Surya Gayet¹ (), Süheyla Aydemir¹; ¹Utrecht University, Helmholtz Institute

Any object that we look at comprises many different properties (e.g., a color, shape, orientation, and spatial location), all of which are processed in separate brain regions. According to Anne Treisman’s influential Feature Integration Theory of attention, these individual properties are processed pre-attentively, whereas attention is needed to ‘bind’ them into integrated objects. Here, we ask during which processing stage attention contributes to binding: during stimulus encoding (to encode stimuli as integrated objects) or during maintenance (to prevent objects from falling apart into their constituents)? In eight lab-based behavioral experiments (N = 8*34), participants memorized four briefly presented visual objects. To manipulate the amount of attention, two of these objects were (pre or post) cued to be likely targets (with 75% validity). After a delay, one object was probed, and participants had to report its features from memory (e.g., color, location, orientation, category). Recall performance was far better for cued than for uncued items, indicating the success of the attention manipulation. Also, stimulus properties were generally bound, so that recall performance on one feature strongly predicted recall performance on another feature of that same object. However, attention only increased binding strength between features (e.g., color and orientation), but not between features and locations, suggesting that feature-location binding is a pre-attentive process. Finally –and most surprisingly– the effect of attention on (feature-feature) binding did not depend on whether the cue preceded or succeeded the memory array, showing that attention contributes to binding exclusively during memory maintenance. These findings expand the Feature Integration Theory of attention, by showing that (1) not only the encoding of features occurs pre-attentively, but also the binding of features to locations, and by showing that (2) the role of attention is not to integrate objects during stimulus encoding, but to keep features bound together during stimulus maintenance.