Hodgkin and Huxley (H-H) model for action potential generation has held firm for half a century because this relatively simple and experimentally testable model embodies the major features of membrane nonlinearity: namely, voltage-dependent ionic currents that activate and inactivate in time. However, experimental and theoretical developments of the past 20 years force one to re-evaluate its usefulness. First, the H-H model is, in its original form, limited to the two voltage-dependent currents found in the squid giant axon and it must be extended significantly if it is to deal with the excitable soma and dendrites of neurons. Second, the macroscopic and deterministic H-H model does not capture correctly the kinetics of the Na(+) channel and it cannot account for the stochastic response to current injection that arises from the discrete nature of ion channels. Third, much simpler integrate-and-fire-type models seem to be more useful for exploring collective phenomena in neuronal networks. Is the H-H model threatened, or will it continue to set the fundamental framework for exploring neuronal excitability?