Research across numerous domains has highlighted the current--and presumably temporary--effects of frames on preference and behavior. Yet people often encounter information that has been framed in different ways across contexts, and there are reasons to predict that certain frames, once encountered, might tend to stick in the mind and resist subsequent reframing. We propose that loss frames are stickier than gain frames in their ability to shape people's thinking. Specifically, we suggest that the effect of a loss frame may linger longer than that of a gain frame in the face of reframing and that this asymmetry may arise because it is more difficult to convert a loss-framed concept into a gain-framed concept than vice versa. Supporting this notion, loss-to-gain (vs. gain-to-loss) reframing had a muted impact on both risk preferences (Study 1) and evaluation (Study 2). Moreover, participants took longer to solve a math problem that required reconceptualizing losses as gains than vice versa (Studies 3-5), and reframing changed gain-based conceptualizations but not loss-based ones (Study 6). We discuss implications for understanding a key process underlying negativity bias, as well as how sequential frames might impact political behavior and economic recovery.