On determining what counts while counting: Aspects of language testing where diversity is the standard