Although agonists and competitive antagonists presumably occupy overlapping binding sites on ligand-gated channels, these interactions cannot be identical because agonists cause channel opening whereas antagonists do not. One explanation is that only agonist binding performs enough work on the receptor to cause the conformational changes that lead to gating. This idea is supported by agonist binding rates at GABA(A) and nicotinic acetylcholine receptors that are slower than expected for a diffusion-limited process, suggesting that agonist binding involves an energy-requiring event. This hypothesis predicts that competitive antagonist binding should require less activation energy than agonist binding. To test this idea, we developed a novel deconvolution-based method to compare binding and unbinding kinetics of GABA(A) receptor agonists and antagonists in outside-out patches from rat hippocampal neurons. Agonist and antagonist unbinding rates were steeply correlated with affinity. Unlike the agonists, three of the four antagonists tested had binding rates that were fast, independent of affinity, and could be accounted for by diffusion- and dehydration-limited processes. In contrast, agonist binding involved additional energy-requiring steps, consistent with the idea that channel gating is initiated by agonist-triggered movements within the ligand binding site. Antagonist binding does not appear to produce such movements, and may in fact prevent them.