Surveilling the scene: learning about the world from multiple 2D dynamic views

CCTV surveillance is an ideal model to understand the nature and time course of spatial learning when viewing the world indirectly through multiple simultaneous videos, as increasingly happens in our everyday life. We reproduced a six-scene CCTV multiplex display on a computer screen showing videos of different rooms of a public building. Participants tracked an actor walking across the rooms. We used eye movements as an index of spatial learning. After a training phase, with the same scene content and display arrangement repeated across a series of trials, we interrupted learning by changing the display arrangement (Expt. 1) and the viewpoint of each room (Expt. 2) or the walked path (Expt. 3), which then remained constant in the trials of test phase. We analysed the number of gaze transitions between scenes, comparing the real data in the test phase to what would be predicted from the training phase if no interruption had occurred. The interruption caused a significant learning disruption in each experiment, increasing the number of gaze transitions. However, this effect was short lasting (1-3 trials). The most disruptive manipulation was the path change, which tests understanding of the relationships between the scenes. Overall, our findings suggest that learning was not limited to the display arrangement or the content of each room, but extended to the layout of the whole 3D environment. Indeed, it included representation of how the rooms are connected to each other in the real space, even though this aspect appeared relatively harder to learn.