In response to an argument by Baird, Andrich, Hopfenbeck and Stobart (2017), Michael Kane states that there needs to be a better fit between educational assessment and learning theory. In line with this goal, Kane will examine how psychometric constraints might be loosened by relaxing some psychometric "rules" in some assessment contexts. In this vein, the author will try to make two points. First, the assumptions inherent in an assessment programme are not necessarily so clear a priori, and as an initial step in designing and evaluating the programme, it is useful to be as clear as possible about the assumptions being adopted (Kane, 2006, 2013), and in particular, to avoid making assumptions that are not necessary in the case at hand. Second, depending on the intended interpretation and use of the scores, the population to be assessed, and the context in which the assessment occurs, various technical requirements may or may not be critical. In cases where violating a particular psychometric "rule" does not interfere with the goals of the programme, there is no need to be constrained by the rule. As noted by Baird et al. (2017), "our assessment tools constrain and shape the metaphors of learning with which we operate" (p. 320); the author suggests that by limiting psychometric requirements to those that are necessary to the case at hand, it can minimise their impact. [This article is a commentary on Jo-Anne Baird, et al.'s "Assessment and Learning: Fields Apart" (EJ1149539).]