Relational approaches in economic geography have grown in popularity and influence, but have not been critically evaluated or discussed. This article argues that poststructural and network‐based versions of relational economic geography undoubtedly open up new research issues and provide tools for certain purposes, but questions whether they provide a coherent research agenda and new theoretical paradigm that can guide the reconceptualization of economic geography. The article addresses two main cases for a relational approach: its correspondence with a knowledge‐ and network‐based capitalism and the claim that it provides an improved philosophy and ontology. It finds problems with both cases and argues that the approach provides an imprecise and selective ontology that is preoccupied with microscale processes. As a result, relational economic geography does not lend itself to causal explanations and schemas, is incapable of discriminating among different economic theories, and could become immune to empirical evaluation. In many cases, it seems to disregard many of the valuable insights of institutionalist and critical realist approaches, including the implications of emergence. The article concludes that instead of searching for a microlevel relational perspective or a new vocabulary, economic geography's analysis of connections and relations would be better set within an evolutionary and historical institutionalism that understands economic relations as forms of institutional rules and practices and does not privilege ties and networks over nodes and agents.