Abstract
Previous studies have demonstrated that the storage capacity of visual short-term memory (VSTM) for single features could be enhanced by grouping the features into objects (e.g., Luck & Vogel, 1997; Wheeler & Treisman, 2002). While such an object-based encoding benefit is usually found when the features are from the same part of an object (e.g., Delvenne & Bruyer, 2004; Walker & Cuthbert, 1998), it is uncertain whether it can also be observed when the features are located on different parts of an object. In a recent change detection study, Xu (2002) proposed that these features might still benefit from an object-based encoding when they come from different dimensions. She demonstrated that features (colors and orientations) from different parts of an object are still better encoded in VSTM than features from spatially separated objects. However, the use of the orientation feature in Xu's (2002) study may have led participants to encode a complex pattern (or configuration) of orientations as a gestalt figure rather than a set of individual items. Here, we tested whether an object-based encoding benefit could still be observed for features from different parts of an object when spatial configuration is properly controlled. Results showed that, when participants were encouraged to encode the orientations as individual items rather than as part of a global configuration, the object-based encoding benefit was significantly reduced. Furthermore, when the orientation feature was replaced by other features, such as shapes or a position of a gap on a square, no object-based encoding benefit was found. These findings suggest important limits of bottom-up perceptual organization on the object-based encoding, as features might not benefit from an object-based encoding when they come from different parts of an object.