The rational expectations hypothesis (REH) is the standard approach to expectations formation in macroeconomics. We discuss its compatibility with two strands of Karl Popper's philosophy: his theory of knowledge and learning, and his "rationality principle" (RP). First, we show that the REH is utterly incompatible with the former. Second, we argue that the REH can nevertheless be interpreted as a heuristic device that facilitates economic modeling and, consequently, it may be justified along the same lines as Popper's RP. We then argue that, our position as to the resolution of this paradox notwithstanding, Popper's philosophy provides a metatheoretical framework with which we can evaluate the REH. Within this framework, the REH can be viewed as a heuristic device or strategy that fulfils the same function as, for instance, the optimizing assumption. However, we believe that the REH imparts a serious methodological bias, since, by implying that macroeconomic instability is caused exclusively by "exogenous" shocks that randomly hit the economy, it precludes the analysis of any sources of inherent instability caused by the making of (nonrandom) errors by individuals, and hence it favors the creation of an institutional configuration that may be ill suited to address this type of instability.