After decades of extensively examining memory-loss patients, we still do not have a clear-cut understanding of how amnesia materializes. Due to differentiating theories, it has been particularly difficult to formulate a definitive model for declarative memory and as a result, describing the etiology of amnesia has been challenging. Squire (1991) proposes an associative model for declarative memory, made up of a unitary system. Others advocate a dissociative model; Aggleton and Brown (1999) propose that declarative memory is composed of separate systems with distinctive anatomical structures, which serve different functional purposes.
After giving a brief introduction to both models, this essay will concentrate on the conflicting evidence, limiting methodology, and lack of justifiable hypotheses within both models. This essay will demonstrate the difficulty in deciding which model best describes amnesia and therefore, propose the desperate need for evidence of double dissociations in amnesic patients in order to close the gaps in our biological understanding of amnesia.
The most famous amnesic patient, H.M, contributed to learnings in both models, highlighting the hippocampus as a crucial component in memory consolidation.
In addition, his pathology helped establish that the deterioration of declarative memory results in amnesia (Eichenbaum, 2013). Whilst these learnings are now widely accepted, the specific functions of different anatomical brain areas and how declarative memory is systematised within these regions still remains disputable. Using H.M’s results, Squire developed an associative model called the medial temporal lobe memory system (MTLMS). This system is made up of brain structures within the medial temporal lobe (MTL), including the hippocampus and adjacent structures; the entorhinal, perirhinal and parahippocampal cortices. Squire (1991) proposed that these structures work together to bind and encode declarative memories and then distribute them into the neocortex, noting that after time passes, these memories eventually become independent of the MTL. The structure of this system has important implications; specifically, that the larger the damage to the MTLMS, the bigger the deficit on declarative memory and thus, the greater the severity of amnesia. However, evidence later emerged demonstrating that amnesiacs could achieve relatively normal scores on familiarity tasks without the ability to recollect information, leading Aggleton & Brown (1999) to formulate an alternate model to Squire’s MTLMS. This insinuated that episodic and semantic memory are detached systems. Aggleton and Brown’s model specifies a dual process, proposing that the declarative memory system has two functional distinctions; recollection (in the hippocampal-anterior thalamic system) and familiarity (in the perirhinal cortex). Despite these differences, both models have an important similarity; they both acknowledge that the hippocampus performs a significant role in memory, particularly in recollection (Squire et al., 2007). That said, there is contention between both models with regards to how the brain organises declarative memory, making it difficult to fully comprehend amnesia’s neurobiological basis.
Whilst compelling evidence initially supports a dissociative model, there are theoretical issues with proving this dissociation between familiarity and recollection in episodic memory, making it difficult to judge with certainty that the dissociative model is the better explanation of amnesia’s neural framework. Both models argue over the functionality of individual MTLMS structures, specifically regarding the neural basis behind familiarity and recollection. Squire’s model supports the idea that hippocampal regions and the adjacent areas are essential for establishing and subsequently storing declarative memory (Zola-Morgan and Squire, 1993). However, other research indicates that the hippocampus may not play as important a role in recognition memory as previously thought (Aggleton and Shaw, 1996); it has been shown that subjects with hippocampal lesions, but without additional damage to the perirhinal cortex, can still recognise objects or places as familiar. Evidence for a binary system has been demonstrated using multiple methodologies; case studies of amnesic individuals, such as KN and YR, with impaired hippocampal but spared perirhinal cortices have demonstrated recollection deficits without impairments in familiarity tasks (Aggleton et al., 2005); receiver operating characteristics (ROC) curves have indicated that hippocampal lesions prevent recollection but not familiarity (Yonelinas et al, 1998; Fortin et al, 2004); and FMRI research has demonstrated that during familiarity tasks, there is activity in the perirhinal cortex but none in the hippocampus and vice versa for recall tasks (Yonelinas et al., 2001; Eldridge et al., 2000). Despite this, Squire (2007) rejects the interpretation of these results and establishes an alternative interpretation to these findings, suggesting that the results are due to recall tasks utilising far more cognitive energy in the MTLMS than the familiarity tasks. Since recognition is far easier to accomplish than recall, testing familiarity is likely to result in a ceiling effect. This is a reasonable explanation for the findings, which can not be contradicted by isolated examples of single dissociations. Moreover, other evidence has directly discredited the dissociations found; one study determined that patients with hypoxia with a specific hippocampal deficit were impaired in both familiarity and recall (Manns et al., 2003), demonstrating that dissociations within declarative memory cannot be replicated in every sample. Whilst the dissociative model gives compelling evidence, until proof of a double dissociation is discovered in human amnesics, there is not strong enough evidence to completely reject Squire’s claim that the findings are examples of strong and weak memories. This highlights the difficulty in disputing the accuracy of the associative model and therefore, it is not currently possible to conclude that a dissociative model is the most appropriate mechanism for describing amnesia.
Although recent animal studies have solved the double dissociation conundrum, it is still uncertain that the episode-like memory seen in animals is analogous to the human experience of episodic memory. Eacott & Gaffan (2005) found a double dissociation in rats, demonstrating that rats with hippocampal lesions were impaired on what-where-which tasks but scored normally on what-where or what-which tasks. The reverse effect was shown for the cortices outside the hippocampus. These findings were not due to some tasks being harder than others, as rats with post-rhinal lesions performed normally in what-where-which tasks, but were impaired in the simpler what-which tasks. This supports the existence of a dual process for familiarity and recollection and thus, helps us reject Squire’s interpretation. Eacott et al. (2005) even went on to demonstrate that what-where-which memory recall in rats is still present after completing unlearned tasks. Nevertheless, there are still multiple issues with studying animals, making it difficult to use their results to disprove Squire’s associative model. For instance, due to their lack of language, researchers can only observe the behaviour of animals and therefore, they have no way of actually knowing whether or not they possess autonoetic consciousness (an individual’s conscious experience of themselves as they recall past memories) (Tulving, 2002), a crucial component of episodic memory. Animals also do not possess identical brain structures to humans, therefore, extrapolating animal results in humans is unreliable. Although the findings from animals show a dissociation between what-which memory and episodic-like memory, as they are unable to self-report, they fail to determine a dissociation between episodic and semantic memory. For these reasons, using animal studies has severe limitations as it cannot be decisively concluded that the episodic-like memory observed in animals is the same as the episodic memory seen in humans. Therefore, it is difficult to label Aggleton and Brown’s (2001) model as the preferable theory, as Squire’s (1987) associative MLTMS cannot be explicitly rejected on the basis of results from animal studies.
As indicated by human and later animal studies, it has been difficult to determine whether or not a dissociation exists between semantic and episodic memory within declarative memory, ergo, declaring one model as superior is currently challenging. Squire’s associative model assumes that semantic and episodic memory are inextricably linked, however, alternative views suggest that episodic memory may not be needed to construct semantic memory (Parkin, 1982). Evidence supporting the latter perspective showing that amnesic patients can still collect semantic knowledge after using repetition techniques is plentiful (Tulving et al., 1991; Hayman et al., 1993). For example, after studying, patient K.C was able to acquire semantic knowledge regardless of his severely impaired episodic memory. On account of this, Tulving (1991) suggested that semantic memory can acquire new information independently of the MTL using alternative perceptual systems. However, Squire and Zola (1998) argue that these findings are not conclusive evidence for the sparing of semantic memory in amnesia. They point out that another amnesic patient called, E.P., like K.C., was also given training, however, he lacked the ability to learn new semantic information, supporting their theory that individuals cannot acquire semantic knowledge without a working episodic memory. Squire also raised concerns over the declaration that K.C.’s episodic memory was non-existent (Hayman et al, 1993), as insufficient documentation was recorded for K.C.’s recognition test results, instead just noting that K.C’s scores were comparable to those of other amnesic patients (Tulving et al, 1991). This is an odd comment, as it is rare for amnesic patients to have no working episodic memory whatsoever. Squire and Zola (1998) explained that K.C.’s severely impaired episodic memory is mainly due to his left frontal damage, rather than injury to the MTL, which explains why the MTL may still be functional and able to process semantic information. This theory is consistent with patient E.P, whose amnesia was due to extensive damage in the MTL rather than the frontal lobe, resulting in an inability to acquire both episodic and semantic information. Therefore, patient K.C provides inadequate evidence of a dissociation within the MTL, as his amnesia was primarily due to frontal lobe damage. Although the dual process model is a more viable hypothesis, these examples of single dissociations between episodic and semantic memory in amnesic patients are not capable of invalidating Squire’s associative model by themselves. Therefore, double dissociations in amnesiacs are needed before this conclusion can be widely accepted. As a result, without definitive evidence, it is too difficult to determine whether an associative or dissociative model better illustrates the neural basis behind amnesia.
Recent research on the basal forebrain and acetylcholine projections has further complicated the matter, intensifying the difficulty in determining the better model. The MTL and basal forebrain are interconnected by acetylcholine projections, which extrude two separate pathways; one onto the hippocampus and the other onto the amygdala (Gaffan, Parker & Easton, 2001). If both passages are disconnected, the animal will become acutely amnesic. However, if only one or the other is cut, the volume of errors on memory tests will reduce, demonstrating that declarative memory can be partly spared in the presence of some functional acetylcholine projections. Limiting the basal forebrain projections only results in anterograde amnesia, indicating their unique role in establishing new memories rather than preserving old information. Thus, it has been proposed that the cholinergic projections somehow play an essential role in developing novel memories. Interestingly, it was discovered that targeting the cholinergic cells in the basal forebrain with specialised immunotoxins resulted in identical impairments to disconnecting the white matter pathways (Turchi et al., 2005), indicating that anterograde amnesia is induced purely by disrupting the relationship between the acetylcholine projections in the basal forebrain and the MTL structures. However, this further complicates the debate over which model better elucidates amnesia, as patients with both basal forebrain and MTL damage cannot be used to support the existence of either model as it can not be concluded how much of the amnesia is due to basal forebrain damage or due to deficits in the MTLMS. As a result, basal forebrain damage acts as a confounding variable and therefore amnesic patients with this particular damage cannot be supporting evidence for either model. Therefore, the discovery of acetylcholine projections linking the basal forebrain and MTL contributes to the difficulty in demonstrating which model is a more valid characterisation of amnesia.
Working with amnesic patients has multiple issues, one of which is characterising the extent of the instigating trauma. A common problem with amnesic patients is that the lesions lack neatness and tend to encroach on other cortical areas and so, it is difficult to work out what behaviour is due to which brain damage. Research papers can also be problematic as they often contain either incorrect data or incomplete evidence. For example, none of the studies Zola-Morgan’s (1989) research paper includes coronal sections anterior enough to fully evaluate the status of the perirhinal cortex, which in turn creates a small gap in our knowledge. Without sufficient understanding of the damage, it severely decreases the chance of creating a reliable model for amnesia. Not only do research papers fail to include all the relevant information, but in the past, they have contained the wrong information. For instance, one of the most famous amnesic patients, H.M was often misleadingly labelled as the ideal case study for investigating amnesia. However, despite his fame, his surgery notes, which were reconstructed after the surgery, are not reliable. Although Scoville and Milner (1957) noted that 8cm of H.M’s hippocampus was taken out, MRI later showed that the back 2.5cm was still very much intact, meaning that the lesion size was only 4cm in reality. Subsequent research was influenced by this over-estimation, with Squire’s own theory relying upon the lesion sizes recorded in this paper. Thus, confidence in a model is shaken when it is shown to be dependent on incorrect data. For this reason, without ample information about cortical damage, it is too difficult to decide whether an associative or dissociative approach is a better explanation of amnesia, as incorrect information on amnesic patients may have led to miscalculations in both models.
Understanding why it is so difficult to establish whether an associative model, such as Squire’s, or a dissociative model, such as Aggleton and Brown’s, is better at describing amnesia is dependent on multiple determinants. Whilst findings in amnesic patients show a dissociation in familiarity and recollection, there is debate over the neural explanations behind this division, which can not be resolved until evidence has been found of double dissociations in human amnesic patients. Whilst animal research seemed to clear up the double dissociation controversy, episodic-like memory is not necessarily representative of human episodic memory and does not help elucidate the semantic/episodic memory split. Although there is evidence to support a dissociation of semantic and episodic memory in humans, other research contradicts these findings and so without the support of a double dissociation, the results are not decisive enough to downright reject Squire’s associative model. The discovery of acetylcholine projections has added another layer of complication, as in spite of their fundamental role in anterograde amnesia, they prevent many amnesic case studies from being used as evidence in support of either model. All of these factors, tied in with the challenge of experimentally manipulating amnesic patients, has made it increasingly difficult to determine which model is best at describing the neurological structure of amnesia. Therefore, whilst the dual process system, as proposed by Aggleton and Brown, is more compelling, further research should attempt to find double dissociations within human patients in order to help us conclude whether an associative or dissociative model represents amnesia with greater accuracy.