Extending the Private Language Argument — LessWrong