Seismic anisotropy is a critical material property which allow us to identify structures, deformation behaviour and flow in the Earth’s mantle. Interpreting seismic signals, however, requires extensive knowledge of the underlying physical properties of minerals. Large scale seismic anisotropy can arise from the anisotropy of individual crystals and their alignment due to specific mechanisms. In the upper mantle, strong seismic anistotropy is explained by lattice preferred orientation (LPO) of olivine due to mantle flow. In the mantle transition zone, seismic anisotropy is weaker and there is no consensus on the source of this anisotropy. The 410 km depth discontinuity, which separates these two regions, is attributed to the transformation of olivine into its high pressure polymorph wadsleyite. Experiments show that this transformation can happen either by a martensitic-like mechanism or by a non-martensitic-like mechanism. In the first case, the LPO of the parent olivine will be inherited by the daughter wadsleyite whereas in the second case, the daughter wadsleyite will form with a random orientation. Hence, if this transformation is martensitic-like, it can be a source for seismic anisotropy in the mantle transition zone. Here we present a project in which an experimental study of the olivine-wadsleyite transformation is used to make inferences about the prevalent transformation mechanism at mantle conditions. We rely on laboratory experiments to produce the transformation at the conditions of the 410 km depth discontinuity by compressing either polycrystals or single crystals of olivine in diamond anvils cells at pressures up to 20 GPa and temperatures of 1600-1800 K. Using multigrain crystallography we evaluate the orientation and position of individual grains in the sample. We then use the experimental LPO to model the seismic anisotropy resulting from the transformation. Finally, we simulate different observables as anisotropy or seismic reflections from the Earth’s transition zone and compare our predictions with real seismic data.