Perceptual similarity landscape differs between young and clinically unimpaired older adults and contributes to long-term memory precision

The precision of memory for event features is lower in older relative to young adults. Recent studies that examined continuous reproduction data with a mixture model –– which estimates two parameters putatively reflecting reduced fidelity/precision and outright forgetting –– suggest age-related memory decline is associated with a selective or larger reduction in mnemonic precision vs. forgetting. However, emerging evidence indicates that performance on continuous reproduction paradigms depends not just on memory, but also on the psychophysical similarity function (PSF) of the perceptual feature space, which may also change with age. Using a recently developed one-parameter statistical model of continuous reproduction memory data that mechanistically accounts for the empirically measured PSF, we examined whether age-related differences in continuous reproduction data are at least partially explained by age-related differences in perception. Young adults (age 18-25 yrs, N = 55) and clinically unimpaired older adults (age 68-86 yrs, N = 54) with normal color vision performed (a) a continuous reproduction task probing long-term memory for object color, (b) a Likert color similarity judgement task, and (c) a color matching task. We first replicated previous findings of age-related differences in memory precision for the colors of previously encountered objects and identified age-related changes in color perception on the similarity and matching tasks. We then characterized the age-group specific perceptual landscape by constructing the PSF using the similarity task data. Subsequent model comparison revealed that for both age groups, the more parsimonious one-parameter model, which accounts for the age-group specific PSF, explained continuous reproduction data better than the mixture model. Notably, the PSFs differed between age groups, and the difference was not explained by individual differences in response bias or trial-history effects. When we replaced the PSF estimated from older adults’ perceptual data with that estimated from the younger adults in the one-parameter model, it produced a worse fit to older adults’ continuous reproduction data compared to the model with the age-group appropriate PSF, further supporting a link between age-related changes in perception and the precision of episodic memory.