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 neurocomputational model of tonic and phasic dopamine in action selection: a comparison with cognitive deficits in Parkinson's disease.

The striatal dopamine signal has multiple facets; tonic level, phasic rise and fall, and variation of the phasic rise/fall depending on the expectation of reward/punishment. We have developed a network model of the striatal direct pathway using an ionic current level model of the medium spiny neuron that incorporates currents sensitive to changes in the tonic level of dopamine. The model neurons in the network learn action selection based on a novel set of mathematical rules that incorporate the phasic change in the dopamine signal. This network model is capable of learning to perform a sequence learning task that in humans is thought to be dependent on the basal ganglia. When both tonic and phasic levels of dopamine are decreased, as would be expected in unmedicated Parkinson's disease (PD), the model reproduces the deficits seen in a human PD group off medication. When the tonic level is increased to normal, but with reduced phasic increases and decreases in response to reward and punishment, respectively, as would be expected in PD medicated with L-Dopa, the model again reproduces the human data. These findings support the view that the cognitive dysfunctions seen in Parkinson's disease are not solely either due to the decreased tonic level of dopamine or to the decreased responsiveness of the phasic dopamine signal to reward and punishment, but to a combination of the two factors that varies dependent on disease stage and medication status.

Probabilistic categorization: how do normal participants and amnesic patients do it?

In probabilistic categorization tasks, various cues are probabilistically (but not perfectly) predictive of class membership. This means that a given combination of cues sometimes belongs to one class and sometimes to another. It is not yet clear how categorizers approach such tasks. Here, we review evidence in favor of two alternative conceptualizations of learning in probabilistic categorization: as rule-based learning, or as incremental learning. Each conceptualization forms the basis of a way of analyzing performance: strategy analysis assumes rule-based learning, while rolling regression analysis assumes incremental learning. Here, we contrasted the ability of each to predict performance of normal categorizers. Both turned out to predict responses about equally well. We then reviewed performance of patients with damage to regions deemed important for either rule-based or incremental learning. Evidence was again about equally compatible with either alternative conceptualization of learning, although neither predicted an involvement of the medial temporal lobe. We suggest that a new way of conceptualizing probabilistic categorization might be fruitful, in which the medial temporal lobe help set up representations that are then used by other regions to assign patterns to categories.

Risk and protective haplotypes of the alpha-synuclein gene associated with Parkinson's disease differentially affect cognitive sequence learning.

Alpha-synuclein (SNCA) is a key factor in the regulation of dopaminergic transmission and is related to Parkinson's disease. In this study, we investigated the effects of risk and protective SNCA haplotypes associated with Parkinson's disease on cognitive sequence learning in 204 healthy volunteers. We found that the 3'-block risk SNCA haplotypes are associated with less effective stimulus-reward learning of sequences and with superior context representation of sequences. In contrast, participants with protective haplotypes exhibit better stimulus-reward learning and worse context representation, which suggest that these functions are inversely affected by risk and protective haplotypes. The Rep1 promoter polymorphism does not influence cognitive sequence learning. Because stimulus-reward learning may be mediated by the basal ganglia and context learning may be related to the medial temporal lobe, our data raise the possibility that dopaminergic signals regulated by SNCA inversely affect these memory systems.

Basal ganglia and dopamine contributions to probabilistic category learning.

Studies of the medial temporal lobe and basal ganglia memory systems have recently been extended towards understanding the neural systems contributing to category learning. The basal ganglia, in particular, have been linked to probabilistic category learning in humans. A separate parallel literature in systems neuroscience has emerged, indicating a role for the basal ganglia and related dopamine inputs in reward prediction and feedback processing. Here, we review behavioral, neuropsychological, functional neuroimaging, and computational studies of basal ganglia and dopamine contributions to learning in humans. Collectively, these studies implicate the basal ganglia in incremental, feedback-based learning that involves integrating information across multiple experiences. The medial temporal lobes, by contrast, contribute to rapid encoding of relations between stimuli and support flexible generalization of learning to novel contexts and stimuli. By breaking down our understanding of the cognitive and neural mechanisms contributing to different aspects of learning, recent studies are providing insight into how, and when, these different processes support learning, how they may interact with each other, and the consequence of different forms of learning for the representation of knowledge.

Dopaminergic contribution to cognitive sequence learning.

Evidence suggests that dopaminergic mechanisms in the basal ganglia are important in feedback-guided habit learning. To test hypothesis, we assessed cognitive sequence learning in 120 healthy volunteers and measured plasma levels of homovanillic acid [HVA] (a metabolite of dopamine), 5-hydroxyindoleacetic acid [5-HIAA] (a metabolite of serotonin), and 3-methoxy-4-hydroxypheylglycol [MHPG] (a metabolite of norepinephrine). Results revealed a significant negative relationship between errors in the feedback-guided training phase of the sequence learning task and the plasma HVA level. The HVA level accounted for 10.5% of variance of performance. Participant who had lower HVA level than the median value of the whole sample committed more errors during the training phase compared with participants who had higher HVA plasma level than the median value. A similar phenomenon was not observed for the context-dependent phase of the task and for 5-HIAA and MHPG. These results suggest that dopamine plays a special role in feedback-guided cognitive sequence learning.

Conditional discrimination and reversal in amnesia subsequent to hypoxic brain injury or anterior communicating artery aneurysm rupture.

Human anterograde amnesia can develop following bilateral damage to the hippocampus and medial temporal lobes, as in hypoxic brain injury, or following damage to the basal forebrain, as following anterior communicating artery (ACoA) aneurysm rupture. In both cases, the mnestic deficit may be similar when assessed by standard neuropsychological measures. However, animal and computational models suggest that there are qualitative differences in the pattern of impaired and spared memory abilities following damage to hippocampus versus basal forebrain. Here, we show such a dissociation in human amnesia using a single two-stage task, involving conditional discrimination and reversal. Consistent with a prior study, 10 individuals with anterograde amnesia subsequent to hypoxic brain injury were spared on acquisition but impaired at reversal. However, 10 individuals with amnesia subsequent to ACoA aneurysm showed the opposite pattern of impaired acquisition but spared reversal. The differences between groups cannot be easily ascribed to severity of mnestic or cognitive deficit, since the two amnesic groups performed similarly on neuropsychological tests of memory, intelligence and attention. The results illustrate qualitative differences in memory impairments in hypoxic and ACoA amnesics and highlight the importance of considering etiology in evaluating mnemonic deficits in amnesic populations.

Integrating incremental learning and episodic memory models of the hippocampal region.

By integrating previous computational models of corticohippocampal function, the authors develop and test a unified theory of the neural substrates of familiarity, recollection, and classical conditioning. This approach integrates models from 2 traditions of hippocampal modeling, those of episodic memory and incremental learning, by drawing on an earlier mathematical model of conditioning, SOP (A. Wagner, 1981). The model describes how a familiarity signal may arise from parahippocampal cortices, giving a novel explanation for the finding that the neural response to a stimulus in these regions decreases with increasing stimulus familiarity. Recollection is ascribed to the hippocampus proper. It is shown how the properties of episodic representations in the neocortex, parahippocampal gyrus, and hippocampus proper may explain phenomena in classical conditioning. The model reproduces the effects of hippocampal, septal, and broad hippocampal region lesions on contextual modulation of classical conditioning, blocking, learned irrelevance, and latent inhibition.

Role of the basal ganglia in category learning: how do patients with Parkinson's disease learn?

The purpose of the present study was to gain a deeper understanding of the role of the basal ganglia in learning and memory by examining learning strategies among patients with basal ganglia dysfunction. Using a probabilistic category learning task (the "weather prediction" task) previously shown to be sensitive to basal ganglia function, the authors examined patterns of performance during learning and used mathematical models to capture different learning strategies. Results showed that patients with Parkinson's disease exhibit different patterns of strategy use. Specifically, most controls initially used a simple, but suboptimal, strategy that focused on single-cue-outcome associations; eventually, however, most controls adopted a more complex, optimal learning strategy, integrating single-cue associations to predict outcomes for multiple-cue stimuli. In contrast, the majority of individuals with Parkinson's disease continued to rely on simple single-cue learning strategies throughout the experiment.

Cortico-striatal contributions to feedback-based learning: converging data from neuroimaging and neuropsychology.

The striatum has been widely implicated in cognition, but a precise understanding of its role remains elusive. Here we present converging evidence for the role of the striatum in feedback-based learning. In a prior functional imaging study, healthy controls showed striatal activity during a feedback-based learning task, which was decreased when the same task was learned without feedback. In the present study, we show that individuals with striatal dysfunction due to Parkinson's disease are impaired on the feedback-based task, but not on a non-feedback version of the same task. Parkinson's patients and controls also used different learning strategies depending on feedback structure. This study provides direct behavioural evidence from humans that cortico-striatal systems are necessary for feedback-based learning on a cognitive task. These findings also link between learning impairments in Parkinson's disease and the physiological and computational evidence for the role of midbrain dopaminergic systems in feedback processing.

Human midbrain sensitivity to cognitive feedback and uncertainty during classification learning.

Mesencephalic dopaminergic system (MDS) neurons may participate in learning by providing a prediction error signal to their targets, which include ventral striatal, orbital, and medial frontal regions, as well as by showing sensitivity to the degree of uncertainty associated with individual stimuli. We investigated the mechanisms of probabilistic classification learning in humans using functional magnetic resonance imaging to examine the effects of feedback and uncertainty. The design was optimized for separating neural responses to stimulus, delay, and negative and positive feedback components. Compared with fixation, stimulus and feedback activated brain regions consistent with the MDS, whereas the delay period did not. Midbrain activity was significantly different for negative versus positive feedback (consistent with coding of the "prediction error") and was reliably correlated with the degree of uncertainty as well as with activity in MDS target regions. Purely cognitive feedback apparently engages the same regions as rewarding stimuli, consistent with a broader characterization of this network.

Interactive memory systems in the human brain.

Learning and memory in humans rely upon several memory systems, which appear to have dissociable brain substrates. A fundamental question concerns whether, and how, these memory systems interact. Here we show using functional magnetic resonance imaging (FMRI) that these memory systems may compete with each other during classification learning in humans. The medial temporal lobe and basal ganglia were differently engaged across subjects during classification learning depending upon whether the task emphasized declarative or nondeclarative memory, even when the to-be-learned material and the level of performance did not differ. Consistent with competition between memory systems suggested by animal studies and neuroimaging, activity in these regions was negatively correlated across individuals. Further examination of classification learning using event-related FMRI showed rapid modulation of activity in these regions at the beginning of learning, suggesting that subjects relied upon the medial temporal lobe early in learning. However, this dependence rapidly declined with training, as predicted by previous computational models of associative learning.

Cerebellar substrates for error correction in motor conditioning.

The authors evaluate a mapping of Rescorla and Wagner's (1972) behavioral model of classical conditioning onto the cerebellar substrates for motor reflex learning and illustrate how the limitations of the Rescorla-Wagner model are just as useful as its successes for guiding the development of new psychobiological theories of learning. They postulate that the inhibitory pathway that returns conditioned response information from the cerebellar interpositus nucleus back to the inferior olive is the neural basis for the error correction learning proposed by Rescorla and Wagner (Gluck, Myers, & Thompson, 1994; Thompson, 1986). The authors' cerebellar model expects that behavioral processes described by the Rescorla-Wagner model will be localized within the cerebellum and related brain stem structures, whereas behavioral processes beyond the scope of the Rescorla-Wagner model will depend on extracerebellar structures such as the hippocampus and related cortical regions. Simulations presented here support both implications. Several novel implications of the authors' cerebellar error-correcting model are described including a recent empirical study by Kim, Krupa, and Thompson (1998), who verified that suppressing the putative error correction pathway should interfere with the Kamin (1969) blocking effect, a behavioral manifestation of error correction learning. The authors also discuss the model's implications for understanding the limits of cerebellar contributions to associative learning and how this informs our understanding of hippocampal function in conditioning. This leads to a more integrative view of the neural substrates of conditioning in which the authors' real-time circuit-level model of the cerebellum can be viewed as a generalization of the long-term memory module of Gluck and Myers' (1993) trial-level theory of cerebellar-hippocampal interaction in motor conditioning.

Spectrotemporal sensitivities in rat auditory cortical neurons.

Studies in several mammalian species have demonstrated that auditory cortical neurons respond strongly to single frequency-modulated (FM) sweeps, and that most responses are selective for sweep direction and/or rate. In the present study, we used extracellular recordings to examine how neurons in the auditory cortices of anesthetized rats respond to continuous, periodic trains of FM sweeps (described previously by deCharms et al., Science 280 (1998) pp. 1439-1444, as moving auditory gratings). Consistent with previous observations in owl monkeys, we found that the majority of cortical neurons responded selectively to trains of either up-sweeps or down-sweeps; selectivity for down-sweeps was most common. Periodic responses were typically evoked only by sweep trains with repetition rates less than 12 sweeps per second. Directional differences in responses were dependent on repetition rate. Our results support the proposal that a combination of both spectral and temporal acoustic features determines the responses of auditory cortical neurons to sound, and add to the growing body of evidence indicating that the traditional view of the auditory cortex as a frequency analyzer is not sufficient to explain how the mammalian brain represents complex sounds.

Parallel neural systems for classical conditioning: support from computational modeling.

Classical conditioning has been explained by two main types of theories that postulate different learning mechanisms. Rescorla and Wagner (1972) put forth a theory in which conditioning is based on the ability of the US to drive learning through error correction. Alternatively, Mackintosh (1973) put forth a theory in which the ability of the CS to be associated with the unconditioned stimulus is modulated. We have proposed a reconciliation of these two mechanisms as working in parallel within different neural systems: a cerebellar system for US modulation and a hippocampal system for CS modulation. We developed a computational model of cerebellar function in eyeblink conditioning based on the error correction mechanism of the Rescorla-Wagner rule in which learning-related activity from the cerebellum inhibits the inferior olive, which is the US input pathway to the cerebellum (Gluck et al., 1994). We developed a computational model of the hippocampal region that forms altered representations of conditioned stimuli based on their behavioral outcomes (Gluck & Myers, 1993; Myers et al., 1995). Overall, computational modeling and empirical findings support the idea that, at least in the case of eyeblink conditioning, there may be two different neural systems: the cerebellum which mediates US-based error correction and hippocampus which alters representations of CSs.

Impaired delay eyeblink classical conditioning in individuals with anterograde amnesia resulting from anterior communicating artery aneurysm rupture.

Anterior communicating artery (ACoA) aneurysm rupture can lead to an anterograde amnesia syndrome similar to that observed after damage to the hippocampus and medial temporal lobes (MT). It is currently believed that ACoA amnesia results from basal forebrain damage that disrupts hippocampal processing without direct hippocampal damage. Converging evidence from animal studies and computational modeling suggests that qualitative differences may exist in the pattern of memory impairment after basal forebrain or MT damage. For example, animals with basal forebrain but not hippocampal damage are impaired at delay eyeblink classical conditioning (EBCC). In this study, individuals with ACoA amnesia were shown to be impaired at delay EBCC compared with matched controls; this contrasts with the spared delay EBCC previously observed in MT amnesia. This finding suggests the beginning of a possible dissociation between the memory impairments in MT versus ACoA amnesia.

Basal forebrain stimulation changes cortical sensitivities to complex sound.

Experience affects how brains respond to sound. Here, we examined how the sensitivity and selectivity of auditory cortical neuronal responses were affected in adult rats by the repeated presentation of a complex sound that was paired with basal forebrain stimulation. The auditory cortical region that was responsive to complex sound was 2-5 five times greater in area in paired-stimulation rats than in naive rats. Magnitudes of neuronal responses evoked by complex sounds were also greatly increased by associative pairing, as were the percentages of neurons that responded selectively to the specific spectrotemporal features that were paired with stimulation. These findings demonstrate that feature selectivity within the auditory cortex can be flexibly altered in adult mammals through appropriate intensive training.

A computational model of mechanisms controlling experience-dependent reorganization of representational maps in auditory cortex.

Cortical representations of sound can be modified by repeatedly pairing presentation of a pure tone with electrical stimulation of neuromodulatory neurons located in the basal forebrain (Bakin & Weinberger, 1996; Kilgard & Merzenich, 1998a). We developed a computational model to investigate the possible effects of basal forebrain modulation on map reorganization in the auditory cortex. The model is a self-organizing map with acoustic response characteristics mimicking those observed in the mammalian auditory cortex. We simulated the effects of basal forebrain modulation, using parameters intrinsic to the self-organizing map, such as the learning rate (controlling the adaptability of map nodes) and the neighborhood function (controlling the excitability of map nodes). Previous research has suggested that both parameters can be useful for characterizing the effects of neuromodulation on plasticity (Kohonen, 1993; Myers et al., 1996; Myers, Ermita, Hasselmo, & Gluck, 1998). The model successfully accounts for experimentally observed effects of pairing basal forebrain stimulation with the presentation of a single tone, but not of two tones, suggesting that auditory cortical plasticity is constrained in ways not accounted for by current theories. Despite this limitation, the model provides a useful framework for describing experience-induced changes in auditory representations and for relating such changes to variations in the excitability and adaptability of cortical neurons produced by neuromodulation.

Dissociating entorhinal and hippocampal involvement in latent inhibition.

This study used anatomical cues to suggest a functional dissociation between the roles of the entorhinal cortex and the hippocampus in learning. The authors proposed that the highly convergent inputs to the entorhinal cortex indicate this region may be particularly important for selecting or compressing information. This hypothesis was tested in rabbits (Oryctolagus cunniculus) trained on an associative learning task that is a common index of stimulus selection. In this task, known as latent inhibition, preexposure to a stimulus (such as a tone) leads to slowed learning when the same tone is subsequently paired with an outcome (such as an airpuff to the eye). As hypothesized, rabbits with neurotoxic lesions of the entorhinal cortex failed to show slowed learning following preexposure (no latent inhibition) and learned the association faster than control rabbits. In contrast, hippocampal-lesioned animals showed normal (slowed) learning.

Stimulus exposure effects in human associative learning.

Learning that one cue (CS) predicts a second, salient cue (US) can often be slowed by prior exposure to one or both stimuli. In animals, CS-US learning is more strongly retarded following uncorrelated exposure to both CS and US than following exposure to the US alone. In this paper we present several studies showing a similar effect in humans, using a computer-based task. Experiments 1 and 2 used a between-groups design and demonstrated a strong CS/US exposure effect, whether or not the US was signalled by a neutral cue during exposure. Experiment 3 demonstrated similar effects using a within-subjects design. Overall, these results are consistent with several theoretical interpretations and suggest that uncorrelated CS/US exposure leads to a robust retardation of subsequent CS-US learning in humans.