Estimation of contact state is important to any multi-point interaction that involves frictional stick/slip phenomena. In particular, when there are more kinematic constraints than there are degrees of freedom, some contact interfaces must slip, leading to the need for contact state estimation. Fortunately, supervisory control techniques from adaptive control can he applied to this problem with relatively little modification. We discuss this approach in terms of a distributed manipulation experiment developed to explore overconstrained manipulation. In this context, we show in a simulated model that on-line contact state estimation dramatically improves performance over methods that estimate contact states off-line.