This brings up another problem: Do we try to interact in authentic native- like speech or do we modify our input to be understood better.