didn't resegment because utterance numbers changing stopped because fit experimental structure