Large Language Models as Proxies for Theories of Human Linguistic Cognition