The fast bursting rate is comparable to that of large homogeneous networks. Despite this apparent trend, there is also significant variability in the network bursting rate between different clusters of the same size. Additionally, in some of the clusters, the IBI distribution was characterized by more than one peak due to the fact that many NBs were grouped into bursts of NBs with short intervals between them. In Figure 3b we show the NB width distribution for a Nutlin-3 molecular weight typical cluster. This distribution has a narrow typical time scale, as shown in Figure 3b, with high variability in the mean NB width between different clusters. The effect of the size of the clusters on the network bursting rate and event width is shown in Figures 3c�Cf. Figures 3c,d show results from all analyzed clusters while figures 3e,f show the average rate and width, respectively, calculated in consecutive logarithmic time windows. The most intriguing feature is the onset of network activity at clusters with as few as several tens of cells. Apparently, clusters larger than 5000 mm2, corresponding to about 40 cells, already exhibit synchronized network activity. We AB1010 purchase therefore approximate that the upper limit for the onset of synchrony is about 40 cells. Based on our data we can estimate that a transition occurs within a cluster area range of 2500�C5000 mm2, corresponding to about 20�C40 cells. None of the smaller clusters in our experiments exhibited synchronized network activity. This result suggests the existence of a minimal network size which is required to generate and sustain collective activity. It is important to note that such small clusters do exhibit tonic single spike activity. However, the single neuron firing is not sufficient to generate collective network bursts. There are also silent small clusters that did not exhibit any electrical activity. Those were eliminated from the analysis. As was mentioned above, the NB rate and the NB width appear to increase with cluster size. This increase converged to the NB rate and width of large uniform networks of 106 cells. We now inspect the internal temporal features of the network events. First, similarly to large homogeneous networks, most of the network events recorded in isolated clusters have a stereotypical temporal profile with a fast rise in the activity intensity, followed by a slower activity decay, as is shown in Figure 2a. This activity profile reflects the fact that many neurons are rapidly activated at the onset of NBs and are gradually relaxed or inhibited with time leading to the NB intensity decay. The overall similarity between consecutive NBs described above reflects a much more significant correspondence between them.