Many scientists are convinced that the Modern Synthesis – the current paradigm in evolutionary biology – needs to be expanded within the framework of an Extended Evolutionary Synthesis (EES). A major task for the EES is to provide an explanation for the origins and extant features of biological complexity. Here we address this issue, focusing our investigation on genetic network architectures, motifs, and dynamical behaviour, by developing an intuitive and essen-tially parameter-free evolutionary model of transcriptional regulation where the self-replicating digital organisms are Boo-lean networks, and where fitness is determined by their information-processing capacities. We validate our choice of fit-ness function by demonstrating that our evolved networks exhibit typical biological features of extant genetic regulatory networks: sparse connectivity, scale-free out-degree (within our range of measurement) and exponentially decaying in-degree distributions, significant clustering, a high proportion of feed-forward loop (FFL) network motifs, a prominence of canalising logic at the promoter, plasticity, and distributed robustness to mutation. In addition, the dynamics of our evolved networks feature simple attractor cycles that are robust to perturbations and exhibit self-organised criticality. In networks evolved without gene duplications, we show that the key architectural signatures noted above are absent. Sur-prisingly, the canalising fraction is much higher in comparison with networks evolved with gene duplications. These re-sults suggest network properties of extant gene networks require gene duplications in order to evolve, and that these prop-erties undergo positive selection, where they contribute to the global stability of the networks. By demonstrating that net-works evolved without gene duplications are robust and, like their scale-free counterparts, also exhibit self-organised criticality, this work highlights the interplay between contingent mechanism, such as gene duplications, and selection, in determining evolutionary outcomes.