n audience expresses appreciation for a good performance by the strength and nature of its applause. The thunder of applause at the start often turns quite suddenly into synchronized clapping, and this synchronization can disappear and reappear several times during the applause. The phenomenon is a delightful expression of social self-organization that provides an example on a human scale of the synchronization processes that occur in numerous natural systems, ranging from flashing Asian fireflies to oscillating chemical reactions 1–3. Here we explain the dynamics of this rhythmic applause. We investigated the mechanism and development of synchronized clapping by making a series of measurements focusing on the collective aspects of the self-organization process as well as on the behaviour of the individuals in the audience. We recorded several theatre and opera performances in Romania and Hungary by using a microphone placed on the ceiling of the hall (Fig. 1a). Typically, after a few seconds of incoherent random clapping, a periodic signal developed (a sign of synchronized clapping), visible in Fig. 1a as pronounced spikes. The transition is also captured by the order parameter (Fig. 1c), which increases as the periodic signal develops, and decreases as it disappears. Although synchronization increases the strength of the signal at the moment of the clapping, it leads to a decrease in the average noise intensity in the room (Fig. 1d). This is surprising, as the driving force for synchronization would be expected to reflect the desire of the audience to express its enthusiasm by increasing the average noise intensity. The origin of this conflict between the average noise and synchronization can be understood by correlating the global signal with the behaviour of an individual in the audience. We demonstrated this by recording the local sound intensity in the vicinity of a group of oblivious individuals (Fig. 1b). In the incoherent phase, the local signal was periodic, with a short period corresponding to the fast clapping of an individual in the audience. However, the clapping period suddenly doubled at the start of the synchronized phase (at about 12 seconds in Fig. 1a,b), and slowly decreased as synchronization was lost (Fig. 1e). The decrease in the average noise intensity is therefore a consequence of the period doubling, because there is less clapping per unit time. An increase in the average noise intensity is possible only by decreasing the clapping period, which can indeed occur (Fig. 1e). However, …