Phylogenetic networks and admixture graphs allow for the representation and modelling of gene flow, introgression and hybridization. Under the assumption that the network is of level-1, most of its topology and edge parameters are known to be identifiable, from various data types and under various models. But little is known about what network features are identifiable if the level-1 constraint is removed. I will show some features that are identifiable, and give examples of networks and substructures that cannot be distinguished, under models with and without incomplete lineage sorting.