The concept of brain modularity stipulates that groups of neurons or brain regions perform functions with some amount of independence. The more modular you think the brain is, the more independently you think brain regions function.
Determining non-arbitrary modules on the basis of connectivity is very difficult (i.e., it is NP-complete). Sohn et al. have taken a creative approach to it on the C. elegans connectome, using the data set here.
As in most clustering analyses, their algorithm attempts to divide the neurons into the groups that maximize intra-cluster (synapse) weights and minimize inter-cluster weights. But they add the constraint that bilateral neuron pairs must be in the same structural cluster, which makes biological sense, to help improve the optimization.
Once the authors define their structural clusters, they can speculate on how it makes sense in terms of neural info flow, which they show with a chart:
To simplify based on synaptic ratios, one might say that the flow of information follows the path of cluster 11 (mostly sensory) -> 13 -> 12 -> 21 -> 22 (mostly motor). In the discussion, the authors suggest a few ways in which these predictions could be tested.
Sohn Y, Choi M-K, Ahn Y-Y, Lee J, Jeong J (2011) Topological Cluster Analysis Reveals the Systemic Organization of the Caenorhabditis elegans Connectome. PLoS Comput Biol 7(5): e1001139. doi:10.1371/journal.pcbi.1001139