onsiderable debate exists regarding the mechanisms that underlie temporally dynamic responses in early processing of olfactory sensory information. Several studies have now shown that within a few hundred milliseconds after onset of stimulation, the response evolves through a sequence of activation patterns (the transient phase) to an odor-specific steady state (attractor). We present a conductance-based computational model that suggests mechanisms for generating both transient and attractor-based dynamic responses. Dynamical systems methods are used to rigorously reduce the biophysical model to a computationally more tractable discrete system. Analysis of the discrete model demonstrates that there is a phase transition separating regions where the network exhibits long transients, dynamically changing patterns of synchronous activity and decorrelation of inputs from regions where these properties do not hold. The analysis demonstrates how the phase transition depends on network properties such as the connectivity and heterogeneities in the intrinsic and synaptic properties of cells. We argue that these discrete networks have implications for understanding biological neuronal networks, particularly because the instability at the phase transition increases the discriminability of sensory patterns.