Abstract
Visual memory is intrinsically linked to the reinstatement of low-level visual features, such as edges and luminosity, within early visual cortex. However, individuals with severely deficient autobiographical memory (SDAM) cannot vividly recollect autobiographical experiences yet display normal everyday functioning. We hypothesized that such individuals would depend on semantic features instead of low-level visual features during a challenging visual recognition task due to impaired communication of low-level visual information between the posterior hippocampus and early visual cortex. Two methods were used to measure the content of memory derived from fMRI data collected at encoding and retrieval: one directly measuring feature-specific neural reactivation within the hippocampus and other cortical regions, and another modeling top–down inference to assess the influence of semantic-based recall on reactivation within early visual cortex. In accord with prior findings, recognition accuracy in non-SDAM individuals was linked to low-level visual reactivation within early visual cortex and posterior hippocampus. As predicted, this association was diminished in SDAM individuals, whose recognition accuracy was instead linked to semantic-based reactivation. In addition, non-SDAM individuals exhibited communication of low-level visual information between early visual cortex and hippocampus, whereas SDAM individuals showed communication of semantic information. Given that SDAM participants' performance on the visual memory task was equivalent to non-SDAM subjects, our findings suggest that SDAM individuals successfully compensate for impaired low-level visual memory through semantic recall and highlight the essential role of feature-specific reactivation measures in identifying distinct neural pathways to memory performance.