This paper presents an architecture for evolving coadapted subcomponents through cooperative coevolution. The key idea is that evolutionary algorithms (EAs) can be extended to allow the emergence of coadapted subcomponents rather than being hand-designed. The architecture models an ecosystem of cooperating species, where each species evolves independently but interacts with others to solve a problem. The paper explores this approach through a simple string-matching task and a more complex case study involving the evolution of artificial neural networks. The paper discusses several challenges in evolving coadapted subcomponents, including problem decomposition, interdependencies between subcomponents, credit assignment, and maintaining diversity. It also reviews previous approaches to evolving coadapted subcomponents, such as classifier systems, REGAL, and models of hosts and parasites. These approaches have shown that coevolution can lead to emergent problem decomposition and better solutions than traditional EAs. The paper then describes the cooperative coevolution architecture, which models an ecosystem of cooperating species. Each species evolves in its own population and interacts with others to solve a problem. The architecture allows for emergent problem decomposition, where the number and roles of species are determined by the evolutionary process rather than being hand-designed. The paper demonstrates this through a case study in string covering, where the algorithm successfully locates and covers multiple environmental niches, evolves to an appropriate level of generality, and adapts as the number and roles of species change. The paper also discusses the benefits of cooperative coevolution, including the ability to maintain diversity, the emergence of appropriate subcomponents, and the ability to adapt to changing environments. It concludes that cooperative coevolution is a promising approach for evolving coadapted subcomponents and that further research is needed to explore its potential in more complex problems.This paper presents an architecture for evolving coadapted subcomponents through cooperative coevolution. The key idea is that evolutionary algorithms (EAs) can be extended to allow the emergence of coadapted subcomponents rather than being hand-designed. The architecture models an ecosystem of cooperating species, where each species evolves independently but interacts with others to solve a problem. The paper explores this approach through a simple string-matching task and a more complex case study involving the evolution of artificial neural networks. The paper discusses several challenges in evolving coadapted subcomponents, including problem decomposition, interdependencies between subcomponents, credit assignment, and maintaining diversity. It also reviews previous approaches to evolving coadapted subcomponents, such as classifier systems, REGAL, and models of hosts and parasites. These approaches have shown that coevolution can lead to emergent problem decomposition and better solutions than traditional EAs. The paper then describes the cooperative coevolution architecture, which models an ecosystem of cooperating species. Each species evolves in its own population and interacts with others to solve a problem. The architecture allows for emergent problem decomposition, where the number and roles of species are determined by the evolutionary process rather than being hand-designed. The paper demonstrates this through a case study in string covering, where the algorithm successfully locates and covers multiple environmental niches, evolves to an appropriate level of generality, and adapts as the number and roles of species change. The paper also discusses the benefits of cooperative coevolution, including the ability to maintain diversity, the emergence of appropriate subcomponents, and the ability to adapt to changing environments. It concludes that cooperative coevolution is a promising approach for evolving coadapted subcomponents and that further research is needed to explore its potential in more complex problems.