Visual short-term memory binding and attentional processes during object-scene integration are preserved in mild cognitive impairment.

Binding, a critical cognitive process likely mediated by attention, is essential for creating coherent object representations within a scene. This process is vulnerable in individuals with dementia, who exhibit deficits in visual working memory (VWM) binding, primarily tested using abstract arrays of standalone objects. To explore how binding operates in more realistic settings across the lifespan, we examined the impact of object saliency and semantic consistency on VWM binding and the role of overt attention. Using an eye-tracking change detection task, we compared younger adults, healthy older adults, and individuals with Mild Cognitive Impairment (MCI). Participants were presented with naturalistic scenes and asked to detect changes in the identity and/or location of objects that were either semantically consistent or inconsistent with their scene context. Across all age groups, semantically inconsistent objects were prioritised during encoding, leading to better change detection than consistent objects. Highly salient objects decreased the inconsistency advantage while being detrimental to detection accuracy when inspected at longer latencies to the first fixation. Longer fixation durations on the critical object were beneficial for recognition. In contrast, delayed initial inspection or frequent subsequent fixations on other objects were detrimental to detection, regardless of age or cognitive impairment. These findings challenge the notion of generalised semantic memory impairment in the prodromal stages of dementia and highlight the importance of efficient attentional control in supporting VWM binding, even in the face of cognitive decline. Overall, preserved low-level and high-level mechanisms of object-scene integration can compensate for age-related cognitive decline, enabling successful binding in naturalistic contexts.