It is a hallmark of social and psychological science that its findings should be reproducible across relevant contexts, playing a key role in showcasing the reliability of detected effects. However, following what some have called a “watershed” moment, a “crisis of confidence” has befallen the discipline, with commentators zeroing in on the Reproducibility Project’s low replication estimate of 36-47% (Open Science Collaboration, 2015; Pashler & Wagenmakers, 2012; Wiggins & Christopherson, 2019). The theory crisis is another perhaps less well-known issue facing social and psychological science, with researchers arguing that its theories are generally of poor quality (Eronen & Bringmann, 2021; Fiedler, 2017; Oberauer & Lewandowsky, 2019). Due to this, theories are often not clearly accepted or refuted, tending to come and go with little of the cumulative progress often associated with scientific knowledge. In this presentation, I aim to look at how the replication and theory crises interact in social and psychological science with particular emphasis on the relationship between replication and theory development. Following an overview of both crises, the first half of the presentation will look at the way in which replications can support theory development, covering the identification of robust phenomena, the identification of  effects’ boundary conditions, and the evaluation of theories’ predictions. The second half of the presentation will cover a number of more recent arguments suggesting that well-specified theory is required for replications to be informative (Irvine, 2021; Klein, 2014; Muthukrishna & Henrich, 2019). I will argue against a strong reading of this position.