Abstract
Knowing what will be tested, and fixating the relevant areas of a scene, is expected to yield better memory. This expectation fails in the case of boundary extension (memory beyond the boundaries of a view). Eye-tracking revealed that, when forewarned about the nature of the memory test, compared to test-naïve observers, test-informed observers more frequently fixated markers for boundary placement — in particular, the place where a boundary cropped the main object (Dickinson, Michod & Intraub, 2008). Yet, at test, boundary extension was at least as great for pictures in which the main object was cropped. Was memory unaffected by these fixations? Or, might memory for the cropped boundary have been accurate, but associated with an increase in boundary extension at the other boundaries, thus resulting in a similar rating? To test this, observers (N = 72) studied 12 single-object-photographs (6 with cropped objects) for 14s each, and were instructed to memorize the objects, layout and background. Half the observers were test-informed and half were test-naïve during presentation. At test, the same views were presented again and observers reported “same” or they adjusted the boundaries to recreate the remembered view (using the mouse). As before, boundary extension was greater in the test-naïve condition (13% area increase) than in the test-informed condition (4% area increase; t (70) = 3.6, pcropped scenes in the test-informed condition revealed that the cropped side was the only one that yielded significant boundary extension (p [[lt]].01). Results run counter to traditional models of eye movements and scene representation and will be discussed in terms of a new multi-source model of scene perception (Intraub & Dickinson, 2008) in which visually-derived information and amodally-derived information play equally important roles.