synthetic rocks. Specialists of modelling can choose any time- and space-scales and introduce in the model any parameter, but the resolution of their models is low compared to natural observations, and mixing short-term and long-term processes is uneasy. It seems now clear that there is not one rheological model applicable to all contexts and that rheological parameters should be adapted to each situation. We will work on exhumed crustal-scale shear zones and describe them in their complexity, focussing on strain localisation and high strain structures that can lead to fast slip events. A number of objects will be studied, starting from geological description (3D geometry, P-T-fluids estimates and dating), experimental studies of rheological properties of natural sampled rocks and numerical modelling. We will set an Argon-dating lab to work on dense sampling for dating along strain gradients in order to overcome local artefacts and quantify rates of strain localisation. We will deform in the lab natural rocks taken from the studied objects to retrieve adapted rheological parameters. We will model processes at various scales, from the lab to the lithosphere in order to ensure a clean transfer of rheological parameters from one scale to another.