Short-interval intracortical inhibition (SICI) is a widely used method to study cortical inhibition, and abnormalities have been found in several neurological and psychiatric disorders. Previous studies suggested that SICI involves two phases and the first phase may be explained by axonal refractoriness. Our objectives are to further investigate the mechanisms of the two phases of SICI. SICI was studied in 11 normal volunteers by a paired transcranial magnetic stimulation (TMS) paradigm applied to the left motor cortex with a subthreshold conditioning stimulus (80% resting motor threshold for rest condition and 95% active motor threshold for active condition) followed by a suprathreshold test stimulus at interstimulus intervals (ISIs) of 1-4.5 ms in steps of 0.5 ms. Motor-evoked potentials (MEPs) were recorded from the right first dorsal interosseous muscle. Three different test stimulus intensities adjusted to produce 0.2, 1 and 4 mV MEPs at rest were studied with the target muscle relaxed and during 20% maximum contraction. Maximum inhibition was observed at ISIs of 1 ms and 2.5 ms for the rest condition and the difference among ISIs was reduced with voluntary contraction. SICI increased with larger test MEP amplitude and decreased with voluntary contraction. At test MEP of 0.2 mV, some subjects showed facilitation and this is likely related to short-interval intracortical facilitation. For rest SICI, the correlation between adjacent ISIs was much higher from 3 to 4.5 ms than from 1 to 2.5 ms or between 1 and 2.5 ms. There was no correlation between SICI at different test MEP amplitudes. We conclude that maximum SICI at ISIs of 1 and 2.5 ms are mediated by different mechanisms. SICI at 1 ms cannot be fully explained by axonal refractoriness and synaptic inhibition may be involved. SICI is a complex phenomenon and inhibition at different ISIs may be mediated by different inhibitory circuits.