Abstract
People can recognize previously viewed objects based on specific visual features, even in incidental recognition tests. This has led to the proposal that abstract, detailed and scene-level visual memories are automatically constructed as a result of attending to objects in a scene. The current study tested if such retention is truly automatic and sensitive to object-context relationships. Subjects viewed object arrays consisting of upright and inverted objects and empty locations while performing a go/nogo cover task. In each trial all objects and locations of a given array were cued twice. For the Object task, responses were contingent on the presence of objects at cued locations (object: go, empty: nogo) and for the Upright task, responses were based on objects' orientations (upright: go, inverted: nogo, empty: nogo). After the cover task trials, subjects received a surprise recognition test. For each recognition trial, subjects chose which array was previously viewed from two alternative choices. The arrays were the same except for a token-level substitution (i.e. original and distracter items were from the same basic-level category). On half of the recognition trials, the distracter item had not been seen by the subject. On the other half, the distracter item was donated from another array the subject had seen during the cover task period. In the Object condition, recognition accuracy was at chance, while in the Upright condition recognition accuracy was above chance and equal for novel and donated distracters. This suggests A) that visual memories are not always created when objects are attended, and B) that objects were remembered as part of a particular array, but not necessarily as individuals.