When describing plant-animal interaction networks, sampling can be performed using plant- or animal-centred approaches. Despite known effects of sampling on network structure, how samplings affect the estimates of interaction beta diversity across networks is still unresolved. We investigated how sampling method affects the assessment of beta diversity of interactions, turnover and rewiring. We contrasted plant- and animal-centred sampling methods applied to pollination networks across habitats in a heterogeneous tropical landscape, the Pantanal Wetlands. We also asked whether plant traits influence the difference in interaction specialization according to sampling.  Plant-centred networks resulted in higher beta diversity of interactions in space than animal-centred networks. Turnover explained most of the beta diversity in both methods, but rewiring is proportionately more important when using the animal-centred method. While the plant-centred method indicated lower network modularity and specialization, floral traits modulated the effects of sampling method on species-level network metrics. Combining animal- and plant-centred approaches returned intermediate values for beta diversity of interactions and network metrics. Distinct methods may also be better suited for answering questions at different scales. Our results point out that the method choice, or combination of methods, should always reflect the appropriate scale of the factors determining the interactions being investigated.