What hypothesis tests are not: a reply to Johnson