Memory monitoring failure in confabulation: evidence from the semantic illusion paradigm

Several prominent models of confabulation characterize the syndrome as a failure in controlled aspects of memory retrieval, such as pre-retrieval cue specification and post-retrieval monitoring. These models have been generated primarily in the context of studies of autobiographical memory retrieval. Less research has focused on the existence and mechanisms of semantic confabulation. We examined whether confabulation extends to the semantic domain, and if so, whether it could be understood as a monitoring failure. We focus on post-retrieval monitoring by using a verification task that minimizes cue specification demands. We used the semantic illusion paradigm that elicits erroneous endorsement of misleading statements (e.g., "Two animals of each kind were brought onto the Ark by Moses before the great flood") even in controls, despite their knowing the correct answer (e.g., Noah). Monitoring demands were manipulated by varying semantic overlap between target and foils, ranging from high semantic overlap to unrelated. We found that semantic overlap modulated the magnitude of semantic illusion in all groups. Compared to controls, both confabulators and non-confabulators had greater difficulty monitoring semantically related foils; however, elevated endorsement of unrelated foils was unique to confabulators. We interpret our findings in the context of a two-process model of post-retrieval monitoring.