Synchronized network level activity is the hallmark of developing neuronal networks. These synchronized patterns are modulated by a combination of intrinsic properties of single neurons and interactions between these neurons as manifested by the network morphology. Consequently, mapping the network bursts (NBs) as a function of the network topology may shed light on the interplay between morphlogical and functional aspects of the network activtiy. To address this issue experimentally, we used poly-d-lysine (PDL) and carbon nanotube (CNT) based network engineering to induce self-organization of cultured neurons into clustered networks of different sizes and topologies. These clusters were coupled to micro electrode arrays (MEAs) for long term electrical activity recordings. By varying the size of the clusters from few to hundreds of neurons, we found that clusters of a few tens of neurons already exhibit NBs and that the duration and rate of NBs increases with the cluster size. Furthermore, the NBs are characterized by innate coherent network oscillations in the range of 25 to 100 Hz. We show that two clusters coupled by a bundle of extensions, exhibit both individual and mutual NBs in which mutual NBs are characterized by long activation delays (tens of ms) between the two clusters. In addition, while small clusters and uniform networks exhibit fast and spatially distributed recruitment during NBs, networks of many connected clusters are characterized by slow sequential activation of the clusters in the network.