This group randomized control trial examined the dose-response effect of varied combinations of linear and nonlinear pedagogy (enriched physical education with specific program led by specialist vs. conventional physical education led by generalist) for improving first-grade children’s motor creativity, executive functions, self-efficacy, and learning enjoyment. We led three physical education classes per group through 12 weeks of combined instruction, based on linear and nonlinear pedagogy: mostly linear (ML; 80% linear, 20% nonlinear; n = 62); mostly nonlinear (MNL; 20% linear, 80% nonlinear; n = 61); and control (C; conventional teaching from generalists; n = 60). MNL improved in (a) motor creativity ability (DMA; 48.7%, 76.5%, and 47.6% for locomotor, stability, and manipulative tasks, respectively); (b) executive functions (working memory and inhibitory control) for RNG task (14.7%) and task errors (70.8%); (c) self-efficacy (5.9%); and (d) enjoyment (8.3%). In ML, DMA improved by 18.0% in locomotor and 60.9% in manipulative tasks. C improved of 10.5% in enjoyment, and RNG task worsened by 22.6%. MNL improvements in DMA tasks, executive functions, and self-efficacy were significantly better than those in C. ML was better than C in DMA task and in executive functions’ task errors. Overall, ML and MNL approaches were more effective than conventional generalist teaching (C), and the MNL combination of 80% nonlinear and 20% linear pedagogy was optimal. We recommend that educators favor the MNL approach.