Cat and dog, sign and wine: The Related Recency Model of proactive interference in verbal working memory

Proactive interference (hereafter, PI) is the unwanted persistence of information from the recent past upon current cognitive processing; resolving PI has been shown to be a critical determinant of WM capacity. Despite the substantial progress that has been made to date, research remains equivocal regarding key aspects of how interference is created, how it might be resolved, and the neural bases of both. The primary goal of the experiments described herein was to outline and test a novel account (the Related Recency Model ) of the mechanisms of PI in working memory in a well known item-recognition task, where the overarching theme was the content-specific nature of those mechanisms. The Related Recency Model was motivated by the following principles: (1) working memory is an emergent property of the long-term memory system, where attention serves to temporarily maintain activation of task-relevant information, (2) items thusly in working memory are composed of distributed features and their representations are content-addressable, and (3) in the absence of attention/rehearsal processes, the activation of information decays from working memory. Content-specific PI was predicted to occur in the task when the current probe was semantically or phonologically similar to a prior list item; resolution was predicted to be dependent upon the ability to maintain item and contextual information, and the integrity of neural regions within the left inferior frontal gyrus (LIFG). Across five experiments, we found evidence consistent with the new model. Specificially, Experiments 1-3 showed that (a) age affected the binding of item and associated context information at the time of encoding such that older adults demonstrated larger semantic and phonological interference effects relative to younger adults, and (b) younger adults\u27 susceptibility to PI could be increased by in essence degrading their ability to establish and maintain context information. Experiments 4-5 showed that strength and specificity of exaggerated PI could be determined by a combination of content-specific STM deficits and lesion damage. Taken together, these experiments provide converging behavioral and neuropsychological evidence in support of the Related Recency Model