Публикация

I guess it depends on expectations of what the data should look like. eg for vowel formants big outliers are highly likely to be measurement errors