The effect of direction of comparison on the selection of causal explanations