Functional fixedness involves difficulty with conceptualizing creative object uses. When it obstructs problem-solving, individuals must reframe their approach. We examined how different training techniques--chunk decomposition (i.e., considering an object's basic parts and physical properties) and constraint relaxation (i.e., considering an object's different functions)--might rely upon different routes to creative reframing. Additionally, we investigated how different forms of cognitive load interact with these dual routes. Participants learned one of three techniques. Chunk decomposition participants created object breakdown diagrams; constraint relaxation participants created object functions lists; and free association (control) participants wrote a word that they associated with each of several concrete nouns. After training, participants attempted to solve five functional fixedness problems. E1 investigated how increasing germane cognitive load via either direct or indirect prompting affected training transfer. Experiment 2 investigated how reducing extraneous cognitive load by providing no transfer instructions and using an eye-closure strategy. Across both experiments, results supported differences in accuracy and response latency by training. However, chunk decomposition and constraint relaxation did not follow the same pattern, suggesting different mechanisms of the effect. We discuss possible applications to increase innovation in real-world domains such as education, business, and engineering.