In the neostriatum, GABAergic inhibition arises from the action of at least two classes of inhibitory interneurons, and from recurrent collaterals of the principal cells. Interneurons receive excitatory input only from extrinsic sources, and so act in a purely feedforward capacity. Feedback inhibition arises from the recurrent collaterals of the principal cells. These two kinds of inhibition have functionally distinct effects on the principal cells. Inputs from interneurons are not very convergent. There are few inhibitory neurons, and so each principal cell receives inhibitory synaptic input from very few interneurons. But, they are individually powerful, and a single interneuron can substantially delay action potentials in a group of nearby principal cells. Recurrent inhibition is highly convergent, with each principal cell receiving inhibitory input from several hundred other such cells. Feedback inhibitory synaptic inputs individually have very weak effects, as seen from the soma. The differences in synaptic strength are not caused by differences in the release of transmitter or in sensitivity of the postsynaptic membrane. Rather, they arise from differences in the number of synaptic contacts formed on individual principal cells by feedforward or feedback axons, and from differences in synaptic location. Interneurons form their powerful synapses near the somata of principal cells, while most feedback synapses are more distal, where they interact with the two-state nonlinear properties of the principal cells' dendrite. This arrangement suggests that feedforward inhibition may serve in the traditional role for inhibition, adjusting the excitability of the principle neuron near the site of action potential generation. Feedback inhibitory synapses may interact with voltage-sensitive conductances in the dendrite to alter the electrotonic structure of the spiny cell.