Knock in a HCCI engine was examined by comparing subjective evaluation, recorded sound radiation from the engine, and cylinder pressure. Because HCCI combustion involved simultaneous heat release in a spatially large region, substantial oscillations were often found in the pressure signal. The time development of the audible signal within a knock cycle was different from that of the pressure trace. Thus the audible signal was not the attenuated transmission of the cylinder pressure oscillation but the sound radiation from the engine structure vibration excited by the initial few cycles of pressure oscillation. A practical knock limited maximum load point for the specific 2.3 L I4 engine under test (and arguably for engines of similar size and geometry) was defined at when the maximum rate of cycle-averaged pressure rise reached 5 MPa/ms.