In my 60s I can attest to my weakened ability to recall. It is ridiculous. This post looks at a paper that is written most prominently by the authors of fuzzy trace theory, Brainerd and Reyna. “Dual-Retrieval Models and Neurocognitive Impairment” appeared online on August 26, 2013 in the Journal of Experimental Psychology. (The post also uses an online source, The Cornell Chronicle, in an article dated September 5, 2013, entitled: “Breakthrough discerns normal memory loss from disease”, and was written by Karene Booker.) It comes up with some interesting conclusions.
Brainerd’s dual-retrieval model of recall posits that items are recalled via a recollective operation, called direct access, and a nonrecollective one, called reconstruction. Reconstruction is accompanied by a slave judgment operation that evaluates the familiarity of reconstructed items before outputting them. (It is a “slave” operation in the sense that it is not activated unless reconstruction is successful.) The recollective operation accesses verbatim traces of list items’ prior presentations directly, without searching through traces of other items, and is therefore the faster of the two forms of retrieval. Direct access is also more accurate than reconstruction because it produces errorless recall: When an item’s verbatim trace is directly accessed, its surface form is symbolically reinstated, so that the item can be recalled by simply reading it out of consciousness. Direct access is a recollective operation because it reinstates vivid, realistic details of prior presentations.
The nonrecollective operation, reconstruction, regenerates items from stable episodic traces of partial-identifying information, especially semantic information (e.g., “animal” and “farm” for horse). This is the tip-of-the-tongue and feeling of-knowing phenomena that old guys like me experience way too much. According to the model we have been found to access a range of partial-identifying information about list items before they can be recalled. Reconstruction searches for items that match partial identifying features and generates sets of candidate items (e.g., horse, cow, goat, sheep) that are small enough to be processed within the time constraints of a recall test. As the features that generate such sets do not uniquely identify studied targets, the sets normally include nontargets (cow, goat, sheep). A judgment operation performs familiarity checks on reconstructed items before outputting them, which is how the dual-retrieval model implements the familiarity notion of dual process conceptions of recognition.
The authors wondered if they could look at memory test results and distinguish between or predict which people would progress from a healthy elderly memory to mild cognitive impairment and on to Alzheimer’s disease. They designed three experiments using the results of 10 to 15 minute recall tests of people diagnosed as healthy, mildly cognitively impaired, and with Alzheimer’s. These recall tests are in wide use and often given to individuals at intervals to test progression. Specifically, they used data from two longitudinal studies of older adults – a nationally representative sample of older adults, the Aging, Demographics and Memory Study, and the Alzheimer’s Disease Neuroimaging Initiative.
The authors’ hypothesis was that there are qualitative shifts in the processes that are responsible for declines during healthy aging versus MCI and AD. That hypothesis is predicated on the following considerations. Before healthy adults reach age 70 and the MCI conversion rate accelerates, cumulative declines in traditional measures of recollection are already quite large. Let me expand on this. The authors seem to be saying that normal people have lost so much direct access recall ability by the time they are 70 compared to when they are 30 that it is hard to differentiate between healthy elderly and diseased elderly based on that measure alone. This leaves nonrecollective processes, such as familiarity or reconstruction, as remaining possibilities.
The focus is on the retrieval processes that are measured by the model’s parameters, which are defined in Table 1. There are separate parameters that measure recollective retrieval (D), reconstructive retrieval (R), and familiarity judgment (J). These parameters do not map with simple, observable features of recall performance; that is, there are no observable aspects of performance that one can point to and say that they are uniquely due to one of the parameters. The memory abilities affected by cognitive impairment differ from those affected by healthy aging, the authors say, resulting in unique error patterns on neuropsychological tests of memory. Their theory-driven mathematical model detects these patterns by analyzing performance on such tests and measuring the separate memory processes used.
The results of the experiments/studies were that the model could predict progression from healthy to MCI to AD about 2 times in 3 to overgeneralize–but far better than chance and better than the best genetic markers. These enabled several findings.
- Mathematical models of memory can extract reliable measurements of underlying retrieval processes from simple instruments that are administered every day in clinics worldwide. This could create a reliable method to distinguish memory declines associated with healthy aging from the more-serious memory disorders years before obvious symptoms emerge.
- The notion that memory declines continuously throughout adulthood appears to be incorrect, they say. “When we separated out the cognitively impaired individuals, we found no evidence of further memory declines after the age of 69 in samples of nationally representative older adults and highly educated older adults,” said Reyna. Specifically, the researchers found that declines in reconstructive memory (recalling a word or event by piecing it together from clues about its meaning, for example, recalling that “dog” was presented in a word list by first remembering that household pets were presented in the list) were associated with mild cognitive impairment and Alzheimer’s dementia, but not with healthy aging. Declines in recollective memory – recalling a word or event exactly – were a feature of normal aging. “Reconstructive memory is very stable in healthy individuals, so declines in this type of memory are a hallmark of neurocognitive impairment,” Reyna said. The authors do note the caveat should be added that these results are cross-sectional, and, hence, they may be contaminated by selective survival effects. It may be that subjects who survive to older and older ages are progressively healthier, on average, than age-mates who do not survive and that this is masking age declines that would be detected with longitudinal comparisons. In other words, using these data bases they did not follow individuals from age 70 to 90 to be able to specifically measure reconstructive memory over time.
- Younger adults rely heavily on recollection, Brainerd said, but this method becomes increasingly inefficient throughout mid-adulthood. “Training people how to make better use of reconstructive recall as they age should assist healthy adult memory function,” he said.