Long term memory and consolidation

Long term memory is information maintained over a significant period of time. This period of time is debatable but generally thought to be of the order of minutes to hours rather than seconds. Alternatively, material in long term memory can be considered as the material that can be recollected following distraction (Squire, 1986). It can be split according to the types of information stored into episodic memory (personal recollection of events in our lives) and semantic memory

(knowledge of facts and concepts that are learnt, rather than experienced) (Tulving, 1972).

Memory is not fixed from the point of learning but over time continues to stabilise (Squire, 1986). The idea of consolidation has been around for over a century, since it was first proposed by Muller and Pilzecker ((Müller & Pilzecker, 1900) cited in (McGaugh, 2000)). Finding that newly learned information was disrupted by learning other information shortly after, they proposed that the processes that underlie the formation of new memories are fragile and consolidate over time. This hypothesis was supported by Duncan‟s research in 1949 that showed rats given an electric shock to the head shortly after learning a maze had much poorer memory for the maze than those who had no electric shock. Rats having shocks at longer delays had progressively better memory for the maze, with rats with a shock 1 hour or more after learning had equal memory to controls. This demonstrated that there may be a period after learning where the material learnt is susceptible to disruption, for example by electro-convulsive shock, but the material becomes more stable over time (Duncan, 1949).

Declarative memory relies on the hippocampus and medial temporal lobe

over time retrieval is thought to become independent of the hippocampus (Kapur & Brooks, 1999). However, when and which memories become independent of the hippocampus is an issue of debate, as will be discussed.

The standard model of consolidation suggests that the hippocampus has a time- limited role, being vital for early storage and retrieval followed by a gradual reorganisation process whereby information is transferred to neocortical networks (Sutherland & McNaughton, 2000). A short-term consolidation process, lasting seconds to minutes, binds information into a memory trace, followed by a long- term consolidation process to stabilise it (Nadel & Moscovitch, 1997). The exact mechanism by which this happens is unclear, but Squire suggests that an important concept is that „information in the medial temporal lobe directs consolidation by gradually changing the organisation of cortical representations, for example,

strengthening connections between the cortical sites that participate in representing a memory‟ (Squire & Alvarez, 1995). Hippocampal synapses can change quickly, so the hippocampus acts as a short term temporary memory store, while neocortical synapses change slowly. If the hippocampal system repeatedly reactivates

representations in the neocortex, strong interconnections between cortical sites form, able to support the memory independent of the hippocampus (Squire & Alvarez, 1995). According to this model, with a purely hippocampal lesion there would be temporally-limited retrograde amnesia, whereas more extensive damage to the temporal neocortex would result in more extensive retrograde amnesia (Squire & Alvarez, 1995). For example, a patient with hippocampal damage might have amnesia for a period of time before the damage, while the memories were still being consolidated, but have intact memory for events longer ago which had already been fully established in the neocortex independent of the hippocampus.

The Multiple Memory Trace theory (MMT) is another suggestion, which involves the hippocampus having a lifetime role in the retrieval of episodic autobiographic memories (Nadel & Moscovitch, 1997; Nadel et al., 2000). This is that the

damage (Nadel et al., 2000). As memory is created, a code in the hippocampus binds information stored in other brain regions to create a memory of a specific episode or scene, so interaction between the hippocampus and the other brain regions

involved is required indefinitely. However, semantic memories are thought to be dependent on the hippocampus while they are consolidated and can later become independent (Nadel et al., 2000), demonstrating a crucial distinction between episodic and semantic memories.

The cellular mechanism by which consolidation occurs is thought to be LTP (long term potentiation) (Squire & Alvarez, 1995). This is a form of synaptic plasticity, which involves the persistent enhancement of signal transmission between two neurons by their repeated synchronous stimulation (Cooke & Bliss, 2006). This follows the principles of Hebb‟s Law, that „if a synapse is active when a post- synaptic neuron is active, the synapse will be strengthened‟ (Hebb, 1949; Gazzaniga et al., 2009). Long term potentiation is thought to be a good model for memory because of its longevity, both processes require protein synthesis, and potentiation is input specific, so a single pathway can be potentiated without impacting other connections to that neuron, increasing the information coding capacity of the brain. Also, association means that a weak stimulus can combine with a strong stimulus, or other weak stimuli, to become potentiated, providing a mechanism for

associating events or entities in our learning (Cooke & Bliss, 2006).

Consolidation does not merely provide a strengthening of memory traces, but also the opportunity for the integration of new memories with existing knowledge networks, so that they are accessible for delayed retrieval (Diekelmann et al., 2009). It also appears to provide an opportunity for experience and emotion to modulate the strength of our memories, for example by the interaction of stress hormones like adrenaline and cortisol released in states of arousal (Gold & Van Buskirk, 1975; Sandi & Rose, 1994; Conrad et al., 1997; McGaugh, 2000).

Sleep is thought to be important in the process of consolidation, and it has been clearly demonstrated that sleep after learning improves declarative memory performance, even when the confounding effects of circadian rhythms and fatigue at recall testing were removed (Gais et al., 2006; Drosopoulos et al., 2007), by

stabilising memory traces and providing increased resistance to interference during consolidation (Ellenbogen et al., 2006). However, there are also suggestions that sleep‟s role in consolidation might be more active, for instance by restructuring brain activity (Orban et al., 2006) and through hormonal changes during sleep (Born & Wagner, 2009).

Another aspect of long term declarative memory is recognition memory, the ability to recognise something that has previously been encountered. This is a matching process, comparing the content of the environment with the content of a memory. Recognition is commonly split into two domains: recollection – remembering details about an experience, and familiarity – awareness that something has been

encountered before but with no further knowledge about it (Eichenbaum et al., 2007). Localising these separate processes to areas of the brain has been attempted, and it is thought that they rely on separate but interlinked structures. In functional neuroimaging studies (Yonelinas et al., 2005; Diana et al., 2007) it has been found that regions in the pre-frontal, parietal and medial temporal cortices interact to provide recognition memory. The hippocampus seems to be crucial for recollection but familiarity is more associated with the peri-rhinal cortex (Montaldi et al., 2006; Diana et al., 2007; Eichenbaum et al., 2007). Within the structures that support recognition there are some very specific areas that fulfil a very specific role, for example the „fusiform face area‟ in the fusiform gyrus of the inferior temporal lobe which is associated with face recognition, and damage to this area leads to

prosopagnosia (failure to recognise faces)(Kanwisher, 2000).