The control of intake air swirl motion is often used in spark-ignition direct-injection (SIDI) engine to improve its in-cylinder fuel-air mixing process especially under engine idle and low load conditions. In this experimental investigation, a novel technique combining the time-resolved particle image velocimetry (PIV) with quadruple proper orthogonal decomposition (POD) is implemented to analyze the time-resolved in-cylinder velocity measurements in an optically-accessible SIDI engine. The intake air swirl motion is introduced into the engine cylinder by a control valve installed in one of two air intake ports. Experimental results show that a strong linear correlation exists between the intake flow swirl ratio and vorticity flow field in the cylinder. This correlation ensures high data reliability of swirl motion control and provides a novel basis to directly compare the flow field measurements under different swirl ratio conditions. The quadruple proper orthogonal decomposition analysis is then applied to the velocity flow fields to separate the highly dynamic in-cylinder flow characteristics into four distinct categories: (1) dominant flow structure; (2) coherent structure; (3) turbulent structure; and (4) noise structure. The results show that the dominant flow structure varies strongly with swirl ratio, and its kinetic energy is also directly related to the swirl ratio. The coherent structure captures the large scale flow characteristics, but its kinetic energy is much lower and exhibits larger cycle-to-cycle variations. The turbulent structure contains similar level of kinetic energy at different swirl ratios but without much cycle-to-cycle variation. Finally, the noise structure contains very low kinetic energy which only alters the dynamic nature of the flow field slightly. In summary, the effect of swirl ratio on in-cylinder flow field is mostly captured by the dominant flow structure and partially captured by the coherent flow structure. The turbulent flow structure can characterize the high-order flow variation. The noise structure can be neglected due to the low energy captured. (C) 2015 Elsevier Ltd. All rights reserved.