The results of an experimental study on the statistical structure of turbulence in an automotive engine are reported, with specific reference to the time-frequency domains. Autocorrelation and autospectral density coefficients were evaluated in consecutive crank-angle intervals throughout the induction and compression strokes. Eulerian time scales were obtained on the analogy of both the micro and integral time scales of turbulence for stationary flows.The spatial distribution of the turbulence structure was investigated in the combustion chamber of a diesel engine with a shallow in-piston bowl and two tangential intake ducts. The study was carried out for different swirl flow conditions, produced by deactivating one intake duct and/or by changing the engine speed. The velocity data were acquired using an advanced HWA technique, under motored conditions.The method of time-frequency spectral analysis used was previously developed in order to estimate the average statistical properties of segmented nonstationary sample records. It is based on an alternative definition of the autocorrelation coefficient so as to reduce this to an even function of the separation time (over which the turbulent fluctuation is correlated), independent of the time instant, within specific correlation periods during the engine cycle. This procedure was applied to both the cycle-resolved turbulent fluctuation and the instantaneous velocity fluctuation, which represents turbulence in the more conventional sense.An insight is also given into the normalized probability density distributions of these fluctuations.