Mental scanning of visual images generated from verbal descriptions: Towards a model of image accuracy