Questions: Are liana–host interactions structured at the community level? Do liana–host interactions differ between species growth form guilds or habitats?\r\n\r\nLocation: Otari-Wilton\\\s Bush on the southern tip of North Island New Zealand. The forest contains 75 ha of mature and regenerating conifer–broadleaf forest.\r\n\r\nMethods: Nine liana species were quantified among 217 trees to test for negative co-occurrence patterns. We also conducted additional analyses within and among compartments embedded in the community-level matrix. Liana and host abundance distributions were assessed across two contrasting habitats.\r\n\r\nResults: Community-level analyses revealed negative co-occurrence patterns. Positive neutral and negative co-occurrence patterns were found among compartments within the community-level matrix. Host species compartments were consistent with randomized expectations while positive co-occurrence patterns were found within the host species matrix. Negative co-occurrence patterns were found inconsistently among lianas that share the same region of host space and those that do not.\r\n\r\nConclusions: Overall results indicate the liana community is structured non-randomly. Liana–host interactions appear to follow an opportunistic growth strategy and interactions are due mostly to habitat partitioning.