The brain must solve a wide range of different temporal problems, each of which can be defined by a relevant time scale and specific functional requirements. Experimental and theoretical studies suggest that some forms of timing reflect general and inherent properties of local neural networks. Like the ripples on a pond, neural networks represent rich dynamical systems that can produce time-varying patterns of activity in response to a stimulus. State-dependent network models propose that sensory timing arises from the interaction between incoming stimuli and the internal dynamics of recurrent neural circuits. A wide-variety of time-dependent neural properties, such as short-term synaptic plasticity, are important contributors to the internal dynamics of neural circuits. In contrast to sensory timing, motor timing requires that network actively generate appropriately timed spikes even in the absence of sensory stimuli. Population clock models propose that motor timing arises from internal dynamics of recurrent network capable of self-perpetuating activity.