The great majority of compositional models in distributional semantics present methods to compose distributional vectors or tensors in a representation of the sentence. Here we propose to enrich the best performing method (vector addition, which we take as a baseline) with distributional knowledge about events, outperforming our baseline.