The paper introduces a new centrality measure called subgraph centrality ($SC$) that characterizes the participation of each node in all subgraphs within a network. Unlike traditional measures such as degree, closeness, betweenness, and eigenvector centralities, $SC$ gives more weight to smaller subgraphs, making it suitable for identifying network motifs. The $SC$ can be mathematically derived from the spectra of the network's adjacency matrix. The authors demonstrate that $SC$ outperforms other centrality measures in discriminating nodes, particularly in real-world networks. They study eight real-world networks, including protein interaction networks, vocabulary networks, and citation networks, and find that $SC$ displays useful properties such as clear node ranking and scale-free characteristics. In the protein interaction network of *Saccharomyces cerevisiae*, $SC$ is more correlated with protein lethality than the number of links per node. The paper also explores the scaling properties of $SC$ and concludes that it shows power-law distributions even in cases where degree centrality does not. Overall, the subgraph centrality measure provides a novel and effective way to analyze and rank nodes in complex networks.The paper introduces a new centrality measure called subgraph centrality ($SC$) that characterizes the participation of each node in all subgraphs within a network. Unlike traditional measures such as degree, closeness, betweenness, and eigenvector centralities, $SC$ gives more weight to smaller subgraphs, making it suitable for identifying network motifs. The $SC$ can be mathematically derived from the spectra of the network's adjacency matrix. The authors demonstrate that $SC$ outperforms other centrality measures in discriminating nodes, particularly in real-world networks. They study eight real-world networks, including protein interaction networks, vocabulary networks, and citation networks, and find that $SC$ displays useful properties such as clear node ranking and scale-free characteristics. In the protein interaction network of *Saccharomyces cerevisiae*, $SC$ is more correlated with protein lethality than the number of links per node. The paper also explores the scaling properties of $SC$ and concludes that it shows power-law distributions even in cases where degree centrality does not. Overall, the subgraph centrality measure provides a novel and effective way to analyze and rank nodes in complex networks.