Musgrave (1981) proposed a typology of assumptions, developed further by Mäki (2000), to defend the idea that the truth of assumptions is often important when evaluating economic theories against those economists who consider only predictive success to be relevant for this purpose. In this paper I propose a new framework for this typology that sheds further light on the issue. The framework consists of a distinction between first‐order assumptions that state the absence or lack of effect of some factor F, and second‐order assumptions that explicate the purposes for which or the reasons why particular first‐order assumptions are imposed. Given this distinction, Musgrave's main contention can be reformulated as the claim that, even though the falsity of first‐order assumptions is often unproblematic, it is important that the second‐order assumptions be true. I go on to introduce the notion of a tractability assumption, which is a second‐order assumption according to which a first‐order assumption is imposed in order to make a particular problem tractable. It is argued that a realist will want to relax a first‐order assumption imposed for reasons of tractability as such assumptions are not even approximately true. These amendments to the Musgrave–Mäki typology are suggested in order to improve our understanding of what moves scientists when they choose particular first‐order assumptions, many of which are false, and in order to argue that the practice of doing so can be supported from a realist perspective of science.