1. Understanding the factors and mechanisms driving the structure of ecological networks is a challenge for community ecologists. Notably, it remains unclear whether observed interaction patterns between two trophic groups are a result of (1) preferential partner selection between groups, or (2) species interactions within groups. 2. We conducted an experiment in order to disentangle these two drivers, using the plant-mycorrhizal symbiosis as a model system. We followed mycorrhizal colonization of plant roots growing either with or without neighbours. This allowed us to assess the relative importance of interaction within trophic groups (here, plant-to-plant interactions) on ecological network assembly. 3. Results showed that plants were not equally affected by the presence of neighbours and that network nestedness was higher when plants grew without neighbours. We also found a poor correlation between the centrality (i.e. standardized number of interactions) of plant species grown in communities and those grown without neighbours, while the reverse was true for mycorrhizal fungi. This suggests that the optimum level of specificity or generalism in mycorrhizal selection is not a fixed plant trait, but a plastic, context-dependent one. 4. Synthesis. Our results show that ecological networks are not only shaped by preferential partner selection, but also by interactions within a given trophic group. This finding should be considered in future modelling exercises on ecological network dynamics. Moreover, our nestedness results suggest that in the presence of multiple host plants, mycorrhizal fungi display preferences for specific interactions, suggesting that local plant diversity may shape mycorrhizal fungal community structure.