A mechanical goal of in vitro testing systems is to minimize differences between applied and actual forces and moments experienced by spinal units. This study quantified the joint reaction forces and reaction flexion–extension moments during dynamic compression loading imposed throughout the physiological flexion–extension range of motion. Constrained (fixed base) and unconstrained (floating base) testing systems were compared. Sixteen porcine spinal units were assigned to both testing groups. Following conditioning tests, specimens were dynamically loaded for 1 cycle with a 1 Hz compression waveform to a peak load of 1 kN and 2 kN while positioned in five different postures (neutral, 100% and 300% of the flexion and extension neutral zone), totaling ten trials per functional spinal unit (FSU). A six degree-of-freedom force and torque sensor was used to measure peak reaction forces and moments for each trial. Shear reaction forces were significantly greater (25.5 N–85.7 N) when the testing system was constrained compared to unconstrained (p < 0.029). The reaction moment was influenced by posture (p = 0.037), particularly in C5C6 spinal units. In 300% extension (C5C6), the reaction moment was, on average, 9.9 N·m greater than the applied moment in both testing systems and differed from all other postures (p < 0.001). The reaction moment error was, on average, 0.45 N·m at all other postures. In conclusion, these findings demonstrate that comparable reaction moments can be achieved with unconstrained systems, but without inducing appreciable shear reaction forces.