Science involves both theory building and fact finding. This chapter focuses on the fact- finding aspect. In this sense, science can be viewed as a process of signal detection for facts. We wish to discover true associations between variables. However, our methods for measurement are imprecise. We sometimes mistake noise for signal, and vice versa. How we conceptualize the scientific enterprise shapes how we go about the business of conducting research as well as how we strive to improve scientific practices. In this chapter, I’ll present several models of science. I’ll begin by showing ways in which the classic “hypothesis testing” model of science is misleading, and leads to flawed inferences. As a remedy, I’ll discuss models that treat science as a population process, with important dynamics at the group level that trickle down to the individual practitioners. Science that is robust and reproducible depends on understanding these dynamics so that institutional programs for improvement can specifically target them.