Vocabulary test development often appears to be based on the design principles of previous tests, without questioning or empirically examining the assumptions underlying those principles. Given the current proliferation of vocabulary tests, it seems timely for the field of vocabulary testing to problematize some of those traditionalised assumptions, based on more current research. This article begins this process by challenging two common assumptions key to vocabulary test development: (1) the counting unit of word families, and (2) the 1,000-word band divisions in the use of frequency. Based on existing literature and an analysis of corpus-based coverage figures, the article frames future research agendas by arguing that the lemma may be a more useful counting unit for vocabulary assessment and pedagogy. Further, it argues that the traditional 1,000-item frequency bands are not optimal. Smaller 500-item bands would be more informative at the higher frequencies, and bands larger than 1,000 items would be adequate at lower frequencies. Because most vocabulary tests are aimed at beginner to intermediate learners of English as a foreign language, these new empirically informed conventions should facilitate the development of more informative vocabulary tests.