Speech perception at its best: Extracting linguistic information from acoustically underspecified input. The case of singing.