An Embedded Computational Framework of Age-Related Memory Change

Research on human memory has expanded rapidly over the past decades, leading to the development of numerous models and experimental paradigms. However, this growth has also resulted in increasing specialization, with limited integration across models, tasks, age groups, and domains. Here, we present the Embedded Computational Framework of Memory (eCFM), a simple yet flexible computational model that integrates encoding, storage, retrieval, and decision processes with structured representations that capture visual, orthographic, phonological, and semantic information. We tested the generalizability of the eCFM across 11 experiments involving approximately 40 younger (18 to 25 years) and 40 older adults (65 to 80 years) per experiment. In Experiments 1 to 4, participants completed order reconstruction tasks using words or images varying in semantic, phonological, orthographic, or visual similarity. Experiments 5 to 7 tested serial recall using the same materials. In Experiments 8 to 11, participants performed old/new recognition tests with studied and unstudied items with similar representational forms. Results showed age-related differences in serial recall, smaller differences in order reconstruction, and none in recognition. Across all tasks and domains, participants consistently showed high false memory rates for related foils, with no age-related differences. The eCFM accounted for all main findings and captured age-related differences, and the lack thereof, by manipulating encoding strength and position discriminability. Overall, the model represents a step toward a general account of human memory and highlights the potential and value of integrating traditionally separate areas of memory research.