The activity of each cluster was recorded by an underlying microelectrode allowing recordings from as many as 60 isolated LY2835219 islands in parallel. To determine the cluster size and cellular arrangement within each cluster, neurons, glia and cell nuclei were specifically labeled and imaged. We found that each cluster was made of both neurons and glia cells spatially arranged as glia carpets with overlying neurons. Moreover, neurons in all clusters exhibited extensive neurite growth and synapse formation. The area of each cluster was used as a measure of its size and the number of cells in each cluster was estimated from the cluster size. After two weeks in vitro, neuronal clusters exhibited noticeable spontaneous activity. We note that since each cluster was recorded by a single electrode, the recorded activity represents the sum of firing of many neurons. Consequently, we defined a population-level cluster activity intensity measure which is estimated from the recorded voltage waveform. The recorded spontaneous activity of isolated clusters was marked by synchronized bursting events or network bursts. These network events are similar to the network GDC-0199 bursts observed in large homogeneous networks composed of hundreds of thousands of cells. More specifically, they are characterized by short time windows of intense neuron firing followed by longer intervals of sporadic firing. Additional similarity is in the intra burst patterns and in the burst statistics. These topics are presented in the next sections. The activity of all the clusters in each neuro-chip was recorded and analyzed. During the self-organization process, some of the clusters became linked by bundles of axons. To distinguish between isolated and linked clusters, we calculated the Pearson correlation between the activities of all cluster pairs. As can be seen in Figure 1c, the activity of some clusters is correlated, but there are some pairs of clusters with vanishingly small correlations. The latter are the isolated clusters whose activity we analyzed while connected clusters, which had significant correlation with at least one other cluster, were eliminated from the analysis. The identification of linked clusters was also validated by visual inspection. Finally, the temporal loci, as well as the width for each NB were detected and the inter burst intervals were calculated. It was previously shown that inter-burst-intervals of NBs are characterized by a long tail multi time-scale distribution. Interestingly, the IBI distribution of our small clusters revealed similar behavior. The general trend is that smaller networks exhibit network bursting with lower rates, while larger clusters exhibit faster network bursting.