Understanding the drivers of food chain length in natural communities has intrigued ecologists since the publication of ‘food cycles’ by Elton in the early 20th century. Proposed drivers of food chain length have included extrinsic controls such as productivity, disturbance regime, and ecosystem size, as well as intrinsic factors including food web motifs. However, current theories have largely assumed simple, two-dimensional habitat architectures, and may not be adequate to predict food chain length in ecosystems which have a complex, branching structure. Here, we develop a spatially explicit theoretical model which provides an integrated framework for predicting food chain length in branching networks. We show food chain length responds independently to both ecosystem size and complexity, and that these responses are contingent upon other extrinsic and intrinsic controls. Our results show that accounting for ecosystem complexity is an important driver of food chain length and may reconcile inconsistent results from empirical studies of food chain length in river ecosystems.