We use sequences of images and magnetograms from Hinode to study magnetic elements in internetwork parts of the quiet solar photosphere. Visual inspection shows the existence of many long-lived (several hours) structures that interact frequently, and
may migrate over distances ~7 Mm over a period of a few hours. About a fifth of the elements have an associated bright point in G-band or Ca II H intensity. We apply a hysteresis-based algorithm to identify elements. The algorithm is able to track elements for about 10 min on average. Elements intermittently drop below the detection limit, though the associated flux apparently persists and often reappears some time later. We infer proper motions of elements from their successive positions, and find that they obey a Gaussian distribution with an rms of 1.57+-0.08 km/s. The apparent flows indicate a bias of about 0.2 km/s toward the network boundary. Elements of negative polarity show a higher bias than elements of positive polarity, perhaps as a result of to the dominant positive polarity of the network in the field of view, or because of increased mobility due to their smaller size. A preference for motions in X is likely explained by higher supergranular flow in that direction. We search for emerging bipoles by grouping elements of opposite polarity that appear close together in space and time. We find no evidence supporting Joys law at arcsecond scales.
