Aim: Nestedness is a common pattern in metacommunities and interaction networks, whose causes are still discussed. Nestedness inference is challenging because, beyond calculating an index, we need to compare observed values with values generated with a null model. There are different null models and the choice between them affects test outcomes. Furthermore, there is no established theoretical basis to guide this choice. Here, we propose a different look at the meaning of nestedness that improves our understanding of its causes and unveils the link between null models and hypotheses.
Innovation: Nestedness of a matrix is a combination of marginal sum inequality and high overlap. The higher the overlap, the more predictable the cell values by marginal sums. Here, we show that nestedness actually measures how better one can predict cell values by marginal sums than by matrix dimensions and total sum alone. From this, we propose that two null models can be used to test for different topological hypotheses. The equiprobable model excludes all nestedness-generating mechanisms and provides the distribution of expected values for nestedness significance tests. The proportional model conserves nestedness-generating mechanisms and excludes nestedness-disrupting mechanisms, and thus, produces highly nested matrices. The proportional model provides the distribution of expected nestedness for nested matrices. Additionally, we evaluate the efficiency of several indices within this new perspective and illustrate our approach using an empirical plant-pollinator network.
Main conclusions: Through a shift of perspective, our approach reconciliates contradictions in null model analysis and delimits the range of possible explanations for nestedness. The only way a process can increase nestedness in a matrix is by promoting marginal sum inequalities, without concomitantly introducing preferences. Consequently, in a species interaction network, explanations for nestedness should explain why some species interact more frequently than others.