Compressor surge is shown by the application of several types of instrumentation (notably a hot-wire anemometer) to consist of two distinct types of phenomena. The whole compressor flow system may be unstable in the manner of a self-excited Helmholtz resonator. The theory of this instability is presented and is shown to explain some of the observed pulsation symptoms. The stalling of the flow through the blade rows, which usually is assumed to be the origin of pulsation, is shown to occur in propagating groups of 1 to 5 regions involving from 2 to 20 blades each. The theory of this “stall propagation” shows the propagation velocity relative to the wheel to be dependent upon boundary-layer growth parameters and hence the frequency (relative to a stationary probe) to be proportional to the wheel speed. Another part of observed compressor pulsation thus is explained. These two phenomena frequently interact to produce complex performance characteristics. The theories presented are essentially correct as shown by experimental verification, but much remains to be done to make quantitative compressor-performance prediction practical.