In-cylinder velocity measurements were acquired in a two-valve, single-cylinder research engine to study the bulk fluid motion and small-scale turbulence. Different port geometries (two), different port orientations (two) and both shrouded and non-shrouded intake valves were tested to vary the intake-generated flow. Tests were performed at engine speeds of 300, 600, 900 and 1200 RPM with an atmospheric intake pressure. Prior to testing on the engine, the different head configurations were tested on a steady flow bench. Particle image velocimetry data were taken on a single plane, parallel to the piston surface, in the engine using both a low magnification to characterize the large-scale flow phenomena, and a high magnification to characterize the turbulence field. The low-magnification results showed that the swirl center location was relatively insensitive to engine speed, but did change position throughout the cycle. The solid-body rotation rate was found to be comparable to the steady-flow bench swirl ratio for swirl-dominated flows, and the decay in the rotation rate during the cycle was measured. The high magnification data (acquired at TDC) were investigated using both ensemble- and spatial-averaging to define the mean flow field, and all of the results from the spatial-average method were investigated as a function of cutoff frequency. The turbulence intensity was found to be linear with engine speed, and the ratio of turbulence intensity to mean piston speed was found to be a unique function of cutoff frequency. Two-point spatial correlation functions were computed, and used to evaluate 8 of the possible 27 integral length scales. The longitudinal length scales were found to range from 5-8 mm, were insensitive to direction of separation, and were on average twice the lateral length scale, indicating a high level of isotropy in the flow. Turbulent kinetic energy spectra were calculated, and were found to show an extended inertial subrange for the higher engine speeds; the spectra were fit well by the model spectrum of Pope. Lower engine speeds and the use of high cutoff frequencies in the spatial-averaging method were found to reduce the presence of the inertial subrange, and may result in a low Reynolds number condition where the turbulence is not fully developed and scale separation is not achieved. The spectral analysis provided length scales (L₁₁) estimates that were approximately 2.5 smaller than those from the correlation results.