Recognition memory shielded from semantic but not perceptual interference in normal aging

Abstract

Normal aging impairs long-term declarative memory, and evidence suggests that this impairment may be driven partly by structural or functional changes in the medial temporal lobe (MTL). Theories of MTL memory function therefore make predictions for age-related memory loss. One theory - the Representational-Hierarchical account - makes two specific predictions. First, recognition memory performance in older participants should be impaired by feature-level interference, in which studied items contain many shared, and thus repeatedly appearing, perceptual features. Second, if the interference in a recognition memory task - i.e., the information that repeats across items - resides at a higher level of complexity than simple perceptual features, such as semantic gist, older adults should be less impacted by such interference than young adults. We tested these predictions using the Deese-Roediger-McDermott paradigm, by creating feature-level (i.e., perceptual) interference with phonemically/orthographically related word categories, and higher-level associative interference with semantically related word categories. We manipulated category size in order to compare the effect of less versus more interference (i.e., small versus large category size), which served to (1) avoid potential item confounds arising from systematic differences between words belonging to perceptually- versus semantically-related categories, and (2) ensure that any effect of interference was due to information encoded at study, rather than pre-experimentally. Further, we used signal detection theory (SDT) to interpret our data, rather than examining false alarm (FA) rates in isolation. The d' measure derived from SDT avoids contamination of the memory measure by response bias, and lies on an interval scale, allowing memory performance in different conditions to be compared without violating assumptions of the statistical tests. Older participants were relatively more impaired by perceptual interference and less impaired by semantic interference than young adults. This pattern is at odds with many current theories of age-related memory loss, but is in line with the Representational-Hierarchical account.

Evidence for adult age-invariance in associative false recognition

“…Thus, they found that in this task age-related memory errors were driven more by perceptual than semantic relatedness. The results converge with those of another study that used intermixed visual word lists without distinctive fonts: Wilson, Potter, & Cowell (2018) also found no significant agerelated increase in false recognition of critical DRM lures, although memory was measured solely in terms of discrimination between studied items and lures (as in Fraundorf, Hourihan, Peters, & Benjamin, 2019).…”

“…We made two sets of alternative predictions. If Tun et al (1998) were correct that older people are more prone to gist memory than young adults when word lists are blocked, an age effect on false recognition would now be observed, unlike in both Burnside et al's (2017) experiments and (D. M. Wilson et al, 2018). Alternatively, if associative false recognition is age-invariant when there is no prior recall test, we would find Bayesian evidence in favour of the null hypothesis, like Burnside et al (2017).…”

“…In terms of theories of cognitive ageing, a finding that the age-related increase in false recognition does not extend to the DRM task would contradict the idea outlined above that older people rely more on prior knowledge. Reliance on prior knowledge also cannot explain errors due to visual perceptual similarity (Burnside et al, 2017;Pidgeon & Morcom, 2014) or phonological similarity (Watson, Balota, & Sergent-Marshall, 2001; D. M. Wilson et al, 2018).…”

“…An alternative view is that older people's memory representations are in some way less specific (Abdulrahman, Fletcher, Bullmore, & Morcom, 2014;Burke & Barnes, 2018;Reagh et al, 2018;Trelle, Henson, Green, & Simons, 2017;D. M. Wilson et al, 2018;I.…”

“…In the two experiments reported here, we aimed to close the gap between studies that have used procedures comparable to the categorised pictures task (e.g., Burnside et al, 2017;D. M. Wilson et al, 2018), and studies using a more typical DRM task (e.g., Kensinger & Schacter, 1999;McCabe & Smith, 2002;Tun et al, 1998).…”