Loss framing selectively reorganizes reward effects on creative problem solving

Whether reward effects on creative problem solving generalize across gain and loss framing remains unclear. Two experiments manipulated reward type (real vs. hypothetical) and reward magnitude (high vs. low) under loss framing in a compound remote associates (CRA) task and a Chinese character chunk-decomposition (CD) task, and the resulting patterns were compared with previously reported gain-frame datasets obtained with the same tasks and reward manipulations. For CRA, no interaction involving frame was detected for solution rate, although a reward type × frame interaction was observed for response time. In contrast, CD showed a significant reward type × magnitude × frame interaction, with a pronounced cross-sample difference in the reward pattern for novel, stroke-level decompositions that required departing from familiar representations; the estimate for ordinary, radical-level decompositions was close to zero. In the loss-frame sample, real-high rewards outperformed hypothetical-high and real-low rewards, whereas hypothetical-low rewards yielded the highest overall performance. Because the two frames were tested in independent samples, these differences cannot be attributed uniquely to framing. We interpret this task-specific pattern within a metacontrol framework, in which reward properties bias the balance between persistence and flexibility, while loss contexts may additionally alter the processing of novel representations. The results point to motivational framing as a possible boundary condition on reward effects in creative problem solving.