The motivation of this study is to understand the seismic anisotropy observations from various subduction regions of the world. In subduction zone backarcs both trench-parallel and trench-normal seismic anisotropy, or fast wave polarization direction of shear wave, are observed. In the mantle the general assumption is that seismic anisotropy is caused by Lattice Preferred Orientation (LPO) of olivine minerals and that the direction of anisotropy is an indicator of the direction of mantle flow. The complex pattern of seismic anisotropy observations suggests that the flow geometry in the vicinity of subduction zones differs at different subduction zones with some subduction zones having trench perpendicular flow, consistent with corner flow in the mantle wedge while other subduction zones have trench parallel flow, consistent with a mode of flow where material from the mantle wedge flows around the edges of the slab. It should be noted that the direction of LPO orientation can also be modified by the presence or absence of water, pressure, and temperature in the mantle and that it is possible that the difference in anisotropy observations reflects a difference in water content or thermal structure of back arcs. The aim of this study is to test whether the flow geometry of mantle in numerical subduction calculations can influence the direction of seismic anisotropy and if we parameters that control the pattern of flow can be identified. In this study we explicitly assume that seismic anisotropy occurs only due to plastic and dynamic re-crystallization of mantle mineral forming LPO. To approach the problem two different models are formulated. In one of the models the trench evolves self-consistently, with no prescribed artificial zones of weakness. The self-consistent model has a sticky-air layer at the top of the model domain that mimics a "free-surface." The other model has the same initial conditions but a trench-migration velocity boundary condition is imposed to the model. The mantle flow pattern for the self-consistent model is consistent with the 2D corner flow with no flow around the trench and no trench migration. However when the trench-migration velocity boundary condition is imposed, 3D flow around the mantle is observed. The stress field from these simulations are used to calculated instantaneous strain axis directions which correlate with LPO directions. The LPO orientations are measured from the models showing that the seismic-anisotropy direction is primarily trench-perpendicular for both models. Because the models have different flow patterns, the trench-perpendicular anisotropy alignment that is calculated for both the models is a bit puzzling. It could be that factors such as high temperature and non-linear rheology cause the LPO direction to align trench perpendicular in both the cases. It can also be possible that the 3D vertical flow is not strong enough to cause change in orientation of the LPO direction. From the present study it can be concluded that by looking at the LPO direction nature of mantle flow might not be predicted. This suggests that in addition to flow direction other factors such as the presence of water in mantle wedge, pressure, and high temperature due to viscous coupling modify the seismic anisotropy directions.