On the Nature of Inference