A two-color Particle Image Velocimetry (PIV) technique has been applied to a single cylinder, cup-in-head two-stroke research engine. In the two-color PIV, two wavelengths are used to successively record, at a known time separation, the positions of the particulate seeds in the flowfield. By separately interrogating the two images of different color and cross-correlating them, a two-dimensional velocity field is obtained. Since the sequence of the images is known, directional ambiguity is eliminated and two-color PIV can be used to study complex flows.The technique is here applied for the first time to an engine in the presence of combustion. The presence of combustion light complicates the application of two-color PIV because narrow band-pass laser line filters can not be used to reject it, however a suitable combination of laser power, colored glass filters and thresholding during interrogation allowed sufficiently high quality images to be obtained. Velocity fields were measured at a plane parallel to the piston crown at the half-height of the cup. Multiple images were made at this plane and ensemble averaged velocity and velocity fluctuation were obtained. This exercise shows that good quality PIV data can be obtained ahead of the flame but with the seeding used (boron nitride and zirconia) the velocity information in the burned gases is erratic due to the low seed density.TURBULENT COMBUSTION in spark ignition engines is strongly influenced by the in-cylinder flow field (eg. ). These flow fields are unsteady with turbulent fluctuations over a wide range of spatial and temporal scales. Much insight into these flow fields has been gained by flow visualization and by using single point LDV measurements (eg. ,  & ). Scanning LDV has provided the ability to obtain nearly simultaneous velocity data along a line through the combustion chamber . To more fully understand the flow field it is highly desirable to obtain quantitative, simultaneous velocity measurements at many points in the field. Such measurements are also particularly suitable for comparison with the results from CFD models. Such measurements are particularly useful if they are capable of resolving the turbulence integral length scale, which in engines is of the order of a few millimeters near TDC . Measurements of this type have recently been made in engines using particle tracking velocimetry, PTV,  and particle image velocimetry, PIV, [8,9,10 & 11]. In the case of PTV, the light source was multiply pulsed, giving particle tracks, consequently, to permit unambiguous identification of individual tracks and the measurement of displacements, a rather low seed density was used. Thus the velocity vector data density is low, randomly distributed and averaged over a number of light pulses. In PIV, typically only two laser pulses are used, and with the appropriate mode of analysis the velocity can be measured on a closely spaced rectangular grid. The PIV technique has been applied in both a motored and fired engine [8,9] with a spatial resolution of 1 mm. In both of these cases the particle images were recorded photographically and the particle displacements were measured using an auto-correlation technique.Two weaknesses are inherent to the auto-correlation technique. The first is the existence of an auto-correlation peak, located at the origin in the spatial separation domain. This peak, corresponding to zero particle displacement, is formed by a particle image correlating with itself. By definition this peak is always greater than any other peak, but it does not represent the particle displacement. The existence of this peak adds complexity to locating the true particle displacement. In addition, particle pair separations less than twice the diameter of an average particle are difficult to distinguish from the auto-correlation peak. The second weakness is the 180° directional ambiguity of the displacement vector that exists in the spatial separation domain. This ambiguity exists because it is not possible to determine which of the two particle images was formed first.Some of the previous attempts to eliminate directional ambiguity have used image shifting techniques. In this technique the image field is displaced a known uniform distance between the first and second light pulse. The displacement is chosen so that the most negative fluid velocities will produce an effective positive displacement. This artificial shift is then subtracted from the results, leaving the actual fluid velocity. This shift can either be obtained by placing a rotating mirror in front of the camera , or by moving the camera/film between the first and second laser pulses, or by an electro-optical image shifting technique involving polarization of the incoming light to displace the image .Another method of eliminating directional ambiguity involves sequentially recording the flow field on a number of different frames, either on film or with a video camera. In digital particle image velocimetry (DPIV) successive digitally recorded video images are used . Two images of the flow field at time t and t+Δt are recorded on different frames of the video system. The separate images are then cross-correlated with each other. The cross-correlation produces an unambiguous displacement vector and does not produce an auto-correlation peak. The elimination of the auto-correlation peak simplifies the identification of the particle displacement and has been shown to allow a greater dynamic range than the auto-correlation method . Although digital particle image velocimetry is a very powerful technique, it is presently limited by the video systems available. The data acquisition rates of currently available cameras of sufficient resolution are too slow to distinguish flow characteristics of interest in engines, as the laser pulse separation used in engine applications is currently on the order of 10 μs.Another method for resolving the directional ambiguity is to use different colors for the illuminating laser sheets and a color sensitive method of interrogation [ 15, 10 & 11 ]; this is the method that was selected for the present work. The method of recording the images was similar to that used by Goss et al. . With this technique, two differently colored sheets of light illuminate the flow field, and this color coding is used to distinguish the first image from the second image. Filters are employed to separate the two colors during interrogation, thus creating two images which are then cross-correlated with each other, similar to the approach of Willert & Gharib . This method and a detailed evaluation of its performance is described by Stucky et al. . A fairly detailed study of the flow in a motored internal combustion engine was carried out by Nino et al. .This article describes the application of the two-color PIV system to a firing internal combustion engine. The major advantages of a two-color system is the ability to resolve the directional ambiguity and the improved dynamic range, so that PIV may be used to study complex, recirculating flows such as those found in an internal combustion engine.