We performed a prospective, randomised trial of 44 patients to compare the functional outcomes of a posterior-cruciate-ligament-retaining and posterior-cruciate-ligament-substituting total knee arthroplasty, and to gain a better understanding of the in vivo kinematic behaviour of both devices. At follow-up at five years, no statistically significant differences were found in the clinical outcome measurements for either design. The prevalence of radiolucent lines and the survivorship were the same. In a subgroup of 15 knees, additional image-intensifier analysis in the horizontal and sagittal planes was performed during step-up and lunge activity. Our analysis revealed striking differences. Lunge activity showed a mean posterior displacement of both medial and lateral tibiofemoral contact areas (roll-back) which was greater and more consistent in the cruciate-substituting than in the cruciate-retaining group (medial p < 0.0001, lateral p = 0.011). The amount of posterior displacement could predict the maximum flexion which could be achieved (p = 0.018). Forward displacement of the tibiofemoral contact area in flexion during stair activity was seen more in the cruciate-retaining than in the cruciate-substituting group. This was attributed mainly to insufficiency of the posterior cruciate ligament and partially to that of the anterior cruciate ligament. We concluded that, despite similar clinical outcomes, there are significant kinematic differences between cruciate-retaining and cruciate-substituting arthroplasties.