LLM language comprehension task tests reveal insensitivity to underlying meaning