Ecological communities are defined by species interacting dynamically in a given location at a given time, and can be conveniently represented as networks of interactions. Pairwise interactions can be ascribed to one of five main types, depending on their outcome for the species involved: amensalism, antagonism (including predation, parasitism and disease), commensalism, competition or mutualism. While most studies have dealt so far with networks involving one single type of interaction at a time, often focusing on a specific clade and/or guild, recent studies are being developed that consider networks with more than one interaction type and across several levels of biological organisation. We review these developments and suggest that three main frameworks are in use to investigate the properties of multiple interactions networks: ‘expanded food‐webs’, ‘multilayer networks’ and ‘equal footing networks’. They differ on how interactions are classified and implemented in mathematical models, and on whether the effect of different interaction types is expressed in the same units. We analyse the mathematical and ecological assumptions of these three approaches, and identify some of the questions that can be addressed with each one of them. Since the overwhelming majority of studies on multiple interactions are theoretical and use artificially generated data, we also provide recommendations for the incorporation of field data in such studies.