Separation of the inlet flow in an internal combustion engine around the valve gap is analysed on the basis of data available in the literature as well as on experimental data obtained under steady and unsteady flow conditions with a moving piston at fixed valve lifts. The analysis comprises mass flow rate and pressure drop measurements, surface flow visualization and spatial and temporal distributions of velocity components in the valve gap measured by laser-Doppler anemometry (LDA). The results show that the simple four-mode model widely used in the literature is not sufficient for proper description of the complex flow separation phenomenon. The how separation is distributed non-uniformly across the valve gap width and around the valve periphery. In the light of these properties, single-point investigations using LDA or hot-wire anemometry techniques must be criticized. More attention must be paid to proper comparison of results obtained by different authors or by experiment and computational fluid dynamics (CFD). The valve lift and valve geometry alone are not sufficient for proper data comparison and interpretation; the valve location in the engine head must also be taken into account. For time-dependent how conditions, two additional parameters, crank angle and engine (piston) speed, are very important.