The brain needs to identify redundant sensory signals in order to integrate them optimally. The identification process, referred to as causal inference, depends on the spatial and temporal correspondence of the incoming sensory signals (‘online sensory causality evidence’) as well as on prior expectations regarding their causal relation. We here examine whether the same causal inference process underlies spatial integration of actions and their visual consequences. We used a basic cursor-control task for which online sensory causality evidence is provided by the correlated hand and cursor movements, and prior expectations are formed by everyday experience of such correlated movements. Participants made out-and-back movements and subsequently judged the hand or cursor movement endpoints. In one condition, we omitted the online sensory causality evidence by showing the cursor only at the movement endpoint. The integration strength was lower than in conditions where the cursor was visible during the outward movement, but a substantial level of integration persisted. These findings support the hypothesis that the binding of actions and their visual consequences is based on the general mechanism of optimal integration, and they specifically show that such binding can occur even if it is previous experience only that identifies the action consequence.