By integrating four lines of thoughts: symmetry breaking originally advanced by Anderson, bifurcation from nonlinear dynamical systems, Landau's phenomenological theory of phase transition, and the mechanism of emergent rare events first studied by Kramers, we introduce a possible framework for understanding mesoscopic dynamics that links (i) fast microscopic (lower level) motions, (ii) movements within each basin-of-attraction at the mid-level, and (iii) higher-level rare transitions between neighboring basins, which have slow rates that decrease exponentially with the size of the system. In this mesoscopic framework, the fast dynamics is represented by a rapidly varying stochastic process and the mid-level by a nonlinear dynamics. Multiple attractors arise as emergent properties of the nonlinear systems. The interplay between the stochastic element and nonlinearity, the essence of Kramers’ theory, leads to successive jump-like transitions among different basins. We argue each transition is a dynamic symmetry breaking, with the potential of exhibiting Thom-Zeeman catastrophe as well as phase transition with the breakdown of ergodicity (e.g., cell differentiation). The slow-time dynamics of the nonlinear mesoscopic system is not deterministic, rather it is a discrete stochastic jump process. The existence of these discrete states and the Markov transitions among them are both emergent phenomena. This emergent stochastic jump dynamics then serves as the stochastic element for the nonlinear dynamics of a higher level aggregates on an even larger spatial and slower time scales (e.g., evolution). This description captures the hierarchical structure outlined by Anderson and illustrates two distinct types of limit of a mesoscopic dynamics: A long-time ensemble thermodynamics in terms of time t→∞ followed by the size of the system N→∞, and a short-time trajectory steady state with N→∞ followed by t→∞. With these limits, symmetry breaking and cusp catastrophe are two perspectives of the same mesoscopic system on different time scales.