Rich gets simpler.

Network science plays a central role in the study of complex systems, offering a range of computational tools and, importantly, a common language to represent systems as diverse as the World Wide Web, the human brain, and social networks (1). Within the framework of network science, a system is modeled as a set of nodes, representing the individual units of the system, and a set of links, representing the dyadic relationships between these units. Many networks have been shown to exhibit a complex organization and yet can often be comprehended by simple and universal mechanisms. However, to properly capture the complexity of real-world interacting systems, standard network models are sometimes not sufficient. For this reason, different attempts have been made to enrich the network language in recent years. Important examples include multiplex or multilayer networks (2), where different types of interactions are accounted for, higher-order networks (3–5), focusing on pathways instead of dyadic interactions, and temporal networks (6, 7), where nodes and links become dynamical entities. In PNAS, Sekara et al. (8) make two important contributions to the latter approach, first by studying a longitudinal, high-resolution dataset on human interactions over an extended time window and second by showing that significant structural patterns naturally emerge from the system when considered at an appropriate time scale (Fig. 1). The study of temporal networks starts from the observation that real-world interacting systems exhibit nontrivial temporal patterns. Nodes and links are often dynamical entities that may emerge and disappear in the course of time. Take a social network, for instance, and the timings at which two individuals communicate via emails or phone calls. The resulting time series exhibit patterns at different scales, from minute-byminute variations to yearly reorganizations. In particular, their rate of interaction is time-dependent due to obvious circadian and weekly rhythms (9), and their interevent times are broadly distributed because of self-exciting, multiplicative processes (10), possibly leading to a cascade of responses. In general, temporal networks are much more complex objects than static ones, and their study requires new computational and mathematical tools, combining structural and dynamical complexity. Temporality alters even the most basic concepts of network science, such as the path between two nodes (11). Nonetheless, as Sekara et al. (8) show, temporality may also help in uncovering information in networks when properly incorporated. Interest in temporal networks has been fueled by the increasing availability of time-stamped interaction data in social systems. Important examples include communication patterns on Twitter (12) or mobile phones (13). Although they are routinely collected by network providers and online services, these data suffer from some important limitations due to their proprietary nature: Strict confidentiality agreements ab