The study explores how a complexity approach empowers science learning. A complexity approach represents systems as many interacting entities. The construct of micro-macro compatibility is introduced, the degree of similarity between behaviors at the micro- and macro-levels of the system. Seventh-grade students' learning about gases was studied using questionnaires and interviews. An experimental group (n = 47) learned with a complexity curriculum that included agent-based computer models, a workbook, class discussions, and laboratory experiments. A comparison group (n = 45) learned with a normative curriculum, incorporating lectures, a textbook, class discussions, and laboratory experiments. Significant learning gains and strong effect sizes were found in the experimental group's overall learning. Diffusion, density, and kinetic molecular theory were learned better with a complexity approach. Pressure, temperature, and the gas laws were learned similarly with both approaches. Learning to notice micro-level behaviors and their probabilistic nature was greater with the complexity approach. Analysis showed that only concepts that have less "micro-macro compatibility" were learned better with a complexity approach. Thus, a complexity approach helps separate the microbehaviors and then relate them to the macrobehaviors when these behaviors are dissimilar. We discuss how micro-macro compatibility helps point to concepts whose learning would benefit strongly from a complexity approach.