*Hidden-variable models aim to reproduce the results of quantum theory and to satisfy our classical intuition. Their refutation is usually based on deriving predictions that are different from those of quantum mechanics. Here instead we study the mutual compatibility of apparently reasonable classical assumptions. We analyse a version of the delayed-choice experiment which ostensibly combines determinism, independence of hidden variables on the conducted experiments, and wave-particle objectivity (the assertion that quantum systems are, at any moment, either particles or waves, but not both). These three ideas are incompatible with any theory, not only with quantum mechanics.Introduction. -Most of the quantum formalism was in place by 1932 [1]. Since then, quantum theory has been spectacularly successful across all the investigated scales and systems. Yet many of its results contradict both common sense and classical physical intuition. Wave-particle duality, superposition, and entanglement are among these counterintuitive features [2,3] and the "strictly instrumentalist" [4] core of quantum theory abandons many familiar traits of classical physics. As a result, there are profound differences of opinion on the meaning of quantum theory and the desire to explain or even to remove its puzzling properties [2][3][4][5].Hidden-variable (HV) theories endeavor to give a satisfactory representation of our intuition while reproducing the experimental predictions of quantum theory [2-6]. Imposing classical concepts (determinism, versions of locality, etc) on HV models constrains the resulting probability distributions. This may lead to "paradoxes," i.e., an incompatibility of the allegedly reasonable assumptions with the predictions of quantum theory.With the advent of quantum technologies [7,8] we can now realize classic gedankenexperiments and develop new tests to confront the predictions of HV theories with those of quantum mechanics. When the latter are experimentally confirmed, HV models fail the crucial test of adequacy and, unless some loophole for the experiment is found [9], should either be abandoned or amended to include deep, possibly unacceptable [5], conspiratorial correlations. The loopholes, in turn, may be countered by more sophisticated set-ups [10].Implicit in these debates is the premise that classically reasonable assumptions form a world view which, although experimentally inadequate, is nevertheless consistent. We question this tacit assumption and investigate the mutual compatibility of three classical requirements (determinism, independence, objectivity). Specifically, in the context of waveparticle duality and delayed-choice experiments [11,12] we inquire if it is possible to find any probability distribution that satisfies all three classical constraints. Here we answer this question in the negative: determinism, independence, and objectivity are incompatible; i.e., no such probability distribu-