Representation and Association in Memory: A Neurocomputational View of Hippocampal Function

Deficits in hippocampal-dependent transfer generalization learning accompany synaptic dysfunction in a mouse model of amyloidosis

Elevated β-amyloid and impaired synaptic function in hippocampus are among the earliest manifestations of Alzheimer's disease (AD). Most cognitive assessments employed in both humans and animal models, however, are insensitive to this early disease pathology. One critical aspect of hippocampal function is its role in episodic memory, which involves the binding of temporally coincident sensory information (e.g., sights, smells, and sounds) to create a representation of a specific learning epoch. Flexible associations can be formed among these distinct sensory stimuli that enable the "transfer" of new learning across a wide variety of contexts. The current studies employed a mouse analog of an associative "transfer learning" task that has previously been used to identify risk for prodromal AD in humans. The rodent version of the task assesses the transfer of learning about stimulus features relevant to a food reward across a series of compound discrimination problems. The relevant feature that predicts the food reward is unchanged across problems, but an irrelevant feature (i.e., the context) is altered. Experiment 1 demonstrated that C57BL6/J mice with bilateral ibotenic acid lesions of hippocampus were able to discriminate between two stimuli on par with control mice; however, lesioned mice were unable to transfer or apply this learning to new problem configurations. Experiment 2 used the APPswe PS1 mouse model of amyloidosis to show that robust impairments in transfer learning are evident in mice with subtle β-amyloid-induced synaptic deficits in the hippocampus. Finally, Experiment 3 confirmed that the same transfer learning impairments observed in APPswePS1 mice were also evident in the Tg-SwDI mouse, a second model of amyloidosis. Together, these data show that the ability to generalize learned associations to new contexts is disrupted even in the presence of subtle hippocampal dysfunction and suggest that, across species, this aspect of hippocampal-dependent learning may be useful for early identification of AD-like pathology.

Novel age-dependent learning deficits in a mouse model of Alzheimer's disease: implications for translational research

Computational modeling predicts that the hippocampus plays an important role in the ability to apply previously learned information to novel problems and situations (referred to as the ability to generalize information or simply as 'transfer learning'). These predictions have been tested in humans using a computer-based task on which individuals with hippocampal damage are able to learn a series of complex discriminations with two stimulus features (shape and color), but are impaired in their ability to transfer this information to newly configured problems in which one of the features is altered. This deficit occurs despite the fact that the feature predictive of the reward (the relevant information) is not changed. The goal of the current study was to develop a mouse analog of transfer learning and to determine if this new task was sensitive to pathological changes in a mouse model of AD. We describe a task in which mice were able to learn a series of concurrent discriminations that contained two stimulus features (odor and digging media) and could transfer this learned information to new problems in which the irrelevant feature in each discrimination pair was altered. Moreover, we report age-dependent deficits specific to transfer learning in APP+PS1 mice relative to non-transgenic littermates. The robust impairment in transfer learning may be more sensitive to AD-like pathology than traditional cognitive assessments in that no deficits were observed in the APP+PS1 mice on the widely used Morris water maze task. These data describe a novel and sensitive paradigm to evaluate mnemonic decline in AD mouse models that has unique translational advantages over standard species-specific cognitive assessments (e.g., water maze for rodent and delayed paragraph recall for humans).

[A simple cognitive test which detects early decline in cognitive function]

AIM

The aim of this study was to examine our new cognitive test which detects early decline in cognitive function.

METHODS

Our newly developed Simple Cognitive test (SC-test) takes 3 minutes, during which participants choose the figures in which both the shape and color are different from the figure on the left end of each line. The top score is 50 points. In a normal control study, the SC-test was given to 271 nursing home staff. We gave the Mini Mental State Examination (MMSE), Frontal Assessment Battery (FAB) and SC-test, to 114 participants in day-rehabilitation (Day group).

RESULTS

For the normal candidates, most young participants obtained nearly full marks. The average mark of those aged 18 to 29 was 48.1 (+/-3.40) points, while that for those 60 to 69 was 36.2 (+/-11.6) points. The average age of the Day group was 80.2 (+/-6.2), the MMSE average score was 23.8 (+/-4.3) points, the average FAB score was 11.2 (+/-3.5) points and the average on the SC-test was 13.1 (+/-11.5) points. The correlation coefficient between the SC-test and MMSE was r=0.569 (p=4.17x10(-11)), while that of the FAB was r=0.664 (p=6.661x10(-16)). In the 3-dimensional scatter chart of the three tests, despite near full marks on MMSE, people who obtained low scores on the FAB test also had low scores on the SC-test.

CONCLUSIONS

The SC-test is very simple to use, and it has very high sensitivity and specificity regarding early decline in cognitive function, especially frontal lobe function.

Cortico-hippocampal interaction and adaptive stimulus representation: a neurocomputational theory of associative learning and memory

Computational models of the hippocampal region link psychological theories of associative learning with their underlying physiological and anatomical substrates. Our approach to theory development began with a broad description of the computations that depend on the hippocampal region in classical conditioning (Gluck and Myers, 1993 and Gluck and Myers, 2001). In this initial model, the hippocampal region was treated as an Information-processing system that transformed stimulus representations, compressing (making more similar) representations of inputs that co-occur or are otherwise redundant, while differentiating (or making less similar) representations of inputs that predict different future events. This model led to novel predictions for the behavioral consequences of hippocampal-region lesions in rodents and of brain damage in humans who have amnesia or are in the earliest stages of Alzheimer's disease. Many of these predictions have, since been confirmed by our lab and others. Functional brain imaging studies have provided further supporting evidence. In more recent computational modeling, we have shown how some aspects of this proposed information-processing function could emerge from known anatomical and physiological characteristics of the hippocampal region, including the entorhinal cortex and the septo-hippocampal cholinergic system. The modeling to date lays the groundwork for future directions that increase the depth of detail of the biological modeling, as well as the breadth of behavioral phenomena addressed. In particular, we are working now to reconcile these kinds of incremental associative learning models with other models of the hippocampal region that account for the rapid formation of declarative memories.

Perirhinal cortex and anterior thalamic lesions: comparative effects on learning and memory

Three learning and memory tasks were used to compare the effects of neurotoxic anterior thalamic nuclei (ATN) and perirhinal cortex (PRC) lesions in rats. Rats with ATN lesions showed impaired spatial memory in a 12-arm radial maze, whereas rats with PRC lesions showed intact spatial memory, despite the use of minimal pretraining and extensive within-session delays (to 40 rain). PRC, but not ATN, lesions produced impairments on a configural learning task using complex visual-tactile cues in the radial maze. Neither ATN nor PRC lesions consistently affected spontaneous object recognition across extended sample-test delays (to 40 min). These findings confirm the differential involvement of the ATN and PRC in learning and memory.

Migration of objects and inferences across episodes

Participants viewed episodes in the form of a series of photographs portraying ordinary routines (e.g., eating at a restaurant) and later received a recognition test. In Experiment 1, it was shown that objects (e.g., a vase of flowers, a pewter lantern) that appeared in a single episode during the study phase migrated between memories of episodes described by the same abstract schema (e.g., from Restaurant Episode A at study to Restaurant Episode B at test), and not between episodes anchored by different schemas. In Experiment 2, it was demonstrated that backward causal inferences from one study episode influenced memories of other episodes described by the same schema, and that high-schema-relevant items viewed in one episode were sometimes remembered as having occurred in another episode of the same schematic type.

The organization of visual object representations: a connectionist model of effects of lesions in perirhinal cortex

We have developed a simple connectionist model based on the idea that perirhinal cortex has properties similar to other regions in the ventral visual stream, or 'what' pathway. The model is based on the assumption that representations in the ventral visual stream are organized hierarchically, such that representations of simple features of objects are stored in caudal regions of the ventral visual stream, and representations of the conjunctions of these features are stored in more rostral regions. We propose that a function of these feature conjunction representations is to help to resolve 'feature ambiguity', a property of visual discrimination problems that can emerge when features of an object predict a given outcome (e.g. reward) when part of one object, but predict a different outcome when part of another object. Several recently reported effects of lesions of perirhinal cortex in monkeys have provided key insights into the functions of this region. In the present study these effects were simulated by comparing the performance of connectionist networks before and after removal of a layer of units corresponding to perirhinal cortex. The results of these simulations suggest that effects of lesions in perirhinal cortex on visual discrimination may be due not to the impairment of a specific type of learning or memory, such as declarative or procedural, but to compromising the representations of visual stimuli. Furthermore, we propose that attempting to classify perirhinal cortex function as either 'perceptual' or 'mnemonic' may be misguided, as it seems unlikely that these broad constructs will map neatly onto anatomically defined regions of the brain.

Right medial temporal-lobe contribution to object-location memory

An important aspect of normal human memory, and one humans share with many other species, is the ability to remember the location of objects in their environment. There is by now strong evidence from the study of epileptic patients undergoing brain surgery that right temporal-lobe lesions that encroach extensively upon the hippocampal and parahippocampal gyrus impair the delayed, but not the immediate, recall of the location of objects within a random array. These findings have now been extended to a multiple-trial, spatial-array learning task; by including not only patients tested after unilateral anterior temporal lobectomy but also those with a selective left or right amygdalohippocampectomy, it has been shown that the deficits associated with right hippocampal lesions are not dependent upon conjoint damage to the lateral temporal neocortex. Furthermore, the fact that on the learning task no group differences were seen on Trial 1, at zero delay, strengthened the view that the impairment was in the maintenance and subsequent retrieval of information rather than in its initial encoding. These results left unresolved the question of whether the deficit was in the mediation of object-place associations or whether it could be reduced to a more general impairment in memory for location as such. Also left unanswered was the neuroanatomical question as to the relative contributions of the hippocampus and the parahippocampal gyrus to the performance of the experimental tasks. These questions were addressed in two blood-flow activation studies that made use of positron emission tomography (PET) and magnetic resonance imaging (MRI) and incorporated computerized versions of object-location and simple-location memory tasks. Taken together, the results point to a special contribution from the anterior part of the right parahippocampal gyrus, probably corresponding to the entorhinal cortex, to the retrieval of object-place associations, a result consonant with neurophysiological findings in non-human primates.

Extending models of hippocampal function in animal conditioning to human amnesia

Although most analyses of amnesia have focused on the loss of explicit declarative and episodic memories following hippocampal-region damage, considerable insights into amnesia can also be realised by studying hippocampal function in simple procedural, or habit-based, associative learning tasks. Although many simple forms of associative learning are unimpaired by hippocampal damage, more complex tasks which require sensitivity to unreinforced stimuli, configurations of multiple stimuli, or contextual information are impaired by hippocampal damage. In several recent papers we have developed a computational theory of hippocampal function which argues that this brain region plays a critical role in the formation of new stimulus representations during learning (Gluck & Myers, 1993, 1995; Myers & Gluck, 1996; Myers, Gluck, & Granger, 1995). We have applied this theory to a broad range of empirical data from studies of classical conditioning in both intact and hippocampal-lesioned animals, and the model correctly accounts for these data. The classical conditioning paradigm can be adapted for use in humans, and similar results for acquisition are obtained in both normal and hippocampal-damaged humans. More recently, we have begun to address an important set of category learning studies in both normals and hippocampal-damaged amnesics. This work integrates experimental studies of amnesic category learning (Knowlton, Squire, & Gluck, 1994) with theoretical accounts of associative learning, and builds on previously established behavioural correspondences between animal conditioning and human category learning (Gluck & Bower, 1988a). Our work to date illustrates some initial progress towards a more integrative understanding of hippocampal function in both animal and human learning, which may be useful in guiding further empirical and theoretical research in human memory and amnesia.

Psychobiological models of hippocampal function in learning and memory

We review current computational models of hippocampal function in learning and memory, concentrating on those that make strongest contact with psychological issues and behavioral data. Some models build upon Marr's early theories for modeling hippocampal field CA3's putative role in the fast, temporary storage of episodic memories. Other models focus on hippocampal involvement in incrementally learned associations, such as classical conditioning. More recent efforts have attempted to bring functional interpretations of the hippocampal region in closer contact with underlying anatomy and physiology. In reviewing these psychobiological models, three major themes emerge. First, computational models provide the conceptual glue to bind together data from multiple levels of analysis. Second, models serve as important tools to integrate data from both animal and human studies. Third, previous psychological models that capture important behavioral principles of memory provide an important top-down constraint for developing computational models of the neural bases of these behaviors.

Late-training amnesic deficits in probabilistic category learning: a neurocomputational analysis

Building upon earlier behavioral models of animal and human learning, we explore how a psychobiological model of animal conditioning can be applied to amnesic category learning. In particular, we show that the late-training deficit found in Knowlton, Squire, and Gluck's 1994 study of amnesic category learning can be understood as a natural consequence of Gluck and Myers's (1993) theory of hippocampal-region function, a theory that has heretofore been applied only to studies of animal learning. When applied to Knowlton et al.'s category learning task, Gluck and Myers's model assumes that the hippocampal region induces new stimulus representations over multiple training trials that reflect stimulus-stimulus regularities in the training set. As such, the model expects an advantage for control subjects over hippocampal-damaged amnesic patients only later in training when control subjects have developed new hippocampal-dependent stimulus representations; in contrast, both groups are expected to show equivalent performance early in training. A potentially analogous early/late distinction is described for animal studies of stimulus generalization. Our analyses suggest that careful comparisons between early and late-training differences in learning may be an important factor in understanding amnesia and the neural bases of both animal and human learning.

Integrating behavioral and physiological models of hippocampal function

In recent modeling of hippocampal function, we have attempted to integrate formal behavioral analyses of classical conditioning with psychobiological data on brain lesions (Gluck and Myers [1993] Hippocampus 3:491-516; Myers and Gluck [1994] Behav Neurosci 108(5):835-847). Based on comparative behavioral analyses, we have argued that animals with hippocampal region damage are unable to alter stimulus similarity based on experience. While hippocampal-damaged animals can still learn whether to respond to an individual stimulus, they are notably impaired at many tasks involving learning relationships between stimuli-especially in the absence of explicit reinforcement. These analyses lead to a computational theory which identifies two representational recoding processes-predictive differentiation and redundancy compression-which alter stimulus similarity relationships in intact animals but are dependent on intact hippocampal region processing. More recent, and ongoing, modeling aims to broaden this model of hippocampal region function in classical conditioning, with an emphasis on physiological and anatomical constraints, including the role of the fornix and subcortical modulation, preprocessing in sensory cortices, and localization of the proposed representational functions within more precisely identified hippocampal region substrates (Myers et al. [1995] Psychology 23(2):116-138; Myers and Gluck [1996] Behav Neurosci; Myers et al. [1996] Neurobiol Learning Memory). Working to bridge between behavioral and physiological levels of analysis, we ultimately hope to develop a more complete understanding of hippocampal region function in memory across a wider range of behavioral paradigms, elucidating how this functionality emerges from underlying physiological and anatomical substrates.

TraceLink: a model of amnesia and consolidation of memory

A model of amnesia is introduced, called TraceLink, that consists of three systems: 1) a trace system (neocortex), 2) a link system (hippocampus), and 3) a modulatory system (hippocampus/fornix/basal forebrain). It aims to explain salient aspects of the neuropsychology of amnesia, such as Ribot gradients in retrograde amnesia, patterns of dissociation between anterograde and retrograde amnesia, recovery from amnesia, and a newly discovered form of amnesia (semantic dementia) that results from certain temporal lobe lesions that do not affect the hippocampus. The model, furthermore, offers a new explanation for the global neuroanatomy of the hippocampus and neocortex based on the assumption that the brain aims to minimize connectivity volume. It also offers various strategies for the consolidation of memory, the effects of which are explored in computer simulations. The paper concludes with ten, largely untested; predictions derived from the TraceLink model.