The work sits alongside research on memory and aging: brain age maps, the influence of white matter on cognition, neural network models for disease recognition. It is already known that older adults more often remember meaning but are worse at details of time and place. This paper tests exactly how this reliance on meaning generates false memories in the standard Deese-Roediger-McDermott paradigm. It is not just another confirmation of the semantic shift, but an attempt to separate the influence of two axes of overlap: semantic and temporal.
Participants (49 young adults aged 18-35 and 41 older adults aged 60-85) studied 18 lists of 16 words each. Words were selected from the Toronto Noun Pool so that within a list they formed groups of semantic associates (for example, 'actor', 'film', 'screen', 'cinema'). Temporal overlap was varied: associates appeared either consecutively or separated by four other words. After studying, participants completed a recognition test of 216 trials: half were old words, half were critical lures that had not been presented but were semantically related. Responses were given on a four-point confidence scale. The analysis included mixed and repeated-measures ANOVAs, as well as Bayes factors to compare models with and without interactions.
False memories were higher in older adults (30%) compared to young adults (22%), but temporal overlap did not affect them in either group. Strong semantic overlap increased false memories in all participants (26% vs 22%). High-confidence false memories grew under any strong overlap: semantic (15% vs 10%) and temporal (14% vs 11%). In recognition of real words, an age difference emerged: young adults recognized words better under strong semantic and strong temporal overlap (74% vs 67%), while older adults did so under strong semantic but weak temporal overlap (74% vs 65%). Memory discriminability (d') did not differ between groups, which rules out a simple liberal response bias in older adults.
The work differs from its neighbors in that it manipulates two axes of overlap simultaneously and measures confidence. This allows the claim that the age-related shift toward semantics does not simply 'spoil' memory but changes the strategy for using information. For young adults, temporal proximity serves as a support; for older adults, it does not. Such a dissociation refines models of cognitive aging and may help in diagnosing early memory impairments, where false recognitions are a marker.
The topic could develop toward neuroimaging: which brain networks support switching between semantic and temporal support at different ages. It would also be promising to study whether training attention to details can reduce false memories in older adults. A separate question is how temporal overlap works over longer intervals, for example between lists rather than within one.
The results will affect geropsychology, cognitive neuroscience of aging, and clinical memory assessment. They are also useful for developing cognitive training and interfaces that take age-related memory strategies into account.
Repeat the experiment with more distant temporal separation of associates and with brain activity recording. Test whether the strategy changes when instructions direct attention to word order.
The work connects to a long-standing question of why episodic memory suffers with aging while semantic memory is preserved. It shows that false memories arise not from a general weakening of memory but from a shift of attention toward meaning. This refines the compensatory hypothesis of aging and leads to the problem of dedifferentiation of neural representations.
Participants scoring below 26 on the Montreal Cognitive Assessment were excluded, so the sample was cognitively healthy. Even in such people, false memories reached 30% under some conditions.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.