Encoding words and pictures: A positron emission tomography study