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