Predictive and reactive control mechanisms involved in object manipulation

Abstract

As humans, we rely on our ability to skillfully manipulate objects to complete tasks in our daily lives. This ability depends critically on our predictions of the mechanical properties of objects (i.e. weight), as well as rapid corrective responses, initiated when we encounter perturbations that interfere with our movement goals. In this thesis, we looked to investigate some of the predictive and reactive mechanisms supporting manipulation in the context of object lifting. When lifting an object, weight is estimated based on learned associations between the visual features of objects (e.g. size, material) and their weight. However, occasionally these predictions are erroneous and corrective responses must be engaged. For example, when an object is heavier than expected and fails to lift off at the expected moment, the corrective lifting response is initiated within ∼90 ms, consisting of an increase in lifting force. In the first study, we had participants perform an object lifting task to investigate the hypothesis that the lifting response can adapt based on memory of an object’s weight history. The results confirmed this hypothesis, demonstrating the flexible adaptability of the lifting response. In the second study, we replicated our previous findings in an augmented reality lifting task, but were not able to show that the lifting response can simultaneously adapt to the independent weight histories of two objects. In the third study, we found that adaptation of the lifting response transfers intermanually. In the fourth study, we moved our focus to the prediction of object weight, specifically the object families hypothesis, which proposes that memories of object dynamics are organized hierarchically in memory based on covariation in their visual and mechanical properties. Our findings showed that one of the benefits of this categorical encoding is an improved ability to learn the weights of multiple families of objects if they each follow a linear family structure. Together, these findings further support the role of object families in the prediction of object weight, and demonstrate the importance of memory for the flexible adaptation of rapid corrective responses in manipulation.