SCOP: Evaluating the Comprehension Process of Large Language Models from a Cognitive View