Splatting is a classical volume rendering technique that has recently gained in popularity for the visualization of point-based suface models. Up to now, there has been few publications on its adaptation to time-varying data. In this paper, we propose a novel frame-to-frame coherent view-aligned sheet-buffer splatting of time-varying data, that tries to reduce as much as possible the memory load and the rendering computations taking into account the similarity in the data and in the images at successive instants of time. The results presented in the paper are encouraging and show that the proposed technique may be useful to explore data through time.