UC San Diego

The dissertation research described herein examined several facets of pattern separation (conceptualized as an encoding process whereby two very similar stimuli are stored as discrete representations). Previous research has conceptualized pattern separation as a discrete, high threshold process and has relied upon performance measures that are derived from that theory. More recently, it has been suggested that pattern separation may be better fit by a continuous, signal-detection-based model. Should that be the case, it would mean that previous work examining pattern separation in aging and dementia has relied upon potentially misleading dependent measures.

The current project had three goals: 1) to determine the most appropriate cognitive modeling for pattern separation, 2) to examine whether or not pattern separation is selectively impaired in aging and dementia populations, and 3) to compare biomarkers of preclinical impairment with performance on behavioral recognition memory and pattern separation tasks. In Chapter 1, ROC and d’ analyses showed that pattern separation is indeed best conceptualized as a continuous signal detection process. In Chapters 2 and 3, we took the framework for pattern separation identified in Chapter 1 and applied it to a population consisting of young adults, older adults, individuals with amnestic mild cognitive impairment (aMCI), and Alzheimer’s disease. The general approach was to first try to equate performance across groups on a recognition memory task and then compare group performance on the pattern separation task. Taken together, Chapters 2 and 3 provided evidence that pattern separation is selectively impaired in aMCI and possibly in healthy older adults as well, and this deficit is not explained by a recognition deficit alone. In Chapter 4 we measured the relationship between behavioral performance measures (on both the recognition memory tasks and pattern separation tasks) and cerebrospinal fluid biomarkers (Aβ42 and tau levels) in healthy aging participants. We found a significant positive correlation between performance on both the recognition memory task and the pattern separation task, but only for pictures of everyday objects. Thus, these biomarkers are predictive of general memory performance for objects but do not selectively predict pattern separation performance.