The natural selection of good science

Scientists in some fields are concerned that many, or even most, published results are false. A high rate of false positives might arise accidentally, from shoddy research practices. Or it might be the inevitable result of institutional incentives that reward publication irrespective of veracity. Recent models and discussion of scientific culture predict selection for false-positive publications, as research labs that publish more positive findings out-compete more diligent labs. There is widespread debate about how scientific practices should be modified to avoid this degeneration. Some analyses suggest that "bad science" will persist even when labs are incentivized to undertake replication studies, and penalized for publications that later fail to replicate. Here we develop a framework for modelling the cultural evolution of research practices that allows labs to expend effort on theory - enabling them, at a cost, to focus on hypotheses that are more likely to be true on theoretical grounds. Theory restores the evolution of high effort in laboratory practice, and it suppresses false-positive publications to a technical minimum, even in the absence of replication. In fact, the mere ability choose between two sets of hypotheses - one with greater chance of being correct than the other - promotes better science than can be achieved by having effortless access to the better set of hypotheses. Combining theory and replication can have a synergistic effect in promoting good scientific methodology and reducing the rate of false-positive publications. Based on our analysis we propose four simple rules to promote good science in the face of pressure to publish.