Attention and scene representations

Human observers have a vast capacity for remembering scenes. At the same time, observers tend to miss large image changes when transients are masked (change blindness), and report remembering more of the scene than was actually presented (boundary extension). This suggests that memory for scenes is not strictly veridical. Instead, we store summary information about the scene, the "gist", and add interpolated information beyond the visible boundaries. Is this an automatic process, or does it require attentional resources? We studied the role of attention in scene representation using dual-task methods. When we used a concurrent visual search task to divert attention from scene processing, we found that boundary extension was actually increased. Furthermore, memory for even the simplest scenes required attention. However, the nature of the attentional task is important. When we used a concurrent multiple object tracking task, however, scene memory was unimpaired. We suggest a distinction between “ventral” and “dorsal” attentional resources. The former represents the bottleneck restricting access to object identification mechanisms, while the latter involves tracking object continuity over time. We do not passively form scene representations without actively attending to the scene.

Back to Program