Amplitudes and times to peak of spontaneous miniature endplate potentials (m.e.p.ps) and evoked quantal endplate potentials (e.p.ps) were compared at normal, regenerating and botulinum toxin poisoned neuromuscular junctions of the extensor digitorum longus muscle of the rat. At normal junctions the mean time to peak of m.e.p.ps was longer and more variable than that of similar-sized e.p.ps. At endplates where nerve regeneration was induced by mechanical crushing of the motor nerve the frequency of m.e.p.ps was reduced and their amplitude distribution was broader than normal. The distribution of times to peak of m.e.p.ps was considerably broader than that of quantal e.p.ps recorded at the same endplates. At neuromuscular junctions poisoned with botulinum toxin type A, spontaneous and evoked transmitter release were greatly reduced. The amplitude distribution of m.e.p.ps was wider than that of e.p.ps and the time to peak of e.p.ps was about twice as fast as and less variable than that of m.e.p.ps. To explain the observed differences in time to peak among m.e.p.ps and between m.e.p.ps and quantal e.p.ps we suggest that some m.e.p.ps, but not e.p.ps, originate from transmitter quanta released from sites at a greater distance from postsynaptic receptors or that the release or diffusion process for acetylcholine is more prolonged when producing some of the m.e.p.ps. Such mechanisms produce at normal junctions a small population of m.e.p.ps with prolonged times to peak, at regenerating junctions a greater proportion of such m.e.p.ps and in botulinum toxin poisoning a majority.