Despite its moral dubiety and negative economic impact, favoritism is ubiquitous. This project investigates distorted memories as driver of favoritism. In an online experiment, managers make incentivized personnel decisions. To causally identify the effect of memory, I vary whether choices are made based on recalled or actual quality of workers. To establish a baseline level of favoritism, I experimentally manipulate preferences for two groups of workers. Managers remember favored workers as being of higher quality than unfavored workers. These distorted memories account for a substantial fraction of the increase in favoritism due to reliance on memory. The remaining gap is due to a shift in weight towards group information.