Abstract
Previous studies have shown that schemas mediate the attention allocation within a scene and the memory for elements present in the scene (Friedman, 1979; Goodman, 1980). People attend to schema-irrelevant information more than schema-relevant information but they are more likely to recall schema-relevant items and to be better at recognising schema-irrelevant objects. This implies that recognition, but not recall, depends on having looked at an object. This was tested by requiring participants to name specific objects within the scene, before their memory was assessed. Objects were classified to be of high (schema-relevant) or low relevance (schema-irrelevant) to a salient action depicted in the scene. Analyses of eye-movements showed that the different task instructions biased participants towards looking at (a) schema-relevant objects, (b) schema-irrelevant objects, or (c) to share attention equally across both. Regardless of the differential allocation of attention, objects relevant to the action depicted were better recalled. In contrast, recognition performance was independent of schema-relevance, but depended on deployment of attention. Having people label items in the scene improved recall for schema-relevant objects but not for those unrelated to the schema. The fact that recall of schema-irrelevant objects was better when participants were biased to search for objects of this type, suggests that attention is critical for recall of objects whose retrieval is not supported by the currently active schema. These results suggest that in natural scenes both attention and the schematic content of a scene determine what is subsequently remembered. Recognition memory for visual detail requires attention, but information which is irrelevant to the action depicted in a scene is only likely to be recalled when it has been attended to. In short, knowledge of a domain allows us to remember what we haven't looked at. Without such knowledge, objects need to be fixated.