Abstract
Theories of visual perception hold that visual scenes are initially decomposed into separate feature dimensions (e.g. color and shape), each represented on a separate map. Representations on these maps are 'unbound', meaning that they code only feature presence but not location or to which object they belong. It is assumed that we have no conscious access to information represented on feature maps, and that conscious perception requires that features belonging to an object be 'bound' by spatial attention. In the present study, we investigated whether unbound representations on feature maps can support at least awareness that 'something' has changed even when binding and recognition of the changing object fail. We presented scenes with tricolored objects in a change blindness paradigm. Each scene was presented several times, and one of the objects was changed from one presentation to the next. Changes either involved colors that were already present elsewhere in the display or entirely new colors. We reasoned that only the second type of change would lead to a substantial change on the color feature map while the first type would not. Observers were first asked to detect whether any change had occurred at all and then to localize the changing object. Observers were generally better at detecting and localizing changes when the change involved a new feature. Importantly, observers sometimes correctly detected the presence of a change without being able to localize it, but only when the change involved a new feature. This finding implies that pre-attentive processing of a change on a map of unbound features can give rise to a particular kind of perception - that something has changed somewhere - even when feature binding and perception of the object's identity fails. This process may underlie the previously demonstrated experience of 'sensing of changes’ in change blindness tasks.
Meeting abstract presented at VSS 2012