Objective. To compare the joint torque pattern and dynamic joint stiffness at the knee and ankle in elderly and young men during stepping down. Background. Adequate joint stiffness is critical during the single support phase to control forward and downward body momentum. Design. Six active elderly men (mean 67.7) and six young men (mean 23.6) of similar body mass and height, were filmed stepping down from one force platform to another. Repeated trials were undertaken at three different step heights (200, 250, and 300 mm). Method. Joint torques were determined for the ankle and knee of the support limb throughout the single support phase. The gradient of the joint torque-angle graph was calculated to define dynamic joint stiffness of the ankle and knee in two phases; (I) from initiation of movement until heel-off of the supporting limb, and (II) from heel-off of the supporting limb to contra-limb touch down. Results. Maximum ankle torque values were lower in the elderly and occurred at a larger dorsiflexion angle (P<0.05). Knee torque patterns were similar in both groups. Phase I ankle stiffness was significantly less in the elderly (4.0-5.2 Nm/°) at all step heights compared to the young (7.6 - 8.7 Nm/°). In both groups ankle stiffness in Phase II increased with step height, while knee joint stiffness decreased. Conclusions. The different torque pattern and lower dynamic ankle stiffness in the elderly, particularly for Phase I, suggested an altered control strategy. These findings highlight the importance of dynamic ankle joint stiffness in stepping down. © 2003 Elsevier Ltd. All rights reserved.