Sampling effort differences can lead to biased conclusions on the architecture of ant–plant interaction networks

• We empirically evaluated how variation in sampling affects ant–plant networks.
• All network descriptors used in this study were influenced by sampling effort.
• We discuss alternative points to reduce biased conclusions within ecological networks.

The study of plant–animal interactions using ecological networks as a tool has been intensified in recent years. However, one of the biggest criticisms of these studies is how sampling effort can affect the observed interaction patterns within these plant–animal networks. In this study we empirically evaluated how variation in sampling effort (comparing cumulative vs monthly sampled networks) influences the values of network descriptors more frequently used in the literature (i.e. number of links per species, network size, diversity of interactions, network specialization, robustness, nestedness, and modularity). For this, we studied interaction networks between ants and plants bearing extrafloral nectar sampled over 12 months in a tropical environment on the coast of Mexico. In general, all network descriptors used in this study were influenced by sampling effort via its effects on the record of new ant–plant interactions throughout the year. Interestingly, network specialization and the modularity tended to decrease with the increase of sampling effort, while all the other network descriptors tended to increase along with the sampling effort. Our study highlights the importance of standardizing data sampling in comparative studies to make sure that the results found in studies dealing with ecological networks are reliable, since the intensity of the sampling effort can directly affect the structure of plant–animal interaction networks.