A Neural Basis for Episodic Memory Deficits in Older Adults
What's the science?
Episodic memory, in which an individual uses contextual details to recall an entire event (as opposed to simply recalling a fact) is disproportionately affected in older adults. Older adults have been found to engage in a process known as ‘hyper-binding’, in which they remember additional irrelevant features while encoding the relationship between an item and its context (forming an episodic memory). Older adults may have to work harder to recall information. However, we don’t know what is happening in the brain when the memory is being encoded in older adults. This week in NeuroImage, Powell and colleagues had young and older adults perform an encoding task while EEG was recorded in order to understand how aging affects episodic memory encoding.
How did they do it?
22 young (18-35) and 21 older (60-80 years) participated in the study. Over a series of mini blocks of trials within four larger blocks, participants were presented with 288 images of objects flanked by a coloured square on one side and a scene on the other side. During the memory encoding phase of the experiment, for each mini block, the ‘target context’ (i.e. the association between the object and either a color or scene) was manipulated: Participants were asked to judge either a) whether the color of the square was a color likely to be associated with the the object or b) whether the object was likely to be found in the scene. In sum, in each trial they were required to memorize either the color or scene associated with the object while the other acted as a distractor. During the test phase of the experiment, the 288 images were presented interspersed with 144 new images/objects. Participants were first asked to recall whether the images were new or old (whether they had seen them before), and, if they correctly guessed old, whether the colour and the scene matched the object as it had originally been presented. Sometimes, the target context matched the image, sometimes the distractor matched the image, and sometimes both matched the image.
Multi-voxel pattern analysis (MVPA; a pattern classification technique) was used to predict the target context of a given trial based on the pattern of brain activity during encoding. The data were divided into bins by frequency (3-80 Hz) and time during the trial (0-2000 ms) and the MVPA classifier was able to ‘learn’ which target context (colour vs. scene) was presented during the memory encoding period. The purpose of the classifier was to see whether EEG oscillatory power could predict selective attention to the target context feature (color or scene). A follow-up analysis/classification was performed in which data was binned into delta, theta, alpha, beta, and gamma frequency bands (different frequencies of brain activation measured by EEG). The relationship between classifier accuracy (correct identification of scene vs. colour as the target context during encoding) and participant accuracy of the target context and distractor during the test phase was explored.
What did they find?
Context memory was worse in older adults compared to young adults, suggesting hyper-binding of both the target and distractor during encoding. Older adults were also more likely to correctly identify the target context if the distractor context also matched the original image, further suggesting hyper-binding. The peak frequency (of brain activity) and time period in which the classifier (which uses brain activity to predict attention to target context) correctly identified the target context presented (colour vs. scene) were 2-20 Hz and 300-1200 ms respectively. When the classifier was run with the data divided into frequency bins, a positive relationship between beta band power and the correct identification of the target context was observed. A negative relationship between beta and alpha band power and correct identification of the distractor (i.e. lower beta and alpha band power, better distractor identification) were found. This indicates that the classifier ‘learned’ the brain activity pattern and used it to correctly identify the target context (color vs. scene) better on trials in which the individuals later recalled the target context correctly (demonstrating better selective attention), but had poor accuracy on trials for which the participant correctly identified the distractor. There was no relationship with age, which suggests that there were no differences between classifier performance across age groups. Electrodes located over the cortex contralateral to the location of the target context exhibited better performance. Therefore, a follow-up analysis was performed using only the contralateral electrodes: greater decline from 0-500 ms to 500-100 ms in classifier accuracy was associated with poorer target context accuracy and better distractor accuracy, suggesting poor selective attention and greater hyper-binding was associated with worse classification performance. This result was driven by the older adults group and could reflect a shift in attention away from target and towards distractor that may result in greater hyper-binding.
What's the impact?
This is the first study to examine age-related changes in memory encoding using a pattern classification of EEG oscillatory activity. Target context was predicted by alpha and beta power features of the classifier, which play roles in behavioural inhibition and attention respectively. Poor selective attention or poor inhibition of a distractor may underlie episodic memory deficits in older adults. These finding improve our understanding of episodic memory impairment associated with aging.
Powell et al., Decoding selective attention to context memory: An aging study. NeuroImage (2018). Access the original scientific publication here.
