This study tests the assumption that processing limitations (in working memory capacity and numeracy) underlie biased decision-making. In these experiments, access to framing information during decision-making was manipulated. Having access to the information led to larger framing biases. Counterintuitively, higher working memory predicted more framing bias, except for in those with high numeracy, suggesting spontaneous conversion between frames for high numerates. In a second experiment, relationships between memory for the problem information and decision-making were analyzed. Crucial for the some-none comparison underlying framing effects, memory for the zero-complement was related to more framing. Memory for the endowment (total lives at risk), which is crucial for spontaneous conversion between frames, led to less bias. Results support fuzzy-trace theory's conception of framing effects, specifically that bias is linked to gist (i.e., meaningful representations of the problem), whereas reduced framing is linked to rote calculation (i.e., verbatim processing).