Rewiring

Rewiring is the operation of changing the endpoints of edges in a network while preserving some set of structural properties — most commonly the degree sequence. A rewired network has the same number of connections at each node as the original, but the specific connections are randomized. This makes rewiring the standard technique for constructing null models: by comparing a real network to its rewired counterpart, researchers can determine which properties are consequences of the degree sequence and which require additional explanation.

The simplest rewiring algorithm, due to Maslov and Sneppen, repeatedly selects two edges at random and swaps their endpoints. After sufficient iterations, the network becomes a random graph conditioned on the original degree sequence — mathematically equivalent to the configuration model but without multi-edges or self-loops.

Rewiring has also been used as a model of network evolution. Preferential attachment rewiring, adaptive rewiring in neural networks, and strategic rewiring in social networks all treat edge rearrangement as a dynamic process rather than a statistical tool. In these contexts, rewiring is not a null model but a mechanism: the network changes its own topology in response to functional pressures.

Rewiring is the closest thing network science has to a controlled experiment. It is also deeply unsatisfying: you can preserve the degree sequence, or the clustering coefficient, or the community structure, but you cannot preserve everything. Every rewiring scheme smuggles in assumptions about what matters and what does not. The null model is never truly null.