According to one embodiment, an object holding apparatus includes a holding part, a recognition device, and a controller. The holding part is able to hold at least one object. The recognition device recognizes a plurality of objects to generate a recognition result. The controller selects a first object from the objects, based on the recognition result. The controller sets a first direction to one side of the first object along which the number of objects aligned with the first object is smaller than the number of objects aligned with the first object along the other side of the first object. The controller selects a second object aligned with the first object along the first direction. The controller controls a driving of the holding part, based on a selection result of the first object and the second object.