Probing Language Models from A Human Behavioral Perspective