1. A study was made of functional and structural changes during degeneration of end-plates in the rat diaphragm after phrenic nerve section at two levels.2. For 8-10 hr after cutting the nerve in the neck, all end-plates retain the ability to transmit impulses. During the following 8-10 hr, an increasing number of end-plates lose this ability so that after a total of about 20 hr, no end-plates can transmit.3. Transmission failure occurs abruptly at most end-plates. This failure is usually accompanied by cessation of spontaneous miniature end-plate potentials (min.e.p.p.s), though in a few cases min.e.p.p.s persist after junctional transmission has failed. Several degenerating junctions were observed where the frequency of min.e.p.p.s was very low, suggesting an intermediate stage in min.e.p.p. failure.4. The time of junctional failure depends on the length of the degenerating nerve stump. For each additional centimetre of nerve, failure is delayed about 45 min.5. Changes in ultrastructure of nerve endings closely parallel those of function. For about 8-12 hr after cutting the nerve, nearly all end-plates appear normal. During the period when transmission is failing, some end-plates are clearly undergoing structural break-down. By the time functional failure is complete, all end-plates appear grossly abnormal.6. During degeneration, the contents of the axoplasm undergo disruption and the nerve terminal breaks up into small fragments. In contrast, the Schwann cell appears to become very active and its processes extend into the synaptic cleft to surround fragments of the nerve terminal. Ultimately, the Schwann cell completely replaces the axon at the end-plate.7. Increasing the length of the peripheral nerve stump delays the onset of structural break-down. Disruption of end-plates near the site of nerve entry into the muscle occurs before those farther away.8. It is suggested that end-plate degeneration is triggered by a signal which passes from the site of injury to the nerve terminal. The duration of the period after transection when end-plates appear to be normal would then reflect the time required for this signal to travel the length of the isolated nerve stump.