About a month ago I had an interesting discussion with Prof. Postma at Utrecht University. During the course of the discussion he recommended a PhD thesis by Olga Meulenbroek on “Neural Correlates of Episodic Memory in healthy aging and Alzheimer’s Disease“. The thesis is a mixture of behavioural experiments and fMRI studies, comparing different groups of adults. There’s loads of really interesting stuff in it and quite a lot of it is beyond what I need. Nevertheless, there was much to learn from. (N.B. the references below are taken directly from Meulenbroek’s thesis, which can be obtained via the link above.)
To recap: the standard, functional model of memory breaks down long-term memory firstly into declarative and non-declarative memory. That is memory which can and cannot be spoken. As other authors have shown just because something can be spoken, doesn’t mean that it is purely linguistic. Spatial memory, which is one of the Prof Postma’s interests and which I’ll come back to, is a bit of an anomaly in this regard as it doesn’t usually feature in the standard model and I’m not sure where to put it: I err towards declarative memory as it can be spoken even though it’s encoding seems to involve a range of modalities.
Anyway… declarative memory is further broken down into semantic memory (knowledge and facts) and episodic memory (personal memory of past events) with autobiographical memory often presented as a subset of episodic. Though, as Meulenbroek reveals this division between episodic and semantic is not as hard and fast as I had thought.
The process of creating episodic memory involves 3 main steps:
1) Encoding – Processing at the moment an event is experienced (memory acquisition), that renders it accessible for future retrieval (Brown and Craik 2000).
2) Consolidation – Post-acquisition stabilisation of initially fragile memory traces, ranging from the molecular/cellular level (cellular consolidation) to the regional level (systems consolidation) and from minutes to months (Dudai 2002, Dudai 2004).
3) Retrieval – Accessing of stored information by recognition (familiarity) or recall (recollection) (Baddeley 2002)”
The standard model often correlates each type of memory with a particular region of the brain. However, as Meulenbroek points out, “[the] complex nature of episodic memory is reflected by the fact that it is supported by many brain structures. However, there are three regions that are central to episodic memory, as lesions in these areas cause amnesic symptoms. These are the medial temporal lobe (MTL) with the hippocampus at its core, the prefrontal cortex (PFC) and the retrosplenial cortex (RSC)” (p12)
“The MTL consists of the hippocampus and adjacent perirhinal, entorhinal and parahippocampal cortices (for review, see Squire et al. 2004). Sensory information flows through associative areas to perirhinal and parahippocampal cortex to the entorhinal cortex, and from there to the hippocampus.” Meulenbroek focuses her attention on the hippocampus.
“The hippocampus is where the highest level of integration (or abstraction) is achieved. Therefore, the hippocampus is able to facilitate the instantaneous storage of the “what”, “where” and “when” of an event. In other words, the hippocampus is responsible for the uniqueness of episodic memories. Finally, the hippocampus is regarded to serve as an index: it links and integrates information (memory traces) in the form of distributed neocortical representations (Marr 1971, Teyler and DiScenna 1986, Moscovitch et al. 2005, Moscovitch et al. 2006).” (p13-14)
“This view has been described in more detail in a recent update of the “Hippocampal-Neocortical Interactions Theory” (Wang and Morris 2010), which states that the hippocampus does not work in isolation, but works together with cortical networks, where it is believed that long-term memory traces are stored (Osada et al. 2008).” (p14)
This point is really helpful. One of the uncertainties associated with the hypothesis that thoughts and memories are supra-modal was that it was unclear to me whether the memory trace was stored in one location with links to the cortices which dealt with each modality or whether the different elements of the memory were distributed across the different parts of the brain with some central part co-ordinating the memories. I need to read further (as ever) but these comments point towards the latter model.
As with the hippocampus in the Medial Temporal Lobes, Meulenbroek focuses her attention on one region within the prefrontal cortex, the (ventro) medial prefrontal cortex (vmPFC). “[The vmPFC] is connected to many areas, including (somato)sensory areas, the amygdala, entorhinal cortex and hippocampus (for review, see Price 2006). These connections point to an integrative nature like the hippocampus.”
“The role of the vmPFC is still a topic of debate and there are many hypotheses […] it is a large heterogeneous area that probably hosts several functional subregions. There are two hypotheses particularly focused on memory:  The post-retrieval monitoring [hypothesis] (also known as the “Feeling of Rightness”) (Elliott et al. 2000, Milner and Petrides 1984, Moscovitch and Winocur 2002) states that the vmPFC monitors the information retrieved and tests its veracity.  The second hypothesis states that the vmPFC takes over (with time, or consolidation) the function of the hippocampus in linking and integrating distributed neocortical representations.” (p14-15)
The third region of interest is the Retrosplenial Cortex (RSC), which “supports spatial memory, like the detection of novel spatial arrangements of objects (e.g. Vann and Aggleton 2002).” It may also have a role in navigation and spatial memory in humans; “it is hypothesised that the RSC serves as a short-term buffer for the translation between allocentric and egocentric representations.”
“Besides spatial memory deficits, patients [with damage to the RSC] also demonstrate impairments at acquiring information presented verbally or visually (Maguire 2001), and they have difficulty retrieving recent, but not very remote, autobiographical memories (Maguire 2001, Osawa et al. 2006, Valenstein et al. 1987). In addition, activation of the RSC is commonly observed in autobiographical memory tasks (Svoboda et al. 2006). Furthermore, activation of the RSC is related to retrieval success (Buckner and Wheeler 2001, Rugg et al. 2002).” (p16)
Meulenbroek then describes the effects of aging and Alzheimer’s disease on these different regions in the brain (p17 – 21). Meulenbroek remains open to possibility of compensatory processes within the brain. That is, as it adapts to the process of the disease and reducing the observed rate of decline in behavioural performance, at least, until the progress of the disease undermines these compensatory processes as well. These processes are one of the areas that she investigates within her thesis.
Regarding route encoding and recognition in young adults and healthy elderly – Meulenbroek’s main conclusion is that “the RSC seems to be a region that can exhibit compensatory processing in healthy aging during spatial memory.
She goes on to speculate that, given the structural and metabolic damage of the RSC [caused by Alzheimer’s Disease] and co-occurrence of disconnection with the hippocampus […], it seems unlikely that this area would display compensatory processing, but this remains to be investigated.” (p145)
Meulenbroek also found compensatory processes at work in healthy older people where young and elderly adults had to remember the location of objects on a grid: “On the brain level, the enrichment of the encoding structure resulted in specific differences between young and elderly subjects. While in young it results in the use of imagery during recall (see De Rover, et al. 2008), the elderly engage their declarative memory system” . Meulenbroek observed other factors at play. Notably elderly participants “difficulty suppressing task irrelevant input (such as scanner noise)” (p73-74)
In Chapter 4, Meulenbroek looked at autobiographical memory (AM). In contrast to the hard division between semantic and episodic memory noted earlier, Meulenbroek points out that autobiographical memory can contain both semantic and episodic elements. “The classical distinction of declarative memory into episodic and semantic elements (Tulving 1972) closely approximates AM content: semantic elements represent facts about the world and our life, unrelated to specific events (lacking contextual details). Conversely, episodic elements are unique and have associated contextual details, allowing for subjective re-experiencing («mental time travel», Tulving 2002). Semantic elements can probably guide the search to episodic elements, making recall of AM an iterative, hierarchical process, with left-lateralised search processes (supported by PFC) followed by recollection (hippocampus, RSC), subjective re-experiencing (supported by occipital areas) and self-referential processing (medial PFC) (Cabeza and St Jacques 2007, Conway, et al. 2002, Daselaar et al. 2008).”
One interesting feature that Meulenbroek draws out of the literature is that “with aging, more semantic elements are reported per probed autobiographical memory, while the amount of episodic elements decreases, but the total number of details remains unaffected (Levine et al. 2002). Therefore, AM seems to “semanticize” with age, which could be related to faster decline of episodic than semantic memories (Piolino et al. 2002). In other words, semantic retrieval might compensate for episodic retrieval failure. Thus, it is not surprising that the episodic-to-semantic shift of AM becomes amplified with memory impairment, like amnestic Mild Cognitive Impairment (aMCI) (Murphy et al. 2008) and is probably further pronounced in Alzheimer’s disease, but studies are lacking.” (p85)
“[The] hippocampus is a core structure in AM retrieval, and its activation enhances with increasing vividness, independent of recency (Addis et al. 2004). This is consistent with the multiple trace theory of consolidation (MTT), stating that retrieval is accompanied by sustained hippocampal dependence of episodic elements like vividness (Moscovitch et al. 2006). Structural hippocampal degradation, a hallmark of Alzheimer’s disease (Blennow et al. 2006, Braak et al. 1999), is therefore bound to have functional consequences on AM retrieval. According to MTT, AM will be less dependent on the hippocampus when it contains fewer episodic elements, like with semantisation. Consequently, AM putatively becomes protected from the hippocampal damage in Alzheimer’s disease.” (p86)
Her experiments looked closely at this episodic to semantic shift and sought to find neural correlates. Her studies showed that “AM in Alzheimer’s patients has undergone an episodic-to-semantic content shift compared to healthy elderly.” The neural correlates of recall changes commensurately, with increased activation of (amongst other areas) the vmPFC. However, “when probed with semantic or autobiographical statements during fMRI, healthy elderly outperformed the patients, corroborating the decline of autobiographical and semantic memory in Alzheimer’s disease.” (p100-102) This indicated that although compensatory processes were at play, they were not entirely successful in maintaining performance.
This is an interesting idea and one that might be utilised experimentally. Might it be possible to devise a way of carrying out an interview before a reminiscence session, run the session and then analyse the contents of a participants conversation for the ratio of episodic to semantic elements?
The discussion in pages 100-102 is nuanced, really interesting and more than I want to put in this post. The role of the vmPFC becomes important to Meulenbroek’s argument. Alongside this connection between the vmPFC and the semantisation of memories (noted above), she also links the vmPFC to the integration and retrieval of long term memories (see the studies cited on p113). Thus the relative decline of the hippocampus and the relative integrity of the vmPFC correlate well with the oft-mentioned observation that people with Alzheimers’ disease often lose their ability to recall recent memories but can recall older memories (p112). However, integration of memories involves the interaction of the hippocampus and the vmPFC. Thus the decline of the hippocampus could potentially affect the interaction between the two, which would have consequence for both the integration of memory and for the capacity of the vmPFC to take on a compensatory role. These are the ideas that Meulenbroek pursues in Chapter 5 (p112ff).
“The results of [the study in Chapter 5] suggest that reduced hippocampal integrity in AD results in a functional vmPFC impairment, potentially by disconnection of these structures. Therefore, the vmPFC appears unable to complement the hippocampus during retrieval of recently acquired information.”
“The main difference between the two memory paradigms is that the [study presented in Chapter 5] probed recent memories, while the [study in Chapter 4] probed remote or well-consolidated memories typically acquired before disease onset. This dissociation suggests and supports the hypothesis that interaction between the hippocampus and mPFC is beneficial to acquisition and consolidation of new information (van Kesteren et al. 2010) and that this interaction is no longer crucial during retrieval of consolidated memories (Meulenbroek et al. 2010, Takashima et al. 2006).” (p129)
So, in summary…
The brain exhibits compensatory processes which help to reduce the decline in performance of certain tasks. However, these processes are dependent on the integrity of connections between different parts of the brain. Once these connections are damaged or broken then the compensatory processes are inhibited. This suggests that there are limits to what can be done to strengthen aspects of autobiographical memory through the course of dementia.
Also, the fMRI studies revealing the correlates of different types of memory correspond well with the often observed characteristic that people with dementia can remember the distant past better than the recent past. I’m not sure how knowing that this is because the vmPFC remains relatively intact during early stages of dementia helps but I like understanding it better.
The course of dementia effects the balance of semantic and episodic elements in recalled autobiographical memory. This does seem like something we can take hold of and make use of practically.