Investigations into the neural basis of explicit memory began over a century ago. The temporal lobes were first suggested to play a critical role in memory in 1900 by Bekhterev [Bekhterev 1900 quoted in Kolb and Whishaw 1996]. In 1954 the importance of the medial temporal lobes was clearly demonstrated by the case of HM [Scoville 1954] who underwent bilateral temporal lobe resections for the relief of epilepsy [Corkin et a l 1997]. Following this surgery he suffered from severe anterograde amnesia [Scoville 1954]. Since then a wealth of clinical and experimental data have accumulated to demonstrate the importance of the medial temporal lobe structures in memory (e.g. see Milner 1970).
Different components of the medial temporal lobes are important for different aspects of explicit memory functions [Squire and Knowlton 2000]. This is indicated by results from lesion studies (in both humans and animals), functional imaging studies, and electrophysiological studies, as well as from the anatomy o f the medial temporal lobes. Although the exact roles o f the individual regions that are necessary and/or sufficient for different aspects of explicit memory are currently unresolved, the hippocampal formation is thought to be critical for episodic memory, whilst the parahippocampal region (particularly the perirhinal cortex) is hypothesised to support non-episodic memory, including familiarity judgements and simple associations [Mishkin et a l 1997].
3.1.3.1 The Parahippocampal Region And Non-Eplsodic Explicit Memories
Many research studies support the hypothesis that the parahippocampal region (but not the hippocampal formation) is necessary for many aspects of non-episodic explicit memories, including familiarity judgements and semantic knowledge.
Lesions to the parahippocampal region in monkeys lead to severe deficits in recognition memory [Horel et a l 1986; Zola-Morgan et a l 1989a; Meunier et a l 1993; Elliott et a l
1997]. As indicated earlier, this would suggest a deficit in familiarity-based recognition. The severity of impairment in recognition is the same even when the lesion is restricted to only the perirhinal cortex [Meunier et a l 1993], underlining the particular role o f this region in recognition. In contrast, damage to the hippocampal formation alone usually results in a much milder (and in some cases no) recognition deficit [Gaffan et a l 1984; Bachevalier et a l 1985; Zola-Morgan et a l 1989b; Murray and Mishkin 1998] suggesting that familiarity-based recognition does not require the hippocampal formation. However, it should be noted that Squire and colleagues have found impaired recognition following hippocampal damage in monkeys [Zola et a l 2000]. It is not understood why such different results have been obtained. Methodological variations (such as removal of monkeys from the apparatus between presentation and test, and differences in lesion extent) may play a role.
The effects of bilateral lesions of the parahippocampal regions in humans have not been reported, as such selective damage rarely occurs. However two patient groups (patients with semantic dementia and patients with lesions encompassing both the parahippocampal region and the hippocampal formation) offer some insights into the neural basis of non- episodic explicit memory.
Semantic dementia is a progressive neurodegenerative disease. The early cognitive profile o f individuals with semantic dementia is a selective loss of semantic knowledge [Hodges et a l 1992]. Performance on recognition tasks is normal (unpublished data reported in
Graham et a l 1999; Simons et a l 2000). The underlying neuropathology of semantic dementia is the degeneration of the cortical, inferior and possibly ventral portions o f the temporal lobes. Although initially thought to be selectively affecting inferolateral temporal cortex with sparing o f the hippocampal formation [Hodges et a l 1992], recent studies suggest that this degeneration also affects the hippocampal formation, even in the early stages o f the disease [Gabon et a l 2000].
The complex nature of a neurodegenerative process means that the functions of the areas affected by the disease need not necessarily be totally disrupted. Similarly, the integrity of areas apparently unaffected can not be assumed. Interpretation of the implications of the profile of semantic dementia with regards to the anatomy must therefore be guarded. With this caveat in mind, two important points should be made. In drawing comparisons with patients with Alzheimer’s disease, it might be argued that the parahippocampal region is involved in semantic knowledge. Patients with Alzheimer’s disease show selective degeneration o f the hippocampal formation, and the typical cognitive profile includes grossly intact semantic knowledge [Hodges and Patterson 1995]. Secondly, the intact recognition skills o f individuals with semantic dementia underlines the possibility that recognition and semantic knowledge are independently supported, but should not be taken as evidence that recognition is independent o f the medial temporal lobes.
The second type of patients who offer some insight into the neural underpinnings o f non- episodic explicit memory are those with lesions to the hippocampal formation and surrounding cortex. These individuals have been reported to have significantly worse recognition memory than individuals with selective hippocampal damage [Aggleton and Shaw 1996; Reed et a l 1997; Buffalo et a l 1998]. This finding is consistent with the parahippocampal region supporting recognition. However it might also be a reflection of the compounding effects o f lesions to the parahippocampal region and the hippocampal formation. Selective damage to either may (at least in theory) leave recognition intact, whilst damage to both produces a recognition impairment.
Perirhinal neurons show encoding of the necessary information to allow solution of a wide range of recognition tasks not requiring spatial and contextual discriminations (see Brown and Xiang 1998). In contrast, no other area in the medial temporal lobe has such extensive encoding capacity. For example, the cells in the hippocampal formation show no evidence o f specific stimulus encoding during recognition tasks [Brown et a l 1987; Sakurai 1990;
Riches et a l 1991].
Many functional imaging studies of recognition per se report both hippocampal formation and parahippocampal region activation (e.g. Stem et a l 1996; Stark and Squire 2000). This might be taken as an indication that the hippocampal formation is involved in familiarity-based judgements. However, successful performance on recognition tasks in normal controls is likely to be supported by both recollection and familiarity. Recently the ‘remember-know’ paradigm has been used in imaging studies to show that hippocampal activation is only associated with recollection, and not familiarity judgements [Eldridge et a l 2000]. These findings therefore support the role o f the parahippocampal region in familiarity judgements.
In summary, research findings are consistent with the hypothesis that the parahippocampal region plays an important role in non-episodic explicit memory. At the present time, however, the necessity of the parahippocampal region for non-episodic explicit memories has not been conclusively demonstrated.
3.1.3.2 The Hippocampal Formation is Necessary For Contextual Memory
There is a wealth of evidence implicating the hippocampal formation in contextual memory. Many different techniques have shown that retrieval o f spatial and temporal information (critical elements in contextual memory) is highly dependent on the integrity of this structure.
Damage to the hippocampal formation results in deficits in association of events across time. This is clearly demonstrated by the differential effects o f hippocampal lesions on two
different types of conditioning. In delayed conditioning, a tone (CS) precedes and overlaps with an unconditioned stimulus (UCS) that evokes a reflexive eyeblink [Pavlov 1927]. The subject learns to blink during the presentation o f the CS. In trace conditioning, the CS onset is abbreviated so there is a short (-500 ms) delay between the end o f the CS and onset of the UCS [Pavlov 1927], thus introducing into trace conditioning a temporal context. Lesions to the hippocampal formation in animals interrupt trace but not delay conditioning (animals: Solomon et a l 1986; Moyer et a l 1990; Kim et a l 1995; McEchron et al
1998; humans: McGlinchey-Berroth et a l 1997; Clark and Squire 1998).
Memory for spatial information is also impaired in animals with lesions to the hippocampal formation. Rats have difficulty finding the platform in the Morris Water Maze when the cues are spatially distributed throughout the maze [Save and Poucet 2000]. Information about spatial arrays (but not individual object recognition) is also disrupted in monkeys with hippocampal damage [Gaffan 1994].
Animals with hippocampal formation lesions process contextual information abnormally, showing unusual dependence on irrelevant contexts and inefficient processing of relevant contexts. Firstly, they tend to process irrelevant or non-predictive information more than controls do. For example, in control animals, the strength of the conditioning to the background is predicted by the reliability with which the background predicts the unconditioned stimulus. However, rats with hippocampal lesions condition strongly to context, even when the existence of an explicit conditioned stimulus precludes context conditioning in the control rats [Winocur et a l 1987].
Animals with hippocampal lesions also have difficulty learning new tasks in the same context as a previously learnt task [Winocur and Gilbert 1984], and performing a previously learnt task in a new context [Winocur and Olds 1978]. Gaffan reported that monkeys with fomix transection were able to perform an object discrimination learning task with varying backgrounds, but were unable to perform the same task with a unique background [Gaffan 1994]. This suggests further that abnormal contextual processing can follow disruption of the hippocampal formation.
Animals with hippocampal lesions also process relevant contextual cues less efficiently. Good and Honey showed that animals with hippocampal lesions failed to learn that a cue in context A signalled a reward, whilst the same cue in context B signalled no reward [Good and Honey 1991].
Cellular recordings have demonstrated that there are cells in the hippocampal formation that encode spatial location (so called ‘place’ cells) (e.g. O'Keefe 1976). A number of cellular protein changes in the hippocampal formation have been associated with performance on contextual memory tasks in animals (e.g. Atkins et a l 1998; Woolf et a l
1999). Such contextual learning paradigms have also been shown to increase neuronal cell generation in the adult animal’s dentate gyrus. This effect is not seen in non-contextual learning paradigms (such as delay conditioning) [Gould et a l 1999].
Individuals with damage to the hippocampal formation are unable to remember episodic information, but may learn at least some semantic information. In particular, individuals with developmental hippocampal atrophy have been shown to have a selective episodic impairment with relatively intact semantic memory [Vargha-Khadem et a l 1997a]. One individual with such damage has been shown to have intact electrophysiological markers of familiarity but not recollection [Duzel et a l 2001]. A number of functional imaging studies have shown that contextual memory retrieval is associated with hippocampal formation activation (e.g. Eldridge et a l 2000; Burgess et a l 2001).