Instrumentalism is an approach to science that treats a theory as a tool and only as a tool for computation; it dispenses with the concept of truth.Conventionalism treats a theory as true by convention if it forms a pattern of observations from which correct predictions can be made.Operationalism denies meaning to the concepts of a theory unless they can be defined operationally. It is argued in this paper that truth-value is indispensable to science, because a theory can be rejected only (...) if an empirical consequence is false and if falsity of a conclusion entails falsity of a premise. This undermines the above positions. The fourth interpretation isinduction. Induction, by contrast, uses the notion of truth-value. What is focused on here is its reliance on the ultimacy ofobservation. The present thesis is that instrumentalism, conventionalism, and induction are different attempts to handle observations. The common problem is the gap between data and theory.All these interpretations share a philosophy of observationalism. The aim of this paper is to show that the several orthodox interpretations of science all fail to solve the problem of the data-theory gap, and to show that they all presuppose a philosophy of observation. (shrink)