1: \begin{abstract}

2: Modular organization characterizes many complex networks occurring in

3: nature, including the brain. In this paper we show that modular

4: structure may be responsible for increasing the robustness of certain

5: dynamical states of such systems. In a neural network model with

6: threshold-activated

7: binary elements, we observe that the basins of attractors,

8: corresponding to patterns

9: that have been embedded using a learning rule,

10: %learnt using Hebbian rule,

11: occupy maximum volume in phase space at an optimal 

12: %degree of

13: modularity. Simultaneously, the convergence time

14: to these attractors decreases as a result of cooperative dynamics

15: between the modules. The role of modularity in increasing global

16: stability of 

17: %the attractors 

18: certain desirable attractors of a system may provide a clue to its

19: evolution and ubiquity in natural systems.

20: \end{abstract}