Biological neural networks are characterized by short average path lengths, high clustering, and modular and hierarchical architectures. These complex network topologies strike a balance between local specialization and global synchronization via long-range connections, resulting in highly efficient communication. Here, we use a geometric network model with either an intermediate or a long-range connection probability to investigate the effects of wiring cost principles on network complexity for different spatial conformations. We find that both long-range and intermediate wiring probabilities only conform to small-world architectures for neurons in dense spatial clusters due to a decrease in wiring cost within clusters. Furthermore, both small-worldness and modularity were reduced in systems with long-range connections caused by a reduction in network clustering, allowing for novel insight into mechanisms underlying adaptive or maladaptive network alterations. Our findings corroborate previous work showing that both wiring probability and spatial distributions play a key role in neural network development.