When explanations compete: the role of explanatory coherence on judgements of likelihood